TSpace Research Repository 
tspace.library.utoronto.ca 
Sensitive tumor detection and classification 
using plasma cell-free DNA methylomes 
 Shu Yi Shen, Rajat Singhania, Gordon Fehringer, Ankur 
Chakravarthy, Michael H. A. Roehrl, Dianne Chadwick, Philip C. 
Zuzarte, Ayelet Borgida, Ting Ting Wang, Tiantian Li, Olena Kis, 
Zhen Zhao, Anna Spreafico, Tiago da Silva Medina, Yadon  
Wang, David Roulois, Ilias Ettayebi, Zhuo Chen, Signy Chow, 
Tracy Murphy, Andrea Arruda, Grainne M. O’Kane, Jessica Liu, 
Mark Mansour, John D. McPherson, Catherine O’Brien, Natasha 
Leighl, Philippe L. Bedard, Neil Fleshner, Geoffrey Liu, Mark D. 
Minden, Steven Gallinger, Anna Goldenberg, Trevor J. Pugh, 
Michael M.  Hoffman, Scott V. Bratman, Rayjean J. Hung, 
Daniel D. De Carvalho
Version Post-print/accepted manuscript 
Citation 
(published version) 
Shen, S.Y., Singhania, R., Fehringer, G. et al. Sensitive tumour detection 
and classification using plasma cell-free DNA 
methylomes. Nature 563, 579–583 (2018). 
https://doi.org/10.1038/s41586-018-0703-0 
How to cite TSpace items 
Always cite the published version, so the author(s) will receive recognition through services that track 
citation counts, e.g. Scopus. If you need to cite the page number of the author manuscript from TSpace 
because you cannot access the published version, then cite the TSpace version in addition to the published 
version using the permanent URI (handle) found on the record page. 
This article was made openly accessible by U of T Faculty.  
Please tell us how this access benefits you. Your story matters. 


 
 
 
 
 
 
 
 
 
 
 
1 
Sensitive tumor detection and classification using plasma cell-free 
2 
DNA methylomes 
3 
4 
Shu Yi Shen1*, Rajat Singhania1*, Gordon Fehringer2*, Ankur Chakravarthy1*, Michael H. 
5 
A. Roehrl1,3,4, Dianne Chadwick1, Philip C. Zuzarte5, Ayelet Borgida2, Ting Ting Wang1,4, 
6 
Tiantian Li1, Olena Kis1, Zhen Zhao1, Anna Spreafico1, Tiago da Silva Medina1, Yadon 
7 
Wang1, David Roulois1,6, Ilias Ettayebi1,4, Zhuo Chen1, Signy Chow1, Tracy Murphy1, 
8 
Andrea Arruda1, Grainne M. O’Kane1, Jessica Liu4, Mark Mansour4, John D. McPherson7, 
9 
Catherine O’Brien1, Natasha Leighl1, Philippe L. Bedard1, Neil Fleshner1, Geoffrey Liu1,4,8, 
10 
Mark D. Minden1, Steven Gallinger9,10, Anna Goldenberg11, Trevor J. Pugh1,4, Michael M. 
11 
Hoffman1,4,11, Scott V. Bratman1,4, Rayjean J. Hung2,8§, Daniel D. De Carvalho1,4,12§ 
12 
13 
1. Princess Margaret Cancer Centre, University Health Network, Toronto, Canada 
14 
2. Lunenfeld-Tanenbaum Research Institute, Sinai Health System, Toronto, Canada 
15 
3. Memorial Sloan Kettering Cancer Center, New York, NY 
16 
4. Department of Medical Biophysics, University of Toronto, Toronto, Canada 
17 
5. Genome Technologies, Ontario Institute for Cancer Research, Toronto, Canada 
18 
6. UMR_S 1236, Univ Rennes 1, Inserm, Etablissement Français du sang Bretagne, F­
19 
35000 Rennes, France 
20 
7. Department of Biochemistry and Molecular Medicine, UC Davis Comprehensive Cancer 
21 
Center, Sacramento, CA, USA 


 
 
 
 
 
 
 
 
 
 
 
 
 
 
22 
23 
24 
25 
26 
27 
28 
8. Division of Epidemiology, Dalla Lana School of Public Health, University of Toronto, 
Toronto, Canada 
9. Fred Litwin Centre for Cancer Genetics, Lunenfeld-Tanenbaum Research Institute, 
Mount Sinai Hospital, Toronto, Canada 
10. Department of Surgery, Toronto General Hospital, Toronto, Ontario, Canada 
11. Department of Computer Science, University of Toronto, Toronto, Canada 
12. Lead Contact 
29 
30 
31 
32 
33 
34 
35 
*These authors made equal contributions 
§Corresponding authors: 
Daniel D. De Carvalho: ddecarv@uhnresearch.ca 
Rayjean J. Hung: rayjean.hung@lunenfeld.ca 
36 
37 
38 
39 


 
 
 
 
40 
45 
50 
55 
60 
The use of liquid biopsies for cancer detection and management is rapidly gaining 
41 
prominence1. Current circulating tumor DNA (ctDNA) detection methods involve 
42 
sequencing somatic mutations using cell-free DNA (cfDNA), but their sensitivity may be 
43 
low among early-stage cancer patients given the limited availability of recurrent 
44 
mutations2-5. In contrast, large-scale epigenetic alterations, which are tissue and cancer-
type specific are not similarly constrained6, thus potentially have enhanced ability to detect 
46 
and classify cancers in early-stage patients. Here, we developed a sensitive, 
47 
immunoprecipitation-based protocol for methylome analysis of small quantities of 
48 
circulating cfDNA and demonstrated the ability to detect large-scale DNA methylation 
49 
changes that are enriched for tumor-specific patterns. We also demonstrated robust 
performance in cancer detection and classification across an extensive collection of plasma 
51 
samples from multiple tumor types, setting the stage for the development of minimally 
52 
invasive biomarkers for early cancer detection, interception, and classification based on 
53 
plasma cfDNA methylation patterns. 
54 
Analysis of ctDNA has numerous potential clinical applications, but certain settings such as 
cancer screening and detection of minimal residual disease after treatment require a degree of 
56 
analytical sensitivity that is often beyond current technical limits of mutation-based ctDNA 
57 
detection methods. The major hurdles to improved sensitivity of these methods include (1) the 
58 
limited number of recurrent mutations available to distinguish between tumor and normal 
59 
circulating cfDNA in a cost-effective manner, and (2) technical artefacts (errors) introduced 
during sequencing. We reasoned that specific enrichment of methylated DNA fragments from 
61 
cfDNA could overcome both of these hurdles. 


 
 
 
62 
To assess whether the higher number of DNA methylation changes in cancers could translate to 
63 
increased sensitivity at lower sequencing costs, we performed bioinformatic simulations 
64 
examining detection probability across varying numbers of Differentially Methylated Regions 
65 
(DMRs), coverage, and ctDNA abundance (Fig. 1a and Extended Data Fig. 1a). We found 
66 
improved sensitivity as the number of DMRs increased, even at lower sequencing depth and 
67 
ctDNA abundance, suggesting the recovery of cancer-specific DNA methylation changes could 
68 
allow for highly sensitive and low-cost detection, classification and monitoring of cancer. 
69 
However, this is challenging in practice due to the low-abundance and fragmented nature of 
70 
plasma cfDNA3, which has restricted most of the previous plasma methylation profiling to locus­
71 
specific PCR-based assays7-9. While Whole-Genome Bisulfite Sequencing (WGBS) of cfDNA 
72 
has been attempted10,11, this approach is inefficient due to degradation of ~84-96% of the input 
73 
DNA during bisulfite conversion12, high-costs, and limited information recovery given the low 
74 
genome-wide abundance of CpGs. Therefore, we developed cfMeDIP-seq (cell-free Methylated 
75 
DNA Immunoprecipitation and high-throughput sequencing) for genome-wide bisulfite-free 
76 
plasma DNA methylation profiling, based on its ability to enrich for CpG rich, potentially more­
77 
informative fragments, thus enhancing cost-effectiveness. 
78 
Briefly, we optimized an existing low-input MeDIP-seq protocol13 that is robust down to 100 ng 
79 
input DNA, using exogenous Enterobacteria phage λ DNA (filler DNA) to inflate starting 
80 
amounts (Extended Data Fig. 1b). This is crucial for applications based on plasma cfDNA 
81 
samples, which yield much less than 100 ng of cfDNA. We then performed extensive 
82 
benchmarking of the optimized protocol. First, comparing low-input cfMeDIP-seq versus gold­
83 
standard MeDIP-seq using colorectal cancer (CRC) HCT116 DNA sheared to mimic cfDNA 


 
 
 
 
85 
90 
95 
100 
105 
84 
showed robust CpG-enrichment (Extended Data Fig. 2a-c) and inter-replicate correlation 
(Extended Data Fig. 2d). cfMeDIP-seq (1 to 10 ng input DNA) also recapitulated profiles from 
86 
gold-standard MeDIP (100 ng), RRBS (Reduced Representation Bisulfite Sequencing) (1,000 ng) 
87 
and WGBS (2,000 ng) (Extended Data Fig. 2e). 
88 
Next, cfMeDIP-seq was compared to ultra-deep, unique molecular identifiers (UMIs) based, 
89 
hybrid capture mutation sequencing14 across a serial dilution of CRC DNA into multiple 
myeloma (MM) MM.1S cell-line DNA (Extended Data Fig. 3a). The former showed near-perfect 
91 
linear associations between observed and expected numbers of DMRs (5% False Discovery Rate 
92 
(FDR) threshold) and signals within DMRs, down to 0.001% dilution (both r2=0.99, p < 0.0001) 
93 
(Fig. 1b and Extended Data Fig. 3b-e). The latter, however, detected CRC specific mutations 
94 
only down to 0.1% and 1% with single strand consensus sequence (SSCS) and duplex consensus 
sequence (DCS) respectively (Extended Data Fig. 3f-g). This highlights the excellent analytical 
96 
sensitivity of cfMeDIP-seq for the detection of cancer-derived DNA. We also evaluated the 
97 
ability of cfMeDIP-seq to enrich ctDNA through biased sequencing of CpG-rich sequences that 
98 
are frequently hypermethylated in cancer when compared to normal tissue15. Plasma from mice 
99 
harboring patient-derived xenografts (PDX) was used for cfMeDIP-seq, and a 2-fold enrichment 
of human tumor-derived cfDNA was found following immunoprecipitation as compared to the 
101 
input sample (Fig. 1c).  
102 
To investigate whether cfMeDIP-seq could detect ctDNA in early-stage cancer, we generated 
103 
cfMeDIP-seq profiles from pre-surgery plasma cfDNA of 24 primary early-stage pancreatic 
104 
cancer (PDAC) patients (cases) and 24 age and sex-matched healthy controls (controls) (Fig. 2a 
and Extended Data Fig. 4a-f). In addition to plasma cfDNA, the microdissected primary tumors 


 
 
106 
and adjacent normal tissue from the same PDAC patients were used to generate DNA 
107 
methylation profiles via RRBS. We identified 14,716 DMRs between cases and controls cfDNA 
108 
(9,931 hypermethylated in cases, 4,785 in controls, based on negative-binomial generalized 
109 
linear model (GLM) of fragment counts at significance level of Benjamini Hochberg FDR 
110 
(BHFDR) of 0.1) (Fig. 2b-c and Supplementary Table 1).  
111 
In comparison, 45,173 Differentially Methylated CpGs (DMCs) were found between tumor and 
112 
normal tissue in RRBS data (Supplementary Table 2). Permutation testing to estimate the 
113 
significance of overlaps between cfMeDIP-seq cell-free DMRs and RRBS tissue DMCs revealed 
114 
significant enrichment for DMR/DMC pairs concordantly hypermethylated (p=3.39e-47), and 
115 
concordantly hypomethylated in case cfDNA and tumor tissue (p=1.43e-22). This significant 
116 
enrichment was not observed in the discordant methylation pattern between cfDNA and tumor 
117 
DNA (Fig. 2d). Furthermore, signals in overlapping plasma cfDNA and tissue DNA methylation 
118 
were correlated (Extended Data Fig. 5a). These findings suggested that cfMeDIP-seq of plasma 
119 
cfDNA could detect tumor-derived DNA methylation events in ctDNA.  
120 
As non-tumor derived cfDNA is mostly released from blood cells, we performed similar 
121 
permutation-based enrichment testing between case-vs-control cfMeDIP-seq DMRs and the 
122 
95,388 RRBS DMCs between PDAC tumor tissue (n=24) and normal peripheral blood 
123 
mononuclear cell (PBMCs) (n=5) (Supplementary Table 3). Again, we observed significant 
124 
enrichment for concordant hypermethylated (p<1e-745) and hypomethylated (p=6.12e-82) sites 
125 
in cfMeDIP-seq DMRs and tumor vs. PBMC DMCs, while discordant calls were 
126 
underrepresented (Fig. 2e). In addition, signals in overlapping DMRs/DMCs were correlated 


 
 
 
127 
(Extended Data Fig. 5b) and altogether indicated that DMRs identified using cfMeDIP-seq, 
128 
between cases and controls, were likely derived from ctDNA (Extended Data Fig. 5c). 
129 
Based on the enrichment of tumor-derived DMRs and the known methylation-specific variable 
130 
binding of transcription factors (TFs)16, we hypothesized that cfMeDIP-seq methylomes could 
131 
identify active transcriptional networks in tumors or other tissues using plasma cfDNA. Upon 
132 
motif enrichment analysis on cfMeDIP-seq DMRs and incorporating methylation preferences of 
133 
candidate transcription factors into account16, we identified 42 TFs as binding in healthy controls 
134 
and 52 as binding in pancreatic cancer cases (Supplementary Table 4-5). As expected, the former 
135 
included hematopoietic-lineage specific TFs such as PU.1, NFE2, and GATA1, while the latter 
136 
included pancreas-associated TFs, PTF1a, Onecut1 (HNF6), and NR5A2 (Extended Data Fig. 6a 
137 
and c). Compared to random sets of TFs, TFs inferred as active in healthy controls are 
138 
overexpressed in blood based on GTEx data, while those inferred as active in pancreatic cancer 
139 
cases are overexpressed in pancreatic tissues in GTEx and PDAC tissue in TCGA (Extended 
140 
Data Fig. 6b, d, and e). Collectively, these findings indicated that cfMeDIP-seq might permit 
141 
non-invasive characterization of active TF-networks in cancer.  
142 
Given that we could detect tumor-specific DMRs in the plasma of PDAC cases relative to 
143 
controls, we then investigated whether cfMeDIP-seq could non-invasively classify multiple 
144 
cancer types from healthy controls. Consequently, we performed cfMeDIP-seq in a discovery 
145 
cohort of 189 plasma samples from 7 different tumor types (PDAC, colorectal cancer (CRC), 
146 
breast cancer (BRCA), lung cancer (LUC), renal cancer (RCC), bladder cancer (BLCA), and 
147 
AML) and healthy controls (Extended Data Fig. 7a-l and Extended Data Fig. 8a).  


 
 
 
 
 
148 
149 
150 
151 
152 
153 
154 
155 
156 
157 
158 
159 
160 
161 
162 
163 
164 
165 
166 
167 
168 
169 
170 
We first identified plasma cell-free DMRs for each tumor type relative to healthy controls and 
asked if these cancer type-specific DMRs identified on the plasma cfDNA were enriched for the 
expected tumor DMRs for each cancer type using tumor tissue methylation data from TCGA 
(n=4032) (Fig. 3a). We observed a marked enrichment of sites hypermethylated in the primary 
tumor tissue (TCGA), within the regions we identified as hypermethylated in the plasma cfDNA 
for each cancer type, coupled with significantly correlated signals between cfMeDIP-seq plasma 
methylation and TCGA 450k tumor data (Extended Data Fig. 8b-h). These results indicate the 
ability to recover ctDNA-associated methylation profiles across a range of cancer types. 
Finally, we carried out a set of machine learning analyses on our discovery cohort to rigorously 
evaluate the utility of cfMeDIP profiles in cancer detection and classification. We initially 
reduced our dataset to 505,027 windows mapping to CpG islands, shores, shelves and 
FANTOM5 enhancers for computational efficiency. Unbiased performance estimates, while 
accounting for training-set biases, were then derived from the reduced dataset. We split the 
discovery cohort 80%-20% into balanced training and test sets. Using only training set samples, 
we selected the top 300 DMRs by limma-trend test statistic for each class versus other classes. 
We then trained a series of one-vs-other-classes regularised binomial GLMs using these features 
on the training set data. The training procedure consisted of 3 rounds of 10-Fold Cross-
Validation (CV) across a grid of values for alpha and lambda with optimisation for Cohen’s 
Kappa. The use of multiple rounds of 10-Fold CV was motivated by a desire to leverage 
additional randomisation for more generalisable model tuning. 
The performance of these classifiers was then evaluated using receiver operating characteristic 
(ROC) statistics derived from the test-set samples that were not used for either DMR selection or 
model training. The whole process was repeated 100 times to prevent training-set biases17, 


 
 
 
 
171 
172 
173 
174 
175 
176 
177 
178 
179 
180 
181 
182 
183 
184 
185 
186 
187 
188 
189 
190 
191 
192 
193 
culminating in a collection of 
800 models, with 100 models for each one-vs-all-others 
comparison (hereby termed E100). High area under the receiver operator characteristic curve 
(AUROC) values were observed for test set samples across classes (Fig. 3b and Extended Data 
Fig. 9a). 
Subsequently, we assessed performance across batches by applying the ensemble to a 199­
sample validation cohort (35 AMLs, 47 PDACs, 55 LUC and 62 healthy controls). Averaging 
the class probabilities output by E100 for each sample, yielded high ROCs for AML vs. others 
(0.980), PDAC vs. others (0.918), LUC vs. others (0.971) and normal vs. others (0.969) (Fig. 3c). 
Notably, performance was similar between early and late stage samples, suggesting applicability 
to cancer early detection (Fig. 3d and Extended Data Fig. 9b). 
We then investigated whether the DMRs (non-zero coefficients) selected during the training of 
E100 were tumor-specific. Visualization using t-distributed stochastic neighbour embedding 
(tSNE) plots showed clear separation by tumor type in the plasma cohort (Fig. 4a). This was 
notably reproduced in the 450k dataset of 4,032 TCGA cancers and normal blood samples and 
400 COSMIC cancer cell lines and PBMCs (Fig. 4b-c). This suggests that our plasma cfDNA 
methylation classifiers are mainly driven by tumor-specific DNA methylation patterns rather 
than fluctuations in blood cells or cell composition on the tumor microenvironment.  
However, these results do not rule out that some plasma cell-free DMRs could originate from 
shifts in proportions of circulating immune cells18,19. To further test our inference, we identified 
38,352 cfMeDIP-windows lowly methylated across a range of leukocyte types in IHEC WGBS 
data, of which 27,088 overlapped with TCGA 450k data (Extended Data Fig. 10a). Out of these 
27,088 regions, we separated those that were identified as hypermethylated through the 
comparisons of plasma cfDNA of each cancer type to healthy controls. We then checked the 


 
 
 
 
 
 
 
 
194 
methylation status of these regions in the tumor tissue compared to PBMCs, using TCGA data 
195 
for each cancer type. For PDAC, we used in house methylation data generated for the matched 
196 
patients (cfDNA and tissue DNA). Indeed, we found these regions to be hypermethylated in 
197 
tumor tissue (Extended Data Fig. 10b), reinforcing the hypothesis that these plasma cell-free 
198 
DMRs are a direct measurement of tumor-derived DNA (i.e., ctDNA). 
199 
In summary, we developed a robust, sensitive and bisulfite-free methodology for 
200 
immunoprecipitation-based profiling of methylation patterns in cfDNA. Our approach awaits 
201 
further validation in completely independent datasets, but our findings underscore the potential 
202 
utility of cfDNA methylation profiles as a basis for non-invasive, cost-effective, sensitive and 
203 
accurate early tumor detection for cancer interception, and for multi-cancer classification. 
204 
Reference 
205 
1 
Diaz, L. A., Jr. & Bardelli, A. Liquid biopsies: genotyping circulating tumor DNA. J Clin 
206 
Oncol 32, 579-586, doi:10.1200/JCO.2012.45.2011 (2014). 
207 
2 
Aravanis, A. M., Lee, M. & Klausner, R. D. Next-Generation Sequencing of Circulating 
208 
Tumor DNA for Early Cancer Detection. Cell 168, 571-574, 
209 
doi:10.1016/j.cell.2017.01.030 (2017). 
210 
3 
Newman, A. M. et al. An ultrasensitive method for quantitating circulating tumor 
211 
DNA with broad patient coverage. Nat Med 20, 548-554, doi:10.1038/nm.3519 
212 
(2014). 
213 
4 
Cohen, J. D. et al. Detection and localization of surgically resectable cancers with a 
214 
multi-analyte blood test. Science, doi:10.1126/science.aar3247 (2018). 
215 
5 
Phallen, J. et al. Direct detection of early-stage cancers using circulating tumor DNA. 
216 
Sci Transl Med 9, doi:10.1126/scitranslmed.aan2415 (2017). 
217 
6 
Hoadley, K. A. et al. Multiplatform analysis of 12 cancer types reveals molecular 
218 
classification within and across tissues of origin. Cell 158, 929-944, 
219 
doi:10.1016/j.cell.2014.06.049 (2014). 
220 
7 
Lehmann-Werman, R. et al. Identification of tissue-specific cell death using 
221 
methylation patterns of circulating DNA. Proc Natl Acad Sci U S A 113, E1826-1834, 
222 
doi:10.1073/pnas.1519286113 (2016). 
223 
8 
Visvanathan, K. et al. Monitoring of Serum DNA Methylation as an Early Independent 
224 
Marker of Response and Survival in Metastatic Breast Cancer: TBCRC 005 
225 
Prospective Biomarker Study. J Clin Oncol, JCO2015662080 (2016). 


 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
230 
235 
240 
245 
250 
255 
226 
227 
228 
229 
9 
10 
Potter, N. T. et al. Validation of a real-time PCR-based qualitative assay for the 
detection of methylated SEPT9 DNA in human plasma. Clin Chem 60, 1183-1191, 
doi:10.1373/clinchem.2013.221044 (2014). 
Chan, K. C. et al. Noninvasive detection of cancer-associated genome-wide 
231 
232 
233 
11 
hypomethylation and copy number aberrations by plasma DNA bisulfite sequencing. 
Proc Natl Acad Sci U S A 110, 18761-18768, doi:10.1073/pnas.1313995110 (2013). 
Sun, K. et al. Plasma DNA tissue mapping by genome-wide methylation sequencing 
for noninvasive prenatal, cancer, and transplantation assessments. Proc Natl Acad 
234 
Sci U S A 112, E5503-5512, doi:10.1073/pnas.1508736112 (2015). 
236 
237 
12 
Grunau, C., Clark, S. J. & Rosenthal, A. Bisulfite genomic sequencing: systematic 
investigation of critical experimental parameters. Nucleic Acids Res 29, E65-65 
(2001). 
238 
239 
13 
Taiwo, O. et al. Methylome analysis using MeDIP-seq with low DNA concentrations. 
Nat Protoc 7, 617-636, doi:10.1038/nprot.2012.012 (2012). 
241 
14 
Newman, A. M. et al. Integrated digital error suppression for improved detection of 
circulating tumor DNA. Nat Biotechnol 34, 547-555, doi:10.1038/nbt.3520 (2016). 
242 
243 
244 
15 
16 
Sharma, S., Kelly, T. K. & Jones, P. A. Epigenetics in cancer. Carcinogenesis 31, 27-36, 
doi:10.1093/carcin/bgp220 (2010). 
Yin, Y. et al. Impact of cytosine methylation on DNA binding specificities of human 
246 
17 
transcription factors. Science 356, doi:10.1126/science.aaj2239 (2017). 
Michiels, S., Koscielny, S. & Hill, C. Prediction of cancer outcome with microarrays: a 
247 
248 
249 
18 
multiple random validation strategy. Lancet 365, 488-492, doi:10.1016/S0140­
6736(05)17866-0 (2005). 
Pedersen, K. S. et al. Leukocyte DNA methylation signature differentiates pancreatic 
cancer patients from healthy controls. PLoS One 6, e18223, 
251 
252 
253 
19 
doi:10.1371/journal.pone.0018223 (2011). 
Teschendorff, A. E. et al. An epigenetic signature in peripheral blood predicts active 
ovarian cancer. PLoS One 4, e8274, doi:10.1371/journal.pone.0008274 (2009). 
254 


 
 
 
256 
Supplementary information 
257 
Supplementary 
Information 
is 
linked 
to 
the 
online 
version 
of 
the 
paper 
at 
258 
www.nature.com/nature. 
259 
Acknowledgements 
260 
This study was conducted with support from the University of Toronto McLaughlin Centre (MC­
261 
2015-02), Canadian Institutes of Health Research (CIHR FDN 148430 and CIHR New 
262 
Investigator Salary award 201512MSH-360794-228629), Ontario Institute for Cancer Research 
263 
(OICR) with funds from the province of Ontario, Canada Research Chair (950-231346), and 
264 
Princess Margaret Cancer Foundation to D.D.C. as well as Canadian Cancer Society (CCSRI 
265 
701717) to R.J.H., CCSRI (704716) to R.J.H. and D.D.C. and CCSRI 703827 to M.M.H. 
266 
Recruitment of healthy individuals was supported by Cancer Care Ontario Chair of Population 
267 
Health and CCSRI 020214 awarded to R.J.H. Collection of lung cancer samples was supported 
268 
by the Alan B Brown chair in molecular genomics and the Lusi Wong Lung Cancer Early 
269 
Detection Program to GL. We would like to acknowledge the Princess Margaret Genomics 
270 
Centre for carrying out the NGS sequencing and the Bioinformatics and HPC Core, Princess 
271 
Margaret Cancer Centre for their expertise in generating the NGS data. 
272 
273 
Author Contributions 
274 
SYS and DDC designed and developed the cfMeDIP-seq protocol. RJH and GF conceived and 
275 
designed the study related to the pancreatic cancer component.  SYS, RS, AC, and DDC 
276 
conceived and designed the study related to the other cancer types. SYS, SVB, TJP, and DDC 
277 
designed the experiments. SYS, DC, MHAR, PCZ, ZC, TL, OK, DR, IE, ZC, SC, GMO, JL, 


 
 
 
 
 
278 
MM and ZZ performed the experiments. TM, YW, and COB performed the mouse experiments. 
279 
RS, AC, GF, TTW, AG, TJP, MMH and DDC analyzed the data with scientific input from RJH. 
280 
GF, AB, DC, AS, TM, AA, NL, MHAR, JDM, PLB, NF, GL, MDM, SG, TJP, and RJH 
281 
collected the clinical data related to the samples, determined the sample selection criteria and 
282 
matching scheme, and provided the clinical samples. SYS, RS, AC and DDC wrote the paper 
283 
with feedback from all authors.  
284 
Competing Interests 
285 
DDC, SYS, AC, SVB, RS, and RJH are listed as inventors/contributors on patents filed related to 
286 
this work. 
287 
Reprints and permissions information is available at www.nature.com/reprints. Correspondence 
288 
and requests for materials should be addressed to Daniel D. De Carvalho.  
289 


 
 
 
 
 
 
 
 
 
291 
292 
293 
294 
295 
296 
297 
298 
299 
300 
301 
302 
303 
304 
305 
306 
307 
308 
309 
310 
311 
312 
290 
Figure Legends 
Fig. 1: cfDNA methylome as a sensitive approach to detect ctDNA in low input DNA. a, 
Simulated probability of detecting at least one epimutation as a function of ctDNA concentration 
(0.001% to 10%) (columns), number of DMRs analyzed (1 to 10,000) (rows), and sequencing 
depth (10X to 10,000X) (x-axis). b, Across a serial dilution series (n=7 dilution points, two 
technical replicates, each replicate was used per protocol) of HCT116 DNA spiked into MM.1S 
multiple myeloma cells, near-perfect correlations are observed between observed vs expected 
methylation signal within DMRs in RPKMs. c, Frequency of ctDNA (human) as a percentage of 
total cfDNA (human + mice) in the plasma from two colorectal cancer, patient-derived 
xenografts (PDX) before and after cfMeDIP-seq. 
Fig. 2: cfMeDIP-seq method can identify thousands of differentially methylated regions 
(DMRs) in circulating cfDNA obtained from pancreatic adenocarcinoma patients. a, 
Experimental design. b, Volcano plot of DMRs from pancreatic cancer (cases, n=24) versus 
healthy donors (controls, n=24) using cfMeDIP-seq. Red dots indicate windows significant at 
BHFDR < 0.1, (Negative Binomial GLM, two-sided p-values). c, Heatmap of the 14,716 DMRs 
identified in the plasma cfDNA from cases and controls (Euclidean distance, Ward clustering). 
Dendrogram shows separation by case/control status. d, e, Overlap between case-vs-control 
plasma-derived DMRs and RRBS tumour DMRs matched normal tissue (d) and PBMCs (e). 
Boxplots represent expected null distribution of overlaps from 1000 permutations (two sided, p-
values computed using standard normal distribution). Extremes of boxes and center-lines define 
upper and lower quartiles and medians. Whiskers indicate 1.5 x interquartile range (IQR). 
Diamonds represent observed overlap (red if significantly enriched, green if significantly 


   
 
 
 
 
313 
314 
315 
316 
317 
318 
319 
320 
321 
322 
323 
324 
325 
326 
327 
328 
329 
330 
331 
332 
333 
334 
335 
depleted, and blue if not significant). Horizontal lines indicate thresholds for statistical 
significance. 
Fig. 3: Methylome analysis of plasma cfDNA allows tumor classification. a, cfMeDIP-seq 
carried out on a discovery cohort consisting of 189 samples from 7 different tumor types: 
pancreatic cancer (PDAC), AML, bladder cancer (BLCA), breast cancer (BRCA), colorectal 
cancer (CRC), lung cancer (LUC) and renal cancer (RCC), including early and late stage tumors, 
and healthy controls (normal). For each cancer type, DMRs between the cancer type and normal 
controls were identified. Overlap is shown between plasma-derived DMRs for each cancer type 
and primary tumor DMRs (tumor tissue versus adjacent normal tissue) for the corresponding 
cancer type using TCGA data. Boxplots represent expected null distribution of overlaps from 
1000 permutations (two sided, p-values computed using standard normal distribution). Extremes 
of boxes and center-lines define upper and lower quartiles and medians. Whiskers indicate 1.5 x 
interquartile range (IQR). Diamonds represent observed overlap (red if significantly enriched, 
green if significantly depleted, and blue if not significant). Horizontal lines indicate thresholds 
for statistical significance. b, Evaluation of classification accuracy on the discovery cohort. The 
discovery cohort (n=189) was partitioned into 100 independent training and test sets in an 80%­
20% manner, consisting of 8 classes (cancer types and healthy controls). Training sets were used 
for DMR-selection and model training, yielding 100 sets of 8 one-class vs-other-classes binomial 
GLMnet classifiers. Y-axis depicts distributions of AUROC (area under the receiver operator 
characteristic curve) for each held-out test set for each class. Dots represent performance in 
individual test sets. 
Ends of boxes represent upper and lower-quartiles, line within box 
represents median, and whiskers represent 1.5 x IQR. c, ROC (receiver operating characteristic) 


 
  
 
 
 
336 
curves constructed using averaged class probabilities for independent validation set samples 
337 
(n=199, 55 LUC, 35 AML, 47 PDAC and 62 healthy controls) from the 100 models for each 
338 
one-class-vs-other-classes comparison trained using the discovery cohort. d, ROC curves for the 
339 
PDAC and LUC validation set divided into early and late stage, showing that the ability to 
340 
discriminate PDAC or LUC samples is similar when considering early and late stage samples of 
341 
that class separately. 
342 
343 
Fig. 4: Plasma-derived DMRs are informative of cancer type. a, The plasma-derived DMRs 
344 
identified as informative of cancer type in the discovery cohort of 189 plasma samples was used 
345 
to generate 3D and 2D tSNE (t-distributed stochastic neighbor embedding) plots for the entire 
346 
cohort of plasma samples (n=388). b, c, The DNA methylation Beta value for probes within the 
347 
plasma-derived DMRs were used to generate 3D and 2D tSNE plots for (b) TCGA cancer tissue 
348 
(n=4,032) and (c) COSMIC cancer cell line (n=400 cell lines). 
349 
Methods 
350 
Bioinformatic simulation of tumor-specific features and probability of detection by 
351 
sequencing depth 
352 
We created 145,000 simulated genomes with 1, 10, 100, 1,000 and 10,000 independent loci with 
353 
0.001 – 10% cancer-specific DMRs in 10-fold increments. 14,500 diploid genomes (expected 
354 
copy number in 100 ng cfDNA) were then sampled from these mixtures and further sampled 10 
355 
– 10,000x in 10-fold increments at each locus. The process was repeated 100 times for each 
356 
combination of parameters. Probability curves were plotted for successful detection of >1 and >5 
357 
DMRs (Fig. 1a and Extended Data Fig.1a). 


 
 
 
 
358 
cfMeDIP-seq 
359 
A schematic representation of the cfMeDIP-seq protocol is shown in Extended Data Fig. 1b. 
360 
Prior to cfMeDIP, the samples were subjected to library preparation using Kapa HyperPrep Kit 
361 
(Kapa Biosystems), following manufacturer’s protocol with minor modifications. Briefly, after 
362 
end-repair and A-tailing, samples were ligated to 0.181 µM of NEBNext adapter (NEBNext 
363 
Multiplex Oligos for Illumina kit, New England BioLabs, Canada) by incubating at 20˚C for 20 
364 
min and purified with AMPure XP beads (Beckman Coulter). The eluted library was digested 
365 
using the USER enzyme (New England BioLabs, Canada) followed by purification with Qiagen 
366 
MinElute PCR Purification Kit (MinElute columns) prior to MeDIP. 
367 
The prepared libraries were combined with the filler λ DNA (to ensure the total amount of DNA 
368 
[cfDNA + filler] was 100 ng) and subjected to MeDIP with Diagenode MagMeDIP kit (Cat# 
369 
C02010021) using the protocol from Taiwo et al. 201213 with some modifications. The filler 
370 
DNA consists of a mixture of unmethylated and in vitro methylated λ amplicons of different 
371 
CpG densities (Supplementary Table 6), similar in size to adapter-ligated cfDNA libraries. Its 
372 
addition ensures a constant ratio of antibody to input DNA and helps maintain similar 
373 
immunoprecipitation efficiency across samples regardless of available cfDNA, while minimizing 
374 
non-specific binding by the antibody and DNA loss due to binding to plasticware. For MeDIP, 
375 
the prepared library/filler DNA mixture was combined with 0.3 ng of control methylated and 0.3 
376 
ng of the control unmethylated A. thaliana DNA provided in the kit, and the buffers. The mixture 
377 
was heated to 95˚C for 10 min, then immediately placed into an ice water bath for 10 min. Each 
378 
sample was partitioned into two 0.2 mL PCR tubes: one for the 10% input control (7.9 µl) and 
379 
the other one for the sample to be subjected to immunoprecipitation (79 µl). The included 5-mC 


 
 
 
380 
385 
390 
395 
400 
monoclonal antibody 33D3 (Cat#C15200081) from the MagMeDIP kit was diluted 1:15 prior to 
381 
generating the diluted antibody mix and added to the sample. Washed magnetic beads (following 
382 
manufacturer instructions) were also added prior to incubation at 4˚C for 17 hours. The samples 
383 
were purified using the Diagenode iPure Kit v2 (Cat# C03010015) and eluted in 50 µl of Buffer 
384 
C. The success of the reaction (QC1) was validated through qPCR to detect recovery of the 
spiked-in methylated and unmethylated A. thaliana DNA. The % recovery of unmethylated 
386 
spiked-in DNA should be <1% (relative to input control, adjusted for input control being 10% of 
387 
overall sample) and the % specificity of the reaction should be >99% (as calculated by (1 – 
388 
[recovery of spiked-in unmethylated control DNA over recovery of spiked-in methylated control 
389 
DNA])*100), prior to proceeding to the next step. The optimal number of cycles to amplify each 
library was determined through qPCR, after which the samples were amplified using Kapa HiFi 
391 
Hotstart Mastermix and NEBNext multiplex oligos, added to a final concentration of 0.3 µM. 
392 
The final libraries were amplified as follows: activation at 95˚C for 3 min, followed by 
393 
predetermined cycles of 98˚C for 20 s, 65˚C for 15 s and 72˚C for 30 s and a final extension of 
394 
72˚C for 1 min. The amplified libraries were purified using MinElute columns, then gel size 
selected with 3% Nusieve GTG agarose gel to remove any adapter dimers. All the final libraries 
396 
were submitted for BioAnalyzer analysis prior to sequencing at the Princess Margaret Genomics 
397 
Centre (PMGC) on an Illumina HiSeq 2500, SBS V4 chemistry, single read 50 bp, multiplexed 
398 
as 7 samples/lane. After sequencing, the sequenced reads were aligned to λ and hg19 using 
399 
Bowtie20 with the default settings. Based on virtually no alignment to λ genome, the filler DNA 
does not interfere with the generation of sequencing data (Supplementary Table 7 and 8).  
401 
The generated SAM files from hg19 alignment were converted to BAM format, ensuring 
402 
removal of duplicate reads, sorting and indexing of reads using SAMtools21 prior to subsequent 


 
 
 
 
 
 
403 
analysis with R package MEDIPS22. CpG Enrichment Score, as a quality control measure for the 
404 
immunoprecipitation reaction, was calculated as part of the MEDIPS package.  
405 
Validation of cfMeDIP-seq against MeDIP-seq 
406 
DNA from human colorectal cancer cell (CRC) line HCT116 (American Type Culture Collection 
407 
(ATCC), mycoplasma free) was extracted using PureLink Genomic DNA Mini Kit 
408 
(Thermofisher Scientific). HCT116 was chosen because of the availability of public DNA 
409 
methylation data. Genomic DNA was sheared to mimic cfDNA using a Covaris sonicator, and 
410 
larger size fragments excluded using AMPure XP beads (Beckman Coulter) to mimic the 
411 
fragment size of cell-free DNA. cfMeDIP-seq was carried out on 1, 5, 10 and 100 ng of sheared 
412 
DNA as input, with 100 ng representing the gold-standard MeDIP-seq protocol, with 2 biological 
413 
replicates per input. The fold enrichment of a methylated human DNA region (TSH2B) over 
414 
unmethylated human DNA region (GAPDH promoter), using primers provided in MagMeDIP kit,  
415 
was determined prior to sequencing libraries to saturation (Extended Data Fig. 2a-c, 
416 
Supplementary Table 7).  
417 
Dilution series of sheared cell line DNA 
418 
As with the CRC DNA, the same extraction and shearing protocol was used with multiple 
419 
myeloma (MM) cell line MM.1S. A dilution series of CRC into MM DNA was carried out 
420 
following Extended Data Fig. 3a scheme. This dilution series was used for cfMeDIP-seq 
421 
(Supplementary Table 9) and for ultra-deep targeted sequencing for CRC point mutation 
422 
detection, using a starting input of 60 ng of DNA. For the mutation detection, DNA libraries 
423 
were prepared using Kapa HyperPrep Kit (Kapa Biosystems) and Illumina compatible molecular 


 
 
 
425 
430 
435 
440 
445 
424 
barcoded adapters with 2-bp in-line barcodes (unique molecular identifiers (UMIs)) to ensure 
optimal analytical sensitivity for mutation detection14. A customized biotinylated DNA capture 
426 
probe panel (xGen Lockdown Custom Probes Mini Pool, Integrated DNA Technologies) 
427 
targeting exons from 5 genes (13kb) was used23. In brief, the barcoded libraries were pooled, and 
428 
hybrid capture was performed according to the manufacturer’s instructions (IDT xGEN 
429 
Lockdown protocol version 4). The amplified post-capture libraries were sequenced 
to >100,000X read coverage using Illumina HiSeq 2500 instrument, SBS V4 chemistry, paired­
431 
end 125 bp, as 4 samples/lane. Average target coverage of unprocessed reads was 186,312X 
432 
(range: 154,419X – 216,434X) (Supplementary Table 9).  
433 
After sequencing, reads were de-multiplexed using sample specific indices into separate paired­
434 
end FASTQ files. A two base pair molecular barcode and a one base pair invariant spacer 
sequence were removed from each read. A thymine base was encoded in the third position for 
436 
adapter ligation and a spacer filter was enforced to remove reads incompliant with this design. 
437 
The extracted barcodes from paired-end reads were grouped and written into the header of each 
438 
sequence for downstream in silico molecular identification24. FASTQ files were mapped to the 
439 
human reference genome hg19 using BWA25, processed using the Genome Analysis ToolKit 
(GATK) IndelRealigner26, and sorted and indexed using SAMtools21 . 
441 
Barcodes were used in combination with endogenous sequence features (genome coordinates, 
442 
mapping alignments, read orientation, and read number in pair) to confer sequences from 
443 
individual molecules. Consensus sequences were formed from two or more reads supporting the 
444 
same molecule with 70% agreement amongst bases above Phred quality scores (Q)27 of 30. 
Reads derived from the same strand of a unique fragment were collapsed to form single strand 


 
 
 
446 
consensus sequences (SSCS), suppressing polymerase and sequencer errors. These condensed 
447 
reads were subsequently combined with their complementary strand into duplex consensus 
448 
sequences (DCS). This enables an additional layer of error suppression as double strand 
449 
consolidated sequences can correct for asymmetric damage accrued during the first cycle of PCR 
450 
or induced by oxidation28 . 
451 
We selected variants based on annotated SNPs from the Cancer Cell Line Encyclopedia 
452 
(CCLE)29 overlapping our target panel. SNVs were called with MuTect30 using the following 
453 
parameters: --enable_extended_output --tumor_f_pretest 0.000001f --downsampling_type 
454 
NONE --force_output --force_alleles --gap_events_threshold 1000 --fraction_contamination 
455 
0.00f --coverage_file30. We force called every base for each variant to assess limit of detection 
456 
and background noise at each stage of barcode-mediated error correction. Analysis of the UMI­
457 
processed error-suppressed reads revealed unique molecule (i.e., SSCS) and DCS average target 
458 
coverage of 6,276X (4,284X – 8,068X) and 1,043X (654X – 1,602X), respectively 
459 
(Supplementary Table 9). 
460 
Specimen processing – Patient Derived Xenograft (PDX) cfDNA  
461 
All mouse work was carried out in compliance with animal use protocol and ethical regulations 
462 
approved by the Animal Care Committee at University Health Network. Human colorectal tumor 
463 
tissue obtained with patient consent and UHN Research Ethics Board approval from the UHN 
464 
Biobank, was digested to single cells using collagenase A. Single cells were subcutaneously 
465 
injected into 4-6 week old NOD/SCID male mouse. Mice were euthanized by CO2 inhalation 
466 
prior to blood collection by cardiac puncture and stored in EDTA tubes. From the collected 
467 
blood samples, plasma was isolated and stored at -80 ˚C. cfDNA was extracted from 0.3-0.7 ml 


 
 
 
 
 
 
 
468 
of plasma using the QIAamp Circulating Nucleic Acid Kit (Qiagen). Two biological samples 
469 
with 10 ng of starting cfDNA were subjected to cfMeDIP-seq protocol as previously mentioned, 
470 
sequenced and analyzed (Supplementary Table 10).  
471 
Donor recruitment and sample acquisition  
472 
All patients have provided written informed consents, and all samples have been obtained upon 
473 
approval of the institutional ethics committees and Research Ethics Boards from University 
474 
Health Network (UHN) and Mount Sinai Hospital (MSH), in compliance with all relevant ethical 
475 
regulations. Pancreatic adenocarcinoma cases (PDAC) were obtained from the Ontario 
476 
Pancreatic Cancer Study and the UHN Biobank. Colorectal and breast cancer plasma samples 
477 
were obtained from the UHN Biobank. Lung cancer plasma samples were obtained from the 
478 
UHN Thoracic Biobank. AML samples were obtained from the UHN Leukemia Biobank. 
479 
Bladder and renal cancer plasma samples were obtained from the UHN Genitourinary (GU) 
480 
Biobank from consenting urologic oncology patients, procured prior to nephrectomy and 
481 
cystectomy respectively. Lastly, healthy controls were recruited through the Family Medicine 
482 
Centre at MSH in Toronto, Canada. 
483 
Specimen processing and methylation analysis of purified tumor and normal cells from 
484 
PDAC samples 
485 
For primary PDAC samples, specimens were processed immediately following resection and 
486 
representative sections were used to confirm the diagnosis. Laser capture microdissection (LCM) 
487 
of freshly liquid nitrogen-frozen tissue samples was performed on a Leica LMD 7000 instrument. 
488 
LCM was performed on the same day when sections were cut to minimize nucleic acid 
489 
degradation. Qiagen Cell Lysis Buffer was used to extract genomic DNA.  


 
 
 
 
 
490 
495 
500 
505 
510 
Quantified 10 ng of genomic DNA for each sample was analyzed using RRBS following the 
491 
protocol from Gu et al., 201131 with minor modifications. DNA libraries ligated to Illumina 
492 
TruSeq methylated adapters were subjected to bisulfite conversion using the Zymo EZ DNA 
493 
methylation kit following manufacturer’s protocol, followed by gel size selection for fragments 
494 
of 160 bp-300 bp in size. After determining the optimal number of cycles to amplify each 
purified library, samples were amplified using Kapa HiFi Uracil+ Mastermix (Kapa Biosystems) 
496 
and purified with AMPure beads (Beckman Coulter). The final libraries were submitted for 
497 
BioAnalyzer analysis prior to sequencing at PMGC on an Illumina HiSeq 2000, using 
498 
sequencing by synthesis (SBS) V3 chemistry, single read 50bp and multiplexed as 4 
499 
samples/lane. After sequencing, the raw data for each sample was trimmed with Trim Galore! 
using the RRBS settings prior to aligning to hg19 using Bismark32 with Bowtie233 
501 
(Supplementary Table 11). The generated SAM files were then converted to BAM format, sorted 
502 
and indexed using SAMtools. 
503 
Specimen Processing – patient cell-free DNA (cfDNA) 
504 
Plasma samples collected using EDTA and ACD (acid citrate dextrose) tubes, were obtained 
from the UHN BioBanks and MSH and were kept frozen until use. cfDNA was extracted from 
506 
0.5-3.5 mL of plasma using the QIAamp Circulating Nucleic Acid Kit (Qiagen) and quantified 
507 
through Qubit prior to use. Sex, age and pathology stage are available in Supplementary Table 
508 
12, while extracted DNA quantities are available in Extended Data Fig. 8a. 
509 
Calculation and visualization of differentially methylated regions from cfDNA of 
pancreatic cancer patients and healthy donors 


 
 
 
 
 
511 
DMRs between cfDNA samples from 24 pancreatic cancer (PDAC) patients and 24 healthy 
512 
donors (controls) were calculated using MEDIPS and DESeq2 R packages22,34. For each sample, 
513 
we computed counts per 300 bp non-overlapping windows, filtered out windows with less than 
514 
10 counts across all samples and fit a negative binomial model to call DMRs at FDR < 0.1 (Wald 
515 
test). Z-scores of DMR RPKM values with Euclidean Distance and Ward clustering were used 
516 
for visualisation. 
517 
Enrichment analyses for plasma-derived DMRs in tumor-specific methylation signals in 
518 
PDAC 
519 
Five normal PBMC samples profiled by RRBS were downloaded from GEO (GSE89473) for 
520 
comparison with the 24 pancreatic cancer tissue RRBS samples. R package MethylKit was used 
521 
to parse files and autosomal CpGs detected in at least 18 out of the 24 PDACs and 4 out of the 5 
522 
PBMCs were retained for further analysis. We obtained DMCs at FDR <0.01, delta-Beta >0.25. 
523 
A null distribution was then generated, from 1000 resamples, preserving the relationship between 
524 
the number of CpGs in windows that were seen in the original intersections between RRBS 
525 
features and cfMeDIP DMRs. Then we computed the frequency of overlap between DMRs 
526 
hypermethylated in both, hypermethylated in one but not the other, hypomethylated in one but 
527 
not the other, and finally, hypomethylated in both comparisons.  The distributions were then 
528 
standardized based on z-scores and used to compute Bonferroni-adjusted p-values to determine 
529 
enrichment. The same procedure was employed for subsequent enrichment tests in the 
530 
manuscript. 
531 
Enrichment analyses for cfMeDIP DMRs in TCGA 450K DMCs relative to normal tissues 
532 
and PBMCs. 


 
 
 
 
 
533 
189 cfDNA samples were obtained across 7 cancer types (AML, bladder (BLCA), breast 
534 
(BRCA), colorectal (CRC), lung (LUC), pancreatic (PDAC) and renal cancer (RCC)) and 
535 
healthy donors (normal) (Supplementary Table 12). After processing of cfMeDIP-seq data from 
536 
these samples, DMRs were calculated using DESeq2 between each cancer type and healthy 
537 
donors as described above. DMCs were also calculated between TCGA 450K methylation array 
538 
samples from each corresponding cancer type (n = 3979) (obtained from SAGE synapse) and 
539 
PBMCs (n = 53, obtained from GEO) samples using limma (FDR < 0.01, absolute delta-Beta 
540 
0.25). Statistical tests for enrichment were performed as described above for PDAC RRBS 
541 
samples. The same procedure was carried out for DMCs calculated between TCGA 450K 
542 
methylation array samples from a cancer type and normal samples from the same tissue, for 
543 
BLCA, BRCA, CRC, LUC and RCC. 
544 
Examination of transcription factors associated with differentially methylated motifs in 
545 
cfMeDIP-seq DMRs. 
546 
RNA-seq data obtained as median RPKMs from the GTEx consortium across 53 human tissues – 
547 
as described in the supplemental R Markdowns in Zenodo (ID 10.5281/zenodo.1205756) 
548 
(Supplementary Table 13), and median expression per tissue was visualized in heatmaps. To 
549 
look for enrichment of TF expression and DMR-associated TF motifs, we selected 1000 random 
550 
sets of TFs. As part of the analysis, we considered the known sensitivity to the methylation status 
551 
of each TF16, yielding 42 TFs that are enriched in healthy donors, and 52 TFs that are enriched in 
552 
pancreatic adenocarcinoma cases.  
553 
We computed ssGSEA (single-sample gene set enrichment analysis) scores for the expression of 
554 
these TFs per sample, for pancreatic cancer (TCGA), blood (GTEx) and normal pancreas (GTEx) 


 
 
 
 
 
 
 
555 
560 
565 
570 
575 
and compared distributions to those from random sets of TFs using Wilcoxon’s Rank Sum Test. 
556 
Violin plots were made as described in the supplemental R Markdown 10.5281/zenodo.1205735 
557 
(Supplementary Table 13). 
558 
Machine learning analyses for evaluation of classification accuracy.  
559 
Model training and evaluation on the discovery cohort. 
In order to evaluate the performance of cfMeDIP data in tumor classification without high 
561 
computational cost, we reduced the initial set of possible candidate features to windows 
562 
encompassing CpG Islands, shores, shelves and FANTOM5 enhancers (“regulatory features”), 
563 
yielding a matrix of 189 samples and 505,027 features. 
564 
We then used the caret R package35 to partition the discovery cohort data into 100 class-balanced 
independent training and test sets in an 80%-20% manner. Then, we selected the top 300 DMRs 
566 
by moderated t-statistic (150 hypermethylated, 150 hypomethylated) on the training data 
567 
partition using limma-trend36  for each class versus other classes. A binomial GLMnet was then 
568 
trained using these DMRs (up to 300 DMRs x 7 other classes = 2100 features) using of 3 
569 
iterations of 10-Fold Cross-Validation (CV) to optimize values of the mixing parameter (alpha, 
values = 0, 0.2, 0.5, 0.8 and 1) and the penalty (lambda, values = 0 – 0.05 in increments of 0.01) 
571 
using Cohen’s Kappa as the performance metric. For each training set, this yielded a collection 
572 
of 8 one-class vs-other-classes binomial classifiers. 
573 
We then estimated classification performance on the held-out test set using the AUROC (area 
574 
under the receiver operating characteristic curve). These estimates represent unbiased measures 
of classification, as the held-out test set samples were not used for either DMR pre-selection or 


 
 
 
 
 
 
576 
GLMnet training and tuning. The 100 independent training and test sets also permitted for 
577 
minimization of optimistic estimates due to training-set bias. 
578 
Model evaluation on the validation cohort. 
579 
For each validation cohort cfMeDIP sample, we estimated class probabilities for the AML, 
580 
PDAC, LUC and normal one-vs-all binomial classifiers trained on the 100 different training sets 
581 
within the discovery cohort. The probabilities from the 100 models were averaged to produce a 
582 
single score that was then used for AUROC estimation. We also evaluated if disease stage 
583 
(applicable to only LUC and PDAC) affected performance by estimating AUROC when either 
584 
early (Stages I and II) or late stage samples (Stages III and IV) of a particular class were left out 
585 
for the one-vs-all classifiers trained to identify the class in question. 
586 
Validation in cell lines 
587 
450K profiles for 1,028 cell lines previously characterised in Iorio et al (2016)37 were obtained as 
588 
IDAT files. The data were then uniformly processed using the ssNoob method in the minfi 
589 
package38. We reduced this dataset to tissue types for which cfMeDIP data were available 
590 
(n=400). 
591 
Data availability statement 
592 
We have deposited R markdowns (either knit or raw) and scripts used to generate the findings in 
593 
this study on Zenodo (DOIs in Supplementary Table 13). All the cell line datasets generated 
594 
during and/or analysed during the current study are available in the GEO repository under 
595 
accession code GSE79838. The cfMeDIP-seq NGS data for patient samples that support the 
596 
findings of this study are available upon request from the corresponding author (D.D.C) 
597 
to comply with institutional ethics regulation. Source data for Fig. 1b and Extended Data Fig. 3e 


 
 
598 
are provided in Supplementary Table 9, and for Fig. 1c are provided in Supplementary Table 10. 
599 
Additional source data can be found on Zenodo (Supplementary Table 13) 
600 


 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
603 
604 
605 
606 
607 
608 
609 
610 
611 
612 
613 
614 
615 
616 
617 
618 
619 
620 
621 
622 
623 
624 
625 
626 
627 
628 
629 
630 
631 
632 
633 
634 
635 
636 
637 
638 
639 
640 
641 
642 
643 
644 
645 
601 
Methods Reference 
602 
20 
Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory-efficient 
alignment of short DNA sequences to the human genome. Genome Biol 10, R25, 
doi:10.1186/gb-2009-10-3-r25 (2009). 
21 
Li, H. et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 
2078-2079, doi:10.1093/bioinformatics/btp352 (2009). 
22 
Lienhard, M., Grimm, C., Morkel, M., Herwig, R. & Chavez, L. MEDIPS: genome-wide 
differential coverage analysis of sequencing data derived from DNA enrichment 
experiments. Bioinformatics 30, 284-286, doi:10.1093/bioinformatics/btt650 
(2014). 
23 
Kis, O. et al. Circulating tumour DNA sequence analysis as an alternative to multiple 
myeloma bone marrow aspirates. Nat Commun 8, 15086, 
doi:10.1038/ncomms15086 (2017). 
24 
Kennedy, S. R. et al. Detecting ultralow-frequency mutations by Duplex Sequencing. 
Nat Protoc 9, 2586-2606, doi:10.1038/nprot.2014.170 (2014). 
25 
Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler 
transform. Bioinformatics 25, 1754-1760, doi:10.1093/bioinformatics/btp324 
(2009). 
26 
DePristo, M. A. et al. A framework for variation discovery and genotyping using next-
generation DNA sequencing data. Nat Genet 43, 491-498, doi:10.1038/ng.806 
(2011). 
27 
Ewing, B. & Green, P. Base-calling of automated sequencer traces using phred. II. 
Error probabilities. Genome Res 8, 186-194 (1998). 
28 
Schmitt, M. W. et al. Detection of ultra-rare mutations by next-generation 
sequencing. Proc Natl Acad Sci U S A 109, 14508-14513, 
doi:10.1073/pnas.1208715109 (2012). 
29 
Barretina, J. et al. The Cancer Cell Line Encyclopedia enables predictive modelling of 
anticancer drug sensitivity. Nature 483, 603-607, doi:10.1038/nature11003 (2012). 
30 
Cibulskis, K. et al. Sensitive detection of somatic point mutations in impure and 
heterogeneous cancer samples. Nat Biotechnol 31, 213-219, doi:10.1038/nbt.2514 
(2013). 
31 
Gu, H. et al. Preparation of reduced representation bisulfite sequencing libraries for 
genome-scale DNA methylation profiling. Nat Protoc 6, 468-481, 
doi:10.1038/nprot.2010.190 (2011). 
32 
Krueger, F. & Andrews, S. R. Bismark: a flexible aligner and methylation caller for 
Bisulfite-Seq applications. Bioinformatics 27, 1571-1572, 
doi:10.1093/bioinformatics/btr167 (2011). 
33 
Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat 
Methods 9, 357-359, doi:10.1038/nmeth.1923 (2012). 
34 
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and 
dispersion for RNA-seq data with DESeq2. Genome Biol 15, 550, 
doi:10.1186/s13059-014-0550-8 (2014). 
35 
Kuhn, M. Building Predictive Models inRUsing thecaretPackage. Journal of Statistical 
Software 28, doi:10.18637/jss.v028.i05 (2008). 


 
 
 
 
 
 
 
 
 
646 
647 
648 
649 
36 
37 
Law, C. W., Chen, Y., Shi, W. & Smyth, G. K. voom: Precision weights unlock linear 
model analysis tools for RNA-seq read counts. Genome Biol 15, R29, doi:10.1186/gb­
2014-15-2-r29 (2014). 
Iorio, F. et al. A Landscape of Pharmacogenomic Interactions in Cancer. Cell 166, 
650 
651 
652 
653 
654 
38 
740-754, doi:10.1016/j.cell.2016.06.017 (2016). 
Aryee, M. J. et al. Minfi: a flexible and comprehensive Bioconductor package for the 
analysis of Infinium DNA methylation microarrays. Bioinformatics 30, 1363-1369, 
doi:10.1093/bioinformatics/btu049 (2014). 
655 
656 


 
 
 
 
657 
Extended Data Figure Legends 
658 
Extended Data Fig. 1 Simulation of the probability of detecting ctDNA as a function of the 
659 
number of DMRs, sequencing depth and percentage of ctDNA in plasma cfDNA and 
660 
proposed method to enrich ctDNA. a, Bioinformatic simulation of scenarios with different 
661 
proportions of ctDNA present in the sample (0.001% to 10%, column facets), and a range of 
662 
tumor specific Differentially Methylated Regions (DMRs), from 1, 10, 100, 1,000, or 10,000, 
663 
determined through the comparison of ctDNA to normal cfDNA (row facets), with reads 
664 
sampled at varying sequencing depths, at each locus (10X, 100X, 1,000X, and 10,000X) (x-axis). 
665 
The probability of detecting at least 5 epimutations/DMR increases as the number of available 
666 
features increases, even at shallow coverage per locus (left y-axis). Each panel depicts 
667 
probability of detection against coverage per candidate DMR for one simulation scenario. b, 
668 
Schematic representation of the cfMeDIP-seq protocol. 
669 
Extended Data Fig. 2 Sequencing saturation analysis and quality controls of MeDIP-seq 
670 
and cfMeDIP-seq carried out on varying starting inputs of HCT116 cell line DNA sheared 
671 
to mimic cfDNA (HCT116 cfDNA mimic DNA). a, The figure shows the results of the 
672 
saturation analysis from the Bioconductor package MEDIPS analyzing cfMeDIP-seq data from 
673 
each replicate, for each starting input amount and including an input control. b, The protocol was 
674 
tested in two biological replicates of four starting DNA inputs (100, 10, 5, and 1 ng) of HCT116 
675 
cfDNA mimic DNA. Specificity of the reaction was calculated using methylated and 
676 
unmethylated spiked-in A. thaliana DNA. Fold enrichment ratio was calculated using genomic 
677 
regions of the fragmented HCT116 DNA (human methylated TSH2B0 and unmethylated 
678 
GAPDH). The horizontal dotted line indicates a fold-enrichment ratio threshold of 25, dots 


 
 
 
680 
685 
690 
695 
700 
679 
represent biological replicate, with line representing mean c, CpG Enrichment Scores of the 
sequenced samples (two biological replicates each of four starting DNA inputs (100, 10, 5, and 1 
681 
ng), and one input control) show a robust enrichment of CpGs within the genomic regions from 
682 
the immunoprecipitated samples compared to the input control. The CpG Enrichment Score was 
683 
obtained by dividing the relative frequency of CpGs of the regions by the relative frequency of 
684 
CpGs of the human genome. The horizontal dotted line indicates a CpG Enrichment Score of 1, 
dots represent biological replicates, with line representing mean. d, Genome-wide Pearson 
686 
correlations of normalized read counts per 300 bp window between cfMeDIP-seq signal for 1 to 
687 
100 ng of input HCT116 DNA sheared to mimic cfDNA (2 biological replicates per 
688 
concentration). e, Genome Browser snapshot of HCT116 cfMeDIP-seq signal across a window 
689 
(chr8:145,095,942-145,116,942) selected out of 4 examined loci, at different starting DNA 
inputs (1 to 100 ng, in biological replicates), compared with RRBS (ENCODE ENCSR000DFS) 
691 
and WGBS (GEO GSM1465024) data (aligned to hg19). For cfMeDIP-seq, y-axis indicates 
692 
RPKMs, for RRBS, yellow and blue blocks represent hypermethylated and hypomethylated 
693 
CpGs, respectively. In WGBS track, peak heights indicate methylation level. 
694 
Extended Data Fig. 3 Sequencing saturation analysis and quality controls of cfMeDIP-seq 
from serial dilution. a, Schematic representation of the CRC DNA (HCT116) dilution series 
696 
into MM DNA (MM1.S). For both CRC and MM DNA, the genomic DNA was sheared to 
697 
mimic cfDNA fragmentation. The entire dilution series was used to carry out cfMeDIP-seq (n=1) 
698 
and ultra-deep sequencing for mutation detection (n=1). b, Specificity of reaction for each 
699 
dilution in the series (n=1) was calculated using methylated and unmethylated spiked-in A. 
thaliana DNA. c, CpG enrichment representing ratio of relative frequency of CpGs in regions to 
701 
relative frequency of CpGs in the human genome for each dilution in the series (n=1), 


 
 
  
 
 
702 
determined via cfMeDIP-seq.  Horizontal dashed line represents CpG enrichment of 1. d, 
703 
Saturation analysis of cfMeDIP-seq sequenced reads from each dilution point in the series (n=1). 
704 
e, Across a serial dilution series (n=7 dilution points, two technical replicates, each replicate was 
705 
used per protocol) of HCT116 DNA spiked into MM.1S multiple myeloma cells, near-perfect 
706 
correlations are observed between observed vs expected numbers of DMRs. f-g, Ultra-deep 
707 
sequencing for mutation detection of three CRC-specific point mutations within BRAF (p.P301P), 
708 
KRAS (p.G13D) and PIK3CA (p.H1047R) in the same dilution series (of CRC into MM DNA) 
709 
(n=1). Unique molecular identifiers (UMIs) were incorporated into the sequencing adapters and 
710 
used to created single strand consensus sequence (SSCS) (f) duplex consensus sequences (DCS) 
711 
(g) for the detection of allele frequency for each mutation at each locus. For each mutation, the 
712 
reference allele is found at the top. Dashed red line: limit of detection. 
713 
Extended Data Fig. 4 Quality controls for cfMeDIP-seq from circulating cfDNA from 
714 
pancreatic adenocarcinoma patients (cases) and healthy donors (controls). a-b, Specificity 
715 
of reaction calculated using methylated and unmethylated spiked-in A. thaliana DNA for (a) 
716 
each case sample and (b) each control sample. Fold enrichment ratio was not calculated due to 
717 
the very limited amount of DNA available after final libraries were generated. c-d, CpG 
718 
enrichment of the sequenced cases (c) and controls (d), Horizontal dashed line represents CpG 
719 
enrichment of 1. e, Principal component (PC) analysis of cfDNA methylation from 24 plasma 
720 
cfDNA samples from healthy donors and 24 plasma cfDNA samples from pancreatic 
721 
adenocarcinoma patients, using the 1 million most variable windows by Median Absolute 
722 
Deviation (MAD) (300bp) genome-wide. Left: PC2 against PC1; right: PC3 against and PC1. f, 
723 
Percentage of variance explained by each PC. 


 
 
 
 
 
 
 
 
 
 
724 
725 
726 
727 
728 
729 
730 
731 
732 
733 
734 
735 
736 
737 
738 
739 
740 
741 
742 
743 
744 
745 
Extended Data Fig. 5 Methylome analysis of plasma cfDNA distinguishes early stage 
pancreatic adenocarcinoma patients (PDAC) from healthy controls. a, Difference in plasma 
cfDNA methylation against difference in tumor DNA methylation for each overlapping window 
(n=547,887). Plasma cfDNA methylation difference is log10 fold change from pancreatic 
adenocarcinoma patients to healthy measured by cfMeDIP-seq. Tumor DNA methylation 
difference is delta-Beta from primary pancreatic adenocarcinoma tumor to normal tissue, 
measured by RRBS. Blue line: trend line, correlation determined by Pearson’s correlation. b, 
Scatter-plot showing the DNA methylation difference for each overlapping window. X-axis 
shows the DNA methylation difference for the primary pancreatic adenocarcinoma tumor versus 
normal PBMCs from the RRBS data. Y-axis shows the DNA methylation difference for the 
plasma cfDNA methylation from pancreatic adenocarcinoma patients versus healthy donors from 
the cfMeDIP-seq data. Correlation determined by Pearson’s correlation. c, Genome browser 
snapshot of RRBS and cfMeDIP-seq signal across a representative chromosomal region selected 
from four candidate regions (chr8:145,095,942-145,116,942) using reference genome hg19. 
RRBS tracks show the methylation signal for the LCM tissues from pancreatic adenocarcinoma 
tumor cases and the matching normal tissue, from the same patient, shown in the same order. 
Each colored block represents differentially methylated CpGs (DMCs), with yellow representing 
hypermethylated and blue representing hypomethylated. cfMeDIP-seq tracks show the 
methylation signal (RPKMs) detected in the cfDNA, with cases representing plasma from the 
same pancreatic adenocarcinoma cases and controls corresponding to plasma from age and sex 
matched healthy controls. For the cfMeDIP-seq tracks, green and blue peaks indicate the 
methylation signal (RPKMs) detected in the cfDNA. 


 
 
 
 
 
  
 
 
746 
Extended Data Fig. 6 Circulating cfDNA methylation profiles can identify transcription 
747 
factor (TF) footprints and infer active transcriptional networks in the tissue of origin. a, 
748 
Expression profile of all TFs (n=42) that were characterized as binding in healthy controls across 
749 
53 human tissues from the Genotype-Tissue Expression (GTEx) project. Several TFs 
750 
preferentially expressed in the hematopoietic system were identified (PU.1, NFE2, and GATA1). 
751 
b, Expression profiles (ssGSEA scores) of all TFs with hypomethylated motifs in controls (n=42) 
752 
are overexpressed versus those of 1,000 random sets of 42 TFs across GTEx whole blood data 
753 
(p<2.2e-16, Wilcoxon’s Rank Sum Test, two-sided). c, Expression profile of all TFs (n=52) 
754 
characterized as binding in pancreatic adenocarcinoma patients. Several pancreas-specific or 
755 
pancreatic cancer-associated TFs were identified. Moreover, hallmark TFs that drive molecular 
756 
subtypes of pancreatic cancer were also identified. d, Expression profile (ssGSEA scores) of all 
757 
TFs with hypomethylated motifs in cases (n=52) are overexpressed versus those of 1,000 random 
758 
sets of 52 TFs in normal pancreas (GTEx data) (Wilcoxon Rank Sum Test, two-sided test, p­
759 
value<2.2e-16). e, Expression profile of all TFs with hypomethylated motifs in cases (n=52) are 
760 
overexpressed versus those of 1,000 random sets of 52 TFs in pancreatic adenocarcinoma tissue 
761 
(TCGA data) (Wilcoxon Rank Sum Test, two-sided test, p-value<2.2e-16). For violin plots (b, d, 
762 
e) the ends of the boxes and middle line represent the lower and upper quartiles and median, 
763 
respectively. Whiskers represent 1.5 times the interquartile range (IQR). Outliers are excluded. 
764 
Rotated kernel densities are also displayed. 
765 
Extended Data Fig. 7 Quality controls for cfMeDIP-seq from circulating cfDNA from 
766 
multiple cancer types. a, c, e, g, i, k, Specificity of reaction and b, d, f, h, j, CpG enrichment 
767 
score for each sample per cancer type. Horizontal dashed line represents CpG enrichment of 1. 


 
 
 
 
 
 
 
 
 
 
 
 
768 
Extended Data Fig. 8 Comparison of plasma cfDNA DMRs with tumor DMCs. a, Yield of 
769 
cfDNA extracted per mL of plasma from healthy donors (n=24), bladder cancer (n=20), renal 
770 
cancer (n=20), lung cancer (n=25), breast cancer (n=25), pancreatic cancer (n=24), colorectal 
771 
cancer (23) and AML (n=28). Horizontal bars represent the mean, with dots representing 
772 
individual samples. b-h, Scatterplot showing the DNA methylation difference for all overlapping 
773 
windows in pancreatic cancer (PDAC) (n=245,980 windows) (b), acute myelogenous leukemia 
774 
(AML) (n=206,735 windows) (c), bladder cancer (BLCA) (n=193,943 windows) (d), breast 
775 
cancer (BRCA) (n= 204,623 windows) (e), colorectal cancer (CRC) (n= 210,645 windows) (f), 
776 
lung cancer (LUC) (n= 193,043 windows) (g), and renal cancer (RCC) (n= 198,390 windows) 
777 
(h). X-axis shows the DNA methylation difference for the primary tumor (TCGA data) versus 
778 
normal PBMCs. Y-axis shows the DNA methylation difference for the plasma cfDNA 
779 
methylation for each cancer type versus healthy controls from the cfMeDIP-seq data. Blue line 
780 
shows the trend line, correlation determined through Pearson’s correlation. 
781 
Extended Data Fig. 9 Circulating plasma cfDNA methylation samples used to distinguish 
782 
between multiple cancer types and healthy donors. a-b, Pathology stage (AJCC/UICC 7th 
783 
Edition) breakdown by tumor type for samples in the (a) training set and in the (b) validation set. 
784 
Pancreatic cancer: PDAC, breast cancer: BRCA, lung cancer: LUC, colorectal cancer: CRC, 
785 
bladder cancer: BLCA, renal cancer: RCC, Non-small cell lung carcinoma: LUC (NSCLC) and 
786 
small cell lung cancer: LUC (SCLC). 
787 
Extended Data Fig. 10 Characterization of hypermethylated regions from cfDNA that are 
788 
not methylated in leukocytes. a, Violin plots for the DNA methylation (Beta value) of 38,352 
789 
regions in normal blood cells selected based on low DNA methylation levels using IHEC whole 


 
 
 
 
 
 
 
790 
genome bisulfite sequencing data. For violin plots, the ends of the boxes and the middle line 
791 
represent the lower and upper quartiles, and medians, respectively. Whiskers represent 1.5 times 
792 
the interquartile range (IQR). Outliers are excluded.  Rotated kernel densities are also displayed. 
793 
b, Volcano plot representing the regions with low DNA methylation levels in normal blood cells 
794 
that overlap with hypermethylated regions in the plasma cfDNA for PDAC (n=3,146 CpG sites) 
795 
relative to normal tissue, and RCC (n=2,767 CpG sites), BLCA (n= 3,286 CpG sites), BRCA (n= 
796 
6,836 CpG sites), CRC (n= 8,360 CpG sites) and LUC (n= 5,239 CpG sites) relative to PBMCs. 
797 
X-axis represents the delta-Beta value in methylation in tumor data from TCGA for cancers other 
798 
than PDAC and RRBS for PDAC. Y-axis represents -log10 q-values (Benjamini Hochberg False 
799 
Discovery Rate, BHFDR).  
800 
801 
802 


a 
Abundance of ctDNA (%) 
log10 of coverage 
Probability of detection (1+ epimutations
Number of DMRs/epimutations 
b
c 
100,000
Expected DNA methylation signal
(RPKM) within DMRs 
1,000,000 
human ctDNA (% of total cfDNA) 
2.5 
2.0 
1.5 
1.0 
0.5 
FDR at 5% 
r2=0.99 p<0.0001 
1 
10 
100 
1,000 
10,000 100,000 1,000,000 
10,000 
1,000 
100 
10 
1 
0.0 
Input 
IP 
Samples
Observed DNA methylation signal (RPKM) within DMRs 
PDX1 
PDX2
Dilution (% of CRC DNA) 
100% 
10% 
1% 
0.1% 
0.01% 
0.001% 


●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●●
●
●
●
●
●
●
●●
●●
●
●●
●●
●
●●●●
●
●
●●●●
●
●●
●
●
●
●
●
●
●●●
●
●
●●
●●
●
●
●
●●
●
●
●●
●
●
●●
●
●●
●
●
●
●●
●●●
●
●
●●
●●
●●
●●
●●●
●●
●
●●
●
●
●
●●
●
●
●●
●
●●●
●●
●●●
●
●●●
●●
●●
●●
●
●
●●
●●
●●
●
●●
●●
●●
●●
●●●
●●
●
●
●●
●
●●●
●●
●
●
●
●
●●
●●●●
●
●
●
●●
●
●●●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●●
●
●
●●●
●
●●●
●●●
●●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●●●
●
●
●●●●
●
●
●
●
●●
●●
●●
●●●
●●
●
●
●
●●
●
●
●●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●●
●●
●●
●●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●●
●
●
●
●
●●
●●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●●
●
●
●●
●●
●
●●
●
●●
●●
●
●
●
●
●
●●
●
●
●●
●
●●
●●
●
●
●●
●●
●
●
●
●●
●●
●
●●●
●
●
●
●
●
●●
●●
●●
●●
●
●●
●
●
●
●●
●●●
●●
●
●
●●
●
●●
●●
●●
●
●●
●●
●●●●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●●
●
●
●●
●
●
●●
●
●
●●
●
●●●
●●
●
●●
●●
●●
●
●
●
●●
●
●●
●●
●●
●●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●●
●●●
●
●
●
●
●
●●
●●
●
●●
●●●
●
●●
●●
●
●
●
●●
●
●
●
●●●
●●
●
●●
●
●●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●●●●
●●
●●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●●●
●
●●
●
●
●●
●●
●
●
●●
●●
●●
●●
●●
●●●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●●●●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●●
●
●●●
●
●●●
●
●●
●
●●●
●●
●
●
●●
●
●
●
●●
●
●●
●
●
●●●
●●
●
●●●●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●●●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●●
●
●
●
●
●
●●
●●
●
●●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●●●
●
●
●●
●
●
●
●●
●
●●
●●
●
●
●
●
●
●●
●●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●●
●
●
●
●
●●●
●
●
●
●
●●
●
●●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●●
●
●
●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●●●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●●
●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●●●
●
●
●
●
●●
●
●●
●
●
●●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●●●●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●●●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●
●
●
●●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●●●
●
●
●●
●
●
●
●●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●●
●●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●●●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●●
●●
●
●●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●●
●●
●●
●
●●●
●
●
●●
●
●●
●
●
●●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●●
●●
●
●
●
●
●●
●
●
●
●●●●
●
●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●●●
●
●
●
●
●●
●●●
●
●
●
●
●●
●●
●●
●●
●
●●
●●
●
●
●●
●●●
●
●●
●
●●
●●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●●●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●●
●●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●●●
●●
●●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●●
●
●●
●●
●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●●
●
●
●
●
●●●
●●
●
●
●
●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●●
●●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●●
●
●
●
●
●
●●●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●●●
●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●●●
●
●
●
●●
●
●●
●
●
●
●
●
●●●
●●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●●●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●●
●●
●
●
●●●
●●
●
●
●
●●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●●●●●
●●
●
●
●●
●
●●●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●●
●
●
●
●
●●●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●●●
●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●●
●
●
●
●●
●
●●●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●●
●
●
●●
●●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●
●●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●
●●
●
●
●
●
●
●●
●●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●●
●●
●
●
●●
●
●
●
●
●●●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●●●●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●
●●
●
●●
●
●
●
●●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●●
●
●
●●●
●
●
●
●
●
●
●
●●●
●●
●
●
●
●●
●
●●●●
●
●●
●●●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●●
●
●●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●
●●●
●●
●
●
●●●
●●●●
●●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●●
●
●●
●
●
●
●●
●●
●●●
●●
●
●●
●
●
●
●
●
●●●
●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●●
●
●
●
●●
●●●●●●
●
●●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●●
●●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●●
●
●
●
●
●
●●
●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●●
●
●
●●
●●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●●
●
●
●
●●●
●
●
●
●●
●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●●
●
●●
●●
●
●
●
●
●●
●
●
●
●●
●●
●
●●●
●●
●●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●●
●●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●●●
●
●
●●
●●
●
●●
●
●●
●
●●●
●●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●●●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●●
●
●
●
●
●●●
●
●
●
●●●
●
●
●
●
●●
●
●
●●●
●
●●●
●●●
●
●●●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●●
●●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●●
●●
●
●
●
●●●
●●●
●
●●
●
●
●
●●
●
●
●
●
●
●●●
●●
●●●
●●
●
●
●
●●
●
●
●●
●●
●●
●
●●
●●
●
●
●
●●
●
●
●
●
●●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●●
●●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●●●●
●
●
●●
●
●●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●●●
●
●
●
●
●●
●
●
●
●●
●
●●●●
●
●●
●●
●●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●
●●●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●●●
●●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●
●●
●●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●●●
●
●
●
●
●
●●●
●
●●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●●●
●●
●
●●●
●
●●
●●
●●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●●
●●
●
●●●●
●●●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●●
●
●●●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●●
●
●
●●
●●
●
●
●
●
●●●
●●●
●
●
●●
●
●
●●
●
●●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●●●
●
●●●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●●
●●
●
●●
●
●
●
●●
●
●
●●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●
●
●●
●
●●
●
●●●●●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●
●
●
●
●●
●
●
●
●●
●
●●
●●●●
●
●
●●
●
●
●●
●
●●●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●●
●
●●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●●●
●
●●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●●
●●●
●●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●●●●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●
●●●
●
●●
●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●●●
●●
●
●
●
●
●●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●●
●●
●
●●
●
●
●
●●
●
●
●●
●
●●
●
●
●●●●●
●
●
●
●
●●
●
●
●
●
●
●
●●●●●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●●●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●●●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●●
●●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●●
●●●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●●
●●
●●
●●●
●
●
●●
●●
●
●
●
●●●●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●●
●●●●
●
●
●
●
●
●
●
●●●
●
●
●●
●●●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●●●
●●●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●●●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●●●
●
●●
●
●●
●
●
●
●
●
●●
●●●
●
●
●
●●●
●
●
●
●●●●
●●
●
●
●
●
●
●●
●●
●
●●
●●
●●
●
●
●●●●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●●
●
●
●
●●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●●
●
●●●
●●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●●●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●●●
●●
●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●●●●
●●
●
●
●●
●
●
●
●
●●●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●●
●●●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●●
●
●●
●●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●●
●
●
●●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●●
●
●●●●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●●
●●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●●
●
●●
●
●●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●●●
●
●
●●
●●
●
●
●
●
●
●●
●●
●●
●
●
●
●●
●●
●
●
●
●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●●
●●
●
●●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●
●
●●●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●●
●
●●●
●●
●
●
●●
●
●●
●
●
●
●
●●
●●
●●
●
●
●
●
●
●
●
●●
●●●
●●
●
●
●●
●
●●
●
●
●●
●
●
●●●
●
●
●
●●
●
●●
●
●
●
●
●●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●●
●●
●
●
●
●
●●
●
●
●
●●
●
●●
●●
●
●
●●●
●
●
●●●
●
●●●
●●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●●
●
●●
●
●
●●
●●
●
●●
●●
●
●
●
●
●●
●●●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●
●●
●●
●
●●
●
●●
●
●●●●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●●●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●●●
●●
●●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●●
●
●
●
●●●
●●●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●●●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●●●
●
●
●●
●●
●
●
●
●
●●●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
Hypermethylated in primary tumor
Hypermethylated in plasma cfDNA
Hypomethylated in primary tumor
Hypomethylated in plasma cfDNA
Hypermethylated in primary tumor
Hypomethylated in plasma cfDNA
Hypomethylated in primary tumor
Hypermethylated in plasma cfDNA
A =
B =
C =
D =
Legend
a
b
Plasma cfDNA (cases vs. controls)
14,716 DMRs
Significance (-log10 q-values)
DNA methylation (log2 fold change)
c
Plasma cfDNA
14,716 DMRs
Loss
Gain
5.0
3.0
2.0
- 2
-1
0
1
2
1.0
0.0
Healthy 
controls
PDAC
Early stage pancreatic
adenocarcinoma patients
(n=24)
e
d
Z-scores
15
10
5
0
-5
-10
Overlap plasma cfDNA (cases vs controls) 
and tumor DNA (primary tumor vs normal tissue)
Overlap plasma cfDNA (cases vs controls)
and tumor DNA (primary tumor vs normal PBMCs)
Z-scores
40
20
0
-20
●
●
●
●●●●
●
●●●
●
●
●
●
●●
●
●
●
●
●
●●●●
●
●●
●
●
●
3.39e−47
0.16
1.43e−22
0.00956
20
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●●●
●
●●
●●
e^−1300
1.64e−36
6.12e−82
7.56e−09
60
1.64e-36
6.12e-82
7.56e-09
3.39e-47
ns
1.43e-22
0.00956
Surgical resection: tumor 
and normal tissue
Plasma: circulating cell-free DNA
Age and sex matched
healthy controls (n=24)
Plasma: circulating cell-free DNA
RRBS: tumor and 
normal tissue
cfMeDIP-seq: 
cell-free DNA
4.0
Row 
z-score
<1e-745
A
B
C
D
A
B
C
D


150 
100 
0 
<1e-745 
8.64e-22 
● 
● 
● 
● 
● 
● 
● 
2.09e-14
2.7e-36 
a 
PDAC 
AML 
BLCA 
BRCA 
b 
Discovery set (n=189) 
●
● 
●
●
● 
● 
Z-scores 
● 
● 
●●●
●
●●
●
●
●
●
● 
● 
● 
● 
●●●●●●● 
0 
40 
<1e-745 
1 
1.35e-77 
 
 
 
 
●
●●
●●●●
●
●
●●●
●
●●●
●●●
●
●
●
●
●
●●●
●
●●
●●
●
●
30
20 
100 
80 
1.87e-1
9.21e-08 
60 
●● 
●● 
●
●●
●
●
● 
● 
●●● 
●● 
● 
● 
2.82e-53 
6.87e-165 
1.43e-22 
<1e-745 
100 
●● 
●● 
●
● 
● 
●
●
●
●●● 
● 
● 
3.53e-295 
6.86e-10 
7.78e-20 
9.22e-47 
0.8 
60 
20 
1.0
150
40
80 
<1e-745 
40 
Sensititvity 
AUROC
50
10 
5.36e-108 
0.6
20 
0 
0 
● 
0 
−10 
1.12e-43 
0.4 
9.68e-143 
AML 
BLCA BRCA 
CRC 
LUC Normal PDAC 
RCC 
−50 
Sample type (class)
−20
−20 
−20 
A
B
C
D 
A
B
C
D 
A
B
C
D 
A
B
C
D 
c 
Validation set (n=199) 
CRC 
LUC 
RCC 
Legend 
●
●●● 
● 
●
●
●
●
● 
● 
●●●● 
●●● 
●● 
2.05e-07 
1.42e-19 
ns 
8.64e-22 
<1e-745 
1.65e-38 
● 
● 
● 
● 
● 
● 
● 
● 
3.09e-50 
e^757 
20 
Hypermethylated
100 
in primary tumor
A = Hypermethylated
in plasma cfDNA 
Hypermethylated
in primary tumor
B = Hypomethylated
in plasma cfDNA 
Hypomethylated
in primary tumor
C = Hypomethylated
in plasma cfDNA 
80 
10 
60 
Z-scores
0
40
50 
20 
0 
−10 
−50 
Hypomethylated
in primary tumor
D = Hypermethylated
in plasma cfDNA 
−20
−20 
A
B
C
D 
A
B
C
D 
A
B
C
D 
d 
Early stage PDAC (n=15) 
Late stage PDAC (n=32) 
Early stage LUC (n=32) 
Late stage LUC (n=23) 
Sensititvity 
1 - Specificity 
1 - Specificity 


 
−50
−40
−30
−20
−10
  0 10
●
15
10
−5
  0
  5
 10
●●●
●
●●
●
●
●●
−6
4
−2
 0
 2
●
● 
●
● ● ●
● 
● 
● 
● ●● ● 
● 
● 
● 
●
● ●● 
● 
●
● ● ● ● 
● 
● 
● 
● 
●
●●
●●● 
●●
● ● 
●
● ● 
●
● ●● ●● ●● 
● ● 
● 
● 
● ●● 
● ●●
●
● 
● ● 
●
●●
●●
● 
● 
●
●●
●
● ● 
● ● 
●●● 
● 
● ● 
●●
●
● ● 
● ● 
●
● 
● ● 
●
● ● 
● 
● 
● ● 
● 
● 
● 
● 
● ●
● 
●
●
●● ●●
● 
● ● 
● ● 
●
● ● ● 
●
● ● 
●
● ● 
●●
●● 
● 
● ●
●● ● 
●
● ● 
● ●
● 
● 
●
● ● 
● 
● 
● ● 
● 
● 
●
●
● ● 
● ● 
● 
● 
● ● 
●● 
● 
● 
●
● 
● ● ●
● 
● 
● ●
●●
● ● 
● ● 
● 
●●● 
●● ● 
● ● 
●● ● 
● ● 
●● 
● 
●
● ●
● ● ● 
● 
●●● 
● ●●
● ● 
●●
●
● ● ● 
● ● 
●● ●● 
● ● 
● ● 
● ● 
●
● ● 
●
●●
● ● ● 
● ● ● 
●
●
● ● 
●
●
●● ●● 
●
●
●
●
●● 
● ● 
● ● 
●●
● ● 
● 
●
●
●●
●● ● 
●●
● ● 
●
●● ● 
●
●●
●
●
● 
●●●
● 
● ●● 
● ● ●
● 
● 
● 
●
●●● ●
●● ● 
●● ●●
● ●
● 
● ● 
●
● ● ●
● ●
●
●●
● 
●
● ● 
● 
● 
● ● 
●
● ●●●● 
●
●●
● ● ● 
●
● ● ● 
●
●
● ● 
● ● 
●● ● 
● ● 
● 
● 
● 
● 
● ● ●●
●
● ● ● 
●
●●
● ● 
●●●
● 
●
● ●
●
●●
● ● 
●●
●●
● ● ●●
● 
●●●
● 
●
● 
●
● 
● 
● 
● 
●
● 
● ● 
●
●
●
● ● 
●
●
●●
●
● ● 
●●
● ●
● 
● ●
● 
● ● 
● ● 
●
●●
● ●● ● 
● ● 
● ● 
●●
●●
●
●
●
● 
● 
● ●
● 
●● ● ● 
● ●
●● ● 
● ● 
●
● ●●
● 
●
●
● 
● 
● 
●● ● ● ●● ●●
● 
●
● 
●●
●
● ● 
● 
● 
● 
● 
●
● ● 
● ● 
● ●●
●
● 
●● ● 
● ●
● 
● ● 
● ●●●
● 
●● 
●● 
● 
● 
●
● ● 
● ● 
● ●● 
●●
●● ● 
●
●● 
●
●
●
●
●
●
●● ● 
● 
●●
●
● ● 
●●
●● ● 
● 
●
● ● 
● 
● 
●
●
●●
●●●
●
●●
●●
●
●
●●●●
●●
●
●●
●
●●●
●
●●●
●●
●●●●
●
●●●
●
●●
●
●●●●
●●●
●
●●
●
●●●
●
●●
●●●●
●
●●●
●●
●●
●●
●
●●●
●●
●
●●
●●●
●●●●
●●●
●
●
●●●
●
●●●
●
●
●
●●●
●●
●
●
●●
●
●●
●
●●●
●
●
●●
●●
●
●●
●●
●●●●●
●
●●
●
●
●
●●●
●●
●
●
●●
●●
●●●●
●
●●
●●●
●
●
●●
●
●●
●●●
●●
●●
●●
●
●
●
●●●
●
●●●
●●
●●
●●●
●
●●
●●
●
●●●
●●
●●
●
●●
●●●
●
●
●●●
●●
●
●●
●●
●
●
●●●
●●
●●●●●
●
●
●
●
●
●●●●
●●●
●●
●●
●
●●●●●
●●
●●●
●
●●
●
●●
●●
●●
●●
●●●
●●●●
●
●●
●
●●
●
●●
●●
●●●●●
●●●
●●
●
●
●●●
●
●●
●●
●●
●●
●
●●
●●
●●
●●
●
●
●●●●
●●
●●●
●●
●●
●
●●
●
●●
●●
●
●●●
●●
●●●●
●●
●●●
●●
●
●●●●
●●●
●●
●
●●●
●●
●●
●●●
●●
●
●
●●
●●
●
●●
●●
●●
●●
●●
●●
●
●●●
●●
●
●●
●●●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●●
●
●●
●●
●
●
●
●●
●●
●●
●●●
●●●
●
●●●
●●
●●
●
●
●●
●●●
●●
●
●●●
●
●●●
●●
●
●
●
●
●
●●
● 
●
●●●
●●
●●
● ●
●
●●
● ●
●
● 
●●
●● ●
●●
●
●
● ●
●●●
●●
●
●
●●●
●●
●●
●●
●● ●●● ●
●● 
●●●
●
●●
●● ●
●●●● ●●
●
●
●● 
●●●
● ●● 
● ● ●●●
●
●●● 
● 
●●●
●●
●
●●
●●
●●
●●
●●
●
●
●●
●
●●●
●●
●
●●● ●
●●
●
●●● ● 
● ● 
●●
● ●
●
●●
●●
●●
●
●●
●● ● ● ● ● ●
●● ● 
●●
●●●●
●
●●
●●●●●
●●
●● ●
●●●
●●
●
●●
●
●
●●
●●
●●●●
●
●●
●
●●
●● ●
●
●●●
●●
●
●
●
●
●●
●
●●●
●
●
●●●
●●
●●
●● 
● ●●●
●
●●●
●
●●●
●
●●●
●● ●● ●
●●● ●●
●● ●
● 
● ● 
●●●
●
●
● ● ●●●
●●
● 
●● ●●● 
● ●● ● ●● ● ● 
●
●●
●
●●
●●●● ●
●● ●
●●
●● ● 
●● ● ●●●
● ●
● ● 
●●
●●●●●
●
●
●●
●●
●
●
●
● ● ●● ● ●
●●
●
●
●●
●●
● 
● ● ● 
● ● ●
●●●●
● ●●
●●●
●●
● ● 
●●
●●
●●
●
●●●
●●
●● ●●
● 
● 
●
●●
●
● 
●●
●●
● 
●
● 
●●●
●●
●●
●
● ● ● 
●
● ● 
●
●●
●
●● ● 
● 
● 
● ● 
● 
●
●
●●●
●●
●●●
● ●● ●
●●●
●●●
●●
●
●●
●
●●●
●
●
●
●
●
●●
● 
●
●
●●●●
●
● 
●
●●
●●●
● ●●
●●●●●
● ●●
●●
●●
● 
● 
● 
●●●
●
● 
● ● ●●
●● ● 
● 
●
● ●●● ●
●●●
●●
● 
● ●
●●
●●
●●●
● ●
●●
●
●● ●
●●●
●●
●●●●
● ●●
● 
● 
● ●●●
●
● ● ● ●●●
●●
●
●●
●
●●
●
●
●●
●
●
●
●
●●
● ●
●
●●
●●●
●●
●
● ● 
●
●●● 
● ● ●●
● ● 
●
●●
●●
●●● ●●●●●
●●
●
●●●
●●●
●
● 
●●● ●●
●●●
●
●
●●●
●● 
●●●
●●
● ●
● ● ● ●● 
●●●
●● 
●●
●●
●
●●●
●
●●
●
●●
●●●
●
●
● 
●●
●● 
● ●● ●
●●
●
●●
●●● 
●●
●
●●●● ● ●●
● ● 
●●
●
●●
●
●●
●●
●●●
● ●
●
●● 
●
● ●
●●●
●●
●●●●●
● ●
●
●● ●
●
●
●●
●
●
●●
●● ●●
●●
●
●●●
●●● ● 
● ●●
●● ●●● ●●
●● ● 
● ●●
●● 
●
●●●
●
●
●
●●
●● ●●
●
●●●
●
●
●●
●●
●
●●
●●●
●●
●●
●●
●●
●●
●●●
●●●
●●
●
●
●
●●●
●● ●
●●●
●
●●●
●●
●
●●
●●
●
●● ●
●●●
●●●
●
●●
●●
●●●
● ● 
●●
●
●●
●●
● 
●●
●
●
●● ●
●●
● 
● ● 
● ● ●●
●
●
●● ● ● ●●
●●
●●
●
●
● ●● ● ● ●
●●● ●●
●● 
●●● 
● 
●●
●●
●●
●● 
● 
●
● 
● ●●
● 
●
●
●●●●
● ●
●●●●●●
●
●●
●●●
●●
●
●
●●
●●
●
●●●
● 
●
●●●
●● 
●
●
●
●● ●
●●
●●●
●●
●●
●●
●
●
●●●
●
●●
●●
●
●●
●
●●●
●
● ●●
●●●
●●
●● ●
●●●
●●
●
●●
●
●●●
●●
●
●●●
● ●
● ● ●
●
●
●
●●●
●
● ●
● 
● 
● ● 
● 
●●
●●●● 
● ●●
●●●
●
●
●
●●●●
●●●●
● ●
●
●●●
●
●
●●● ●
●●●●● ●●
●●
● ●
● 
●
●● ● 
●●
●
●●
● ●
●●
●● ●
●●
●●●
●●●
●
●
●●●
●●
● 
● 
●
● ● 
● ● ●
●
●●
● ● ● ● 
● 
●●●
●● 
●
●
●●
●
●
●● 
●● ● ●●
●● ●●●
●
●●
●●
●●
●●●
●
●●●
●
●●
● ●●
●●●
●
●●
●●
●●
●●
●
●
●
●●
●
● 
●●
●●● 
●
●●●●
●●
●
●
●● ●
●● ● 
● ● 
● 
● ●●
●●
●
●● ●●
● ● ● 
● ● 
● 
●
●●●
●●●●
● 
●● 
● 
● ● ●
●
●●
●●
●●
● 
●
●●●●
●
●●● ●●
●● ●●
● ●●
● ● ●
●●
● ●●●●
● 
●●●
●
●●
●
●●●
●●
●● ●●
● ● ●
●
●●
● ●
●●● 
● 
● ●
● 
●
●
● ● ● ● ● 
●
●
●●
●●
●●
●●
●
●●
●●
● ●● 
● ●● ●●●
● ●● ●●●● ● ●●
● 
● ●●●
●●
●
● ●
●●●●
●●●
●
●
●●
●
●●
●●●● ●●● ●
● 
● 
● ●
●● 
●●
●●●
● 
●●
●● 
● 
● ●
●●● ● ●
●● ● ●
●
● 
● ● 
●
●●
●
●●
●
●
●●
●●
●
●
●●
●●●
●●
●●
●
● ●
●
●●
●●
●
●●
●●
●●
●●
●● ●●●
●●
●●
●● ●● ●● ● 
●● ● 
●
●
●●
●
●●●●
●●
● ● ●●●●
●
●●
●●
●●
●
●●●●
●
●● ●●
●
●●●
●
●●
●
●●
●
●●
●
●●
●
●● ●
●●
●●
●●
● 
●
●
●●
●●●
●●
●●
●
●●
●●
●●
●
●●
●
●●
●●●
●
●
● ●●
●
●
●●
●
●●
●●●
●
●
●●
●
●●
●●
●●
● ●
● 
● ●● 
●
● ●
●
●●
●●●
●
●
●● 
●●
●●
● ●●
●●
●
●●
●●
●
●●
● 
●●●
●●●
●●
●●
●●
●
●●
●●
●
●●●●
●●
●
●
●
●●
●
●
●●●
●
●
●●
●●
●●
●●●
●
●●
●
●●
●●●
●●
●●
●
●●●
●
●●
●
●●
●
●
●
●●
●●●
●●●
● 
● 
●●●
●
●●●
●
●●
●●
●● 
●
●●
●●
●●
●
●●
●
●●●
●
●
●●
●● ●●●
● 
●
●●
●●●
●●
●
●●
●● 
●
●
●●●
●
●●
●
●
●●
●
●●
●●●
●●●
●
●
●●
●
●●
●●
●●
●●●
● ●
●●●
●●
● ●
●●
●
●●
●●
●●
●●
●
●●●
●●
● ● 
●●
●
●●
●●●
●● 
●●
●●●●
●●
●●
●●
●●
●●
●●
●●
●●●
●●●
●
●
●
●
●
●●
●●●
●
● ●●●
● ●
●●●
●
● 
●
●
●●●
●
●●●
●
●●●
●●●
●●●
●●
●●
●
●
●●
●
●●
●●
●
●●●
●●
●●●
●●●
●
●●●
●●●●●
●●
●
●
●●
●
●
●●●●
●
●●●
●●●
●
●●
●●●●
●
●●
●
●●
●
●
●●●
●●●
●●●
●●●
● 
●●
●●● ●●
●●
●●●●
● 
●
●●
●●
●● ●
●●●
●
●●
●● 
●●●
●●
●
●
●
●
● ●●
● 
●
●
●
●
●●
●●
●●●●●
●
●●
●●●
●
●
●
●●
●
● 
●●●
●
●●
●
●●●
●
●●
●
●●●
●
●
●●●
●●●
● 
●
●●
●●
●●
●●
●●
●
●●
●●●
●
●●
●
●●
●●
●
●●●
●●●● ●
●● 
●●
●●
●●
● ● 
●● 
●
●●
● ●●●
●
●●
●●●
●
●●●
●●●
●●
●●
●
●●●
●
● ●
●● 
●●●
●●
●●
●●
●●
●●●
●
●●●
●●●
●●
●●●
●
●
●●
●
●●
●●
●●
●●●●
●
●●
●●
● ● 
●
●
●●
●
●
●
●
●●
●●●
●●
●
●●
●●
●
●●
●
●
●●●
●
●●
●●●
●
●
●●●
●
●●●●●
●
●
●●●●
●●
●
●
●
●●●
●
●●●
●
●●●●
●
●
●●●●●
●
●●
●
●
●
●●
●●
●
●● 
●
●● 
● ● ● 
● ● 
● 
● 
● 
● 
●●
● 
●
●
●●●
●●
●
●●●●
●●
● ●
● 
● 
●●
●
● 
● 
● 
●●
●●
●●
●●●
●●
●●
●●
●
●
●●
●●
●●
●●
●●
●●●●
●●
●●
●
●●
●
●● ● ●
●●
● 
●
●●
●
●●●
●
●●
●●
●●
●●
●●●
● 
●
●
●●
●
●●
●●●
●●●
●●
●
●●
●●
●
●●
●
●●
●
● ●●
●
●●
●
●●
● ● 
●●
●
●●
●
●●
●●
●
●●
● ●
●●
●●
●
●●●●
●
●●
●●
●
●
●
●●●
●
●●●●
●● ●●
●●●
●●●
● 
●● ● 
● ●● ●●
●●●●
●
●
●
●●
●
●
●● 
●
●
●
●
●●●
●
●
●●
●●
●
●●●
●
●●●●
●●●
●●●
●●
●
●
●
●●
●
●
● ● ●
● ● 
●●
●
●
●
●●●
●
●●●
●●●
●●
●●●●
●
●●
●
●●●
●●●●
●
● ● 
●●
●
●
●
●
●●
●
●
●
●●
●●●
●●●
●●
●●
●●
●
●●
●●●
●
●●
●●
●●
●●
●
●●
●●●
●
●● ●
●●
●●
●●
●●●
●●
●● ● ●
●●● 
●● ●
●● ● ● ● ●
●● ●
● 
● ●●
● 
●
● ●●
●● 
●
●
●●●●●
●●
●
●
●●
●●
●●●
●●
●●
●● ● 
●●
●●
●●
●●
●
●
●●●
●
●
●●●
●
●●●
●●
● ● 
●
●●
●
●●
● ● 
●●
●
● ● 
●
●●
●
●
●●
●
●
●●
●●●●
●
●●
●●●●
●
●●
●
●●
●
●●
●●●●
●●
●
●●●
●●●
●●
●●
●
●
●●●●
●
●●
●●●
●●
●●
●●●
●
● ● 
●
● 
●● 
●●●
●
●●
●
●●
●
●●
●
●●
●
●
●●
●
●●●
●●●●
●●●
●●
●●
●●
●●●●
●
●●
●
●
●●●●
●●●
●
●●
●
●
●●●●
●
●●
●●
●●
●●●
●●
●●●
●●
●●
●
●●
●
●●
● ●● ● 
● ● 
●
●●
●●
●
●●
●
●
●
●●●
●
● ● 
● ●●
●●● ● 
●
● ● ●●
●●
●
●●
●
●
●
●
●●
●● ●
●●
●●
●
●
●●
●
●
● ● 
●●
●
●●●
●●●
●
●●
●●
●● ● 
● 
● ● 
●● 
● 
● ●
● 
● ● ●
●
●
●●●
●●
●
●●
●●
●●
●●●●
●●●
●
●
●
●
●
●
●
●
●
● ● ●
●● ●
● ●●
●
●●
●
● 
●●
●
●●
●
●●● ●
●
●●●
● 
●
●
●●
●
●●● ●● 
●● ●●
●●
● ● 
●
●
●●
●
●●
●●●
●
●
●
●●
●●
●● 
●
●●
●●●
●
● ● 
●●●●
●
●●
●
●●●
●
●● 
●●●
●●
●●
●●
●●●
●●
●
●
●●
●●
●
●
●●
●●
●●
●●
●●●
●
●
●●
●● ● 
●●●
●
●●
● ●● 
●
●●●
●●
●●
●●●
●●
●●
●●●
●
●●●●●
●
●●●
●
●●
● ● 
●
●
●●●
●●
●●
●
●●
●●
●●●● ●
●
●● 
●
●
●●
●●●
●●
●●●
●
●
●●●
●●
●●
●●
●
●
●●
●●
●●
●●
●
●●●●
●
●●● ●
●●
●●
●
●●
●
● 
● ●
● ●●●
●●●
●
●●
●●●
●●
●
●●
●●
●
●●●
●
●
●●●●●●
●●
●●
●●
●●
●●
●●
●●●
●●
● ● 
●
●
●●
●
●
●●
●● ●
●●
●●
●●
●●
●●● 
●
●●●
●●
●●
●●
●●●
●●
●●
● 
●●
●●
●
●●●
● ●
●●
●● ● 
●
●●
●●● ●●
●●
●●
●● 
●●●
●●●
●
●●● ●●
●
●●
●
●
●●●●
●●●●
●●
●● 
●
●●
●●●
●
●●●●
●●
●●●●
●●
●
● ●
●●
●●
●
●●
●●●
●
●
●
●●
●●
●●●
●●
●
● ●
●
●
●
●
●
●●
●
●● 
●●
●●
●●
●● ●●●●
●●
●●
●
●●
●●
●●
●●
●●
●●
●●●
●●
●●
●●
●●
●●
● 
●
●●
●●
●●
●●
●●●
●
●●
●
●●
●
●
●●●
● ●
●●●
●●
●●●
●●
●●●
●● 
●
●●●
●●●
●●
●●
●
●
●
●●
●●●
●
●●
●●● 
●
●
●
●●●
●
●
●
●
●
●●
●●
●●
●
●
●●
●●
●
●●
●
●
●
●●●
●● ●●●
●
●●
●●●
●●●
●●
● ●
●●
●●●
●●●●●
●●
●●●
●●
●●
●●●
● ● ●
●●●
●●
●
●
●
●●
●
●
●●
●●
●●● ●●
●
●●
● ●
●
●●●●●
●
● ●● 
● ● ● ● 
● 
● ● 
●●
●●
●
●●
●●
●●
●
●●● ●
● 
●
●●
●
●●
●
●●
●
●
●
●●
●
●●●
● 
●● 
●●●
●● 
●
●●●
●●
●
●●● ●
●●
●● ●●
● 
● ●● 
●
●●●
● ●
●●
●●●●●
●● 
●● ●● 
●●●
●●●
●
●●
●
●●●
●
●● ●
●●●
●●●
●●●
●
● ●● 
● ● 
● ● 
●
●
●●●
●● ●●
●●●
●●●
● 
●●●
●●
●●
●
●●
●
●
●●●
●●
●
●
●●●
●
●
●
●
●●
●●
●●●
●●●
●
●
●●
●●
●●●●
●●●
●
●
●
●●
●
●●●●
●
●
●
●●
●●●
●●
● ●●
●● ●
●●
●●
●●●
●●
●
●
●●
●●
●
●
●●
●●
●●
●●●
●●
●●
●●
●
●●
●●
●●
●
●
●
●●
●
●●●●●●
●
●●
●●●●
●●
●● ● 
●●
●
●●●
●●● ●
●●●
● ●
●●
●●●
●
●●
●
●
●●●
●●●●
●
●
●●●●
●●
●
●●●●
●
●●●
●●●
● 
● ● 
●●
●●●●
●●●●
●●
●
● ●●●●
●●
●●● 
●
●●
●
●●
●
●●●
●
●●●
●
●●
●
●
●●
●
●
●●
●●●●●
●
●●
●●
●●●
● ●
●●
●●
● ●●
●●●
●●
●
●
●●●
●
●
●
●●
●
●●●●
● 
●
●
●
●
●●
●●
●
● ●
●●
●●
● ●
●●●
●●
●●●●
●●●
●●
● 
● ● ● 
● ●
●● 
●
●
●●
●●●
●
●●
●●
●● ●
● 
●●●●
●●●●
●●
●
●
●●
●
●●
●● ●
●
●
●●●
●●
●● 
●●●●
●●●
●●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●●
● 
● ●●
●
●●
●●
●●●
●●
●●●
● 
● ●
●●
●
● 
●
● ●
●
●● 
●●
●●
●●
●●●
●●
●●
●
●●
●●
●●●●
●
●●
●●
● ●●
● 
●●●●
●
●
●
●●●●
●●●
●●
●
●●
●
●●●
●
●
●●
●
●
●●
●●
●
●●
●●
●●
●
●●●
●●
●●●
●●
●
●
●●
●
●
●
●
●●
●
●●●●
●●
●
●●
●●●
●
●●●
●
●●
●
●
●●
●●
●●
●●
●
●
●
●●●●●
●
●●●
●●
●
●●●
●
●
●●
●
●
●
●●
● 
● 
● ● ●●
●● 
●
●
●
●●
●●●
●●●●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●●●
●
●
●●●●
●
●●●
●●
●
●
●●
●
●●
●●●
●●●
●
●
●●
●●
●
●
●
●●
●
●●
●
●●
●●
●●
●
●
●
●
●
●●
●●●
●●●
●●
●●
●
●●
● ●
●●
●
● 
●
●●
● ● 
●
●●
●
●●
●● ●●● 
●●
● ●
● 
●
●
●● ● 
●● 
● ● ● ●● 
● ● 
●●
●●
●
●●●
●●●
●●●
●●●●
●●
●●
●
●●●
●●● 
●
● 
●●
●●
●
●●
●●
●●
●●●●
●
●
●●●● ●
●●
●●
●●
● ● 
●
●
●●
● ●●●●
●
●●●●
●
●●
●
● 
●●
●●
●
●
●●
●
●●
●●●●
●●●●
●
●
●
●●
●
●
●
●
●
●
●●
●●●
●
●
●●
●
●●●
●●
●●●●
●●
●●●●●
●●
● 
●●
●●
●●
●
●●
●● ●●
● 
●●
●●
●●●●●
●
●●
●
●●●
●●
●
●
●●
●
●
●●
●●●●
●●●
●●●
●●●
●
● 
●●● ●●
●●●●
●●●
●
●●
●●
●
● ●
● ●● 
●●
●
●● ● ●●
●●
●●
●●●●
● 
●
●●●●
●
●●
●●● ●● ● ● 
● ● 
●●●
●●●
● ●● 
●●●
●●
●
●●
●
●●
●●
●
●● ● 
●
●●
●●●
●
●
●
●
●●●●●
● ●●● ● 
●● ●
● 
●
●
●●●
●●●
●
●●
● 
●●●● 
●
●●
●●
●● ●●
●
●
●
●●
●●
●●
●●
●●
●●
●●●
● 
●
●●●
●
●
●
●
●●●●
●
●● 
● 
●
●
●
●
●●
●
●●
●●
●●●
●●●
●●●
●
●●
●●
●●●
●●●
●●
●●
●
●●
●●
●
●●
●●
●
●●
●●
●
●
●●●
●●
●●
●●●
●●
●
●●
●
●●● ●
●●
●● 
●
●●
●●
●
●●●
●
●●
●●
●
●
●●●●
●●
●●
●
●●●●●
●●
●●
●●
●●
●●
●●
●
●
●●
●●●
● 
●● ● 
●
●●●
●●●
●●●
●
●●●●
●
●●
●●●
●● ●
●●●
●●●●
●
●●
●
●●●
● ●
●●
●
●●
●●
●●●●
●●
●●
●●●●
● ●●
● 
●
●●
●●
●
●●
●
●●
●●
●●
●
●● ● 
● ●●
●●
●● 
●
●●
●●
●● ●●
●
●
●●●●
●●
●●●
●
●● 
●
●●
●●●
●●
● ●●
●
● 
●●●
●●
●
●●
●●
●●
●●
●
●
●●
●●
●
●
●●
●●
●
●●
●●
●●
●
●●
●●
●
●●
●●●
●●
● 
●
●●
●●
●● 
●
●●
●●●
●
●
●●
●●
●
●
●●
●●
●
●
●●
●
●
●●
●●
●●
●
●
●
●●
●
●●●
●●
●●
●
●●
●
●
●●
●
●
●●●
●
●●
●●
●●
●● ●
●●●●
●●●
●●●
●●●
●
●●
● 
●●
●
●
●
●●●
●●
● ●●
●●
●●
●●
●●
●
●●
●●●●
●●
●
● 
●
●
●
●●●
●
●●●
●
●●
●
●●●
●●● ●●
●●●
●●●
●
●●●●●
●●
●
●
●●●
●●●
●
●●
●● 
●
●
●●
●
● ●●●
●●●
●●
●●●
●●
●● 
● ●
● 
●●●
●
●
●●●●
●●●
●
●●
●●●
●●
●●
●● 
●
●●
●
●●
●●
●●●
●●●
●●
● 
● ● ●
●●
● ●
● 
●
●●
●●●
● 
●
●●
●●
●●
●●●
●●
●
●
● 
●●
●● 
●●● ● ●
●
●
●●
●●●●
●●
●●●
●●●
●●
●
●●
●
● 
●
●●●●
● ● 
●● ●●
●●
●●●●
●
●
●●●
●●
●●●
●
●● ●●
●
●
●●●●● ●●
●● 
●● 
●●
●● 
●●●
●
●● ●
●●
● 
●
●●
●
●
●●● ●
● ● ●●
● 
●●●●
●●●
●●●●
●
●●
●
●●
●●
●●
● ●
● ●
●
●●
●●●
●●●
●●
●●●
●●
●
●●●
●●
●
● ●
●
●
●●
●●
●●
●●
●● ●
●●
●
●●
● 
●
●
●●
●●
●●
●●●
●●
●●●
●
●●●
●
●●
●
●●●
●
●●
●
●●
●●●
●●●●
●●
●
●●
●●
●●●
●●
●● ● ● ●
●●● 
●● ● 
●
●●●
●
●●
●●●
●
●
● ● 
●
●
●●
●●●
●
●
●●●
●●●
●●
●
●●
●●
●●
●●
●
●
●●●
●
●
●●●
●
●●
●●●
●●
●
●●
●
●●
●●●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●●●
●●●●●●
●
●
●
●
●●
●
●
●●
●●
●●
●●
●●●
●
●●●
●●●
●
●
●
●●●
●●
●●●
●
●●●
●
●●●
●
●
●●
●
●
●●
●●
●
●●
●●●
●
●●●●
●●
●●●
●
●●
●●
●
●●
●
●●
●●
●●●
●●
●●●
●●●
●
●●●
●●
●
●
●
●●● ●
●●
● 
●
●
●●
●●●
●●
●●●
●
●● ● 
● ●
● 
●
●●●
●
●
●
●
●●
●
●●
●
●
●●● ● ● 
●●
● ●●●
● 
●
●●
●●
●●
●
●●●
●
●●
●
●
●●
●●
●●●
●
●●
●
●
●●
●●
●● ●● ●
●●
●●●
●
●●
●●●●
●●
●
●●
●
●●
●●
● 
●
● ●
● ●●
●
●
●●
●●●
●
●●
●
●●
●●
● 
●●
●●
●●
●●●
●●
●
●●
●
●●
●●●
●●
● ● 
● ●●●●●
●●
●●●●
●
●
●●
●●●●
●
●●
●
●●
●
●●●
●●●
●
●●● 
●●●
●
●●●
● ● 
●
●
●●
●
●●
●●
●
●
●●●●
●●●
●
●
●●
●●
●●
●
●●
●●
●
●
●●●●
●
●
●●
●●●
●
●
●
●●
●●
●●
●●
●●
●
●●
●
●
●
●●
●●
●●●
●
●
●●● ●
●●●
●●●●●
●●
●●●●
●●●●
●
●
●● 
● ●● 
●
●
●
●●
●
●●
●
●●
●● ● 
●
●●
●
●●
●●●●
●●
●●
●●●●●●
●
●●
●●●
●●●●
●
●●●
● ● ●●●
●●
● 
● ● ● 
●●●
●●
●●
●●
●●
●
●●
●●
●
●
●●●
●
●
●
●●●
●
●
●●● ● 
●●
●
●
●●
●●
●
●●
●
●
●
●●●
●●
●●●
●●
●●
●●●
●●●
●●
●
●
●
●
●
●●●
●
●●●
●●
●
●●
●
●
●●
●
●●
●●●
●●
●●
●●●
●
●
●●
●●●
●
●●
●●
●●●
●●●
●
●●
●●●
●
●●●
●
●
●●
●
●●
●
●●●
● ● 
●
●
●
● ●●
●
● 
●
●
●●
●
●
●
●●
●●
●
●●
●●
●
●
●
●●
●●
●
● 
●
●●
●●
●●
●
●●●●
●●
●
●
●
● ●
●●
● 
● ● 
●
●●
●
●●●
●
●
●●
●●
●● ●
●
●
● 
●
●
●●●
●
● ● 
●
●●
●●
●●●
●●
●●●
●
●●●
●●●
●●●
●
●●●
●●
●
●●●
●●
●●
●
●●
● 
● 
● ● ●●●
●
● 
●●●
●
●●
●●
●
●●● ● ● 
●
●●●
●
●● ●●
●
●●
●● 
● ●
●●●●
● 
●●●
●●
●●●●
●
●● ● 
●●
●
●●
●●●
●●
●●
●●
●●●●
●
●●● ●
●●
●●
●●●
● ●
●●
●
●
●●●
●
●●
● 
●
●●●
●
●
●●●
●● ● 
● ● 
●
●●
●●
●●●
● ●●
● 
●●
●
●●
●
●●
●
●
●
●
●●
●●
●
●●
●
●●●
● 
●●●●
●●
●●
●●
●●●
●●
●●
●
●
●
●●
●●
● 
●●●●●●
● 
●●●●
●
●●●
●●●
●● 
●●
●
●●●
●●
●
●● 
●●
●●●●
●●
● 
●
●●
●
●●
●
● ●
●●● 
●●●
●●●
●
●●●
●
●●
● ● 
●●●
●●
●
●●
●●
●●
●●
●●
●●●
●
●
●● 
●●
●●
●
●●● ●
● ●●●●
●
●
●●●
●●
●●
●●●
●●
●●
●
●
●
●●
●
●●
●
●●●●
●
●●●
●
●
●
●
●●
●●
●●●
●●
●
●●● ●●●●
●
●
●●
●●
●●
●
● 
●
●
●
●●●
●●
●
●● ● ●
●●
●
●●●
●
●●
●●
●
●
●
●
●
●
●●●●
●●
●
●●
●●
●
●
●
●●●
●●
●
●●
●
●●
●●
●
●
● ●
●●
●
●● ● 
●
●
●●
●●
●
●●●●
●●
●●
●●
●
●●
●●●
●●
●
●●
●
●●●●
●
●●
●●●●
●●
●●●
●●
●
●
● ●
●●●
●● ●
●●
● 
● ● 
● 
●
●●●
● ●
●●
●●
● ●●
●
●●
●
●●
●
●●
●●
●
●●●
●●
●
●●
●
●●
●● 
●●●
●●
●●
● 
●
●●
●●
●
●●●
●
●●●
●●
●
●●●
●●●
●
●
●
●●●●
●
● ●
●
●●
●●● ●●
●●
●●●
●
● 
●
● 
●●●
●●●
● ●●
● 
●●
●●
●
●●●
●●
●
●●
●
●●
●
●●
●●●● ●
● 
●●
●
●
●●●●●
●
●
●●●
●
●
●●●
●●●
●
● ●●
● 
●●
● 
●●●
●● ●●
●
●●
●●
●●
●
●
●
●
●●
●●●
●●
●
●●
●
●
●●
●●
● ● ●●● 
●●
●●
●
●●●
●●
● ● 
●
●●
●
●● 
● ●
● ●
●
●
●●●●
●●
●●
●
●●●
●
●
●●
●●
●
●
●
●●
●
●
●●
●
●● ● 
●
● ● 
●
●●●
●
●
●
●●
●●
● ●
● 
●
●●●
●●
●●
●●
●● ● ● 
●●
●●
●●
● ●●
● 
●
●●
●●●
●●
●
●●●
●
●●
●●
●●
●● ● 
●●●
●
●
●●
●●
●
●● ● 
●
● 
●●● ●●
● ●
●
●●
●●
●●
●
●● 
●●
●● 
●
●●
●
●●●●●
●●
●
●●
●●●●●
●●●●
●●
●●
●
●●●● 
●
●●
●
●●
●
●●
●
●●
●●
●●●●
●●
●
●●
●●
●
●
●●
● ● ●● ● ●
● ●●
● ● 
● ● ● 
●●
● ●
●●
●●
●●
● 
●●
●●●
●
●
●
●●
●
● ●●
●
●●
●
●
●●
● 
●●
●● ●
● ●●
●●●
●●
●
● ●●
●
●
●
●
●
●●●●
●●●●
●
●●
●●●
●
●
●●
●
●
●●
●●●
●●
●●
●
●●●
●
●
●●
●
●●● ●
●●
●●
●●●
●●
●
●●● ● 
●
●
●●●
● ● 
●
●●●
● ●
●●
●●● ● 
●●
●● ●
●●
● 
●
●●●●
●
●●●●
●●●
●
●●●
●●
●●
●●
●●
●●
●●
●
●●●
●
●●
●●
●
●
●
●●
●●
●●●
●
●
●●
●●
●
●●
●
●●
●●
●
●●
●●●●
●●
●●
●●●
●●
●●
●●●●
●●●●
●●●●
●●
●●
●●
●
●
● ● ● 
●
● 
● ●●
●
●
●●●●
●●● ●
● ●
●
●●
● 
●
●●●
●
●
●
●
●●●●
●●
●●●
●●
●●●
●● ●
●
● 
●
●
●●
●
●●●
●
●
●
● ●●
●● ●
●●
●●●
●
●●
●
● ●
● ●
● 
●● ● ● ●
●●
●
● 
●
●
●
●●
●
●●●
●● 
●●
● 
●●●
●●
●●●●● ● ●● 
● ● 
●●●
● ● ● 
● ● 
●●
●●
●
●
● ●●●
●● 
● ●
●●
●
●●●
●
●●
● 
●●
● 
●● ● ●●●
●●●
●●
●
● 
●●● ● 
● ●●
●
● 
●●●●●
●
●
●●
●●
●
●●●
● ● ●
●
●
●●
●●
●
●●● ●
●● 
●●● 
●
● 
●●
●●
● ●
●●
● 
●
●
●●
●
●
●
●●
●
●
●
●
●●● ● 
●
●
●●
●
●●
●●
●
●●
●●
●●
●●
●●
●●●
●●
●●●
●●
●
● ● 
● ● 
● ● ●
● 
● ● 
● ● ●
●●
●●●
●●● 
●
●
●●●●●
●
●
●● ●●
●● ● 
●●● ● 
●●●
●●
●●
●
●
●●
●
● ●●
● ● 
●●●
●● ●
●● 
●● ●●
●●
●
●●●● 
●
●
●●●●
●● ● 
●
●
● ●
●
● 
●
●●●
●
●●
●●
●●
●●
●
●
●●●
●●●
●
●
●●
●●●
●
●●
●●●●●
●●●●
●●
●
●●●
●
●
●
●●
● 
●●
●●
●●
●
●
● ● 
●●●
●
●●
●●
●●
●●
●●
●●
●● ●
● ●●
●
●●●
●● ● ●
●●
●● ●●
●●
●● ●●
●●
●●●● ●●
●
●● 
●
●●●
●
●●
●●●
●●
●●
●
●●● ●●● ●
●●
●●
●
●●●●
●● ●
●
●
●● 
●
●●
●●
●
●●●
●●
●●
●●●●
●
●●●
●●●
●●
●
●●● 
●● 
●
●●
●●
●
●
● ● ●
●●
● ●●
● ●●
●●
● ●
●●
●●
●●
●●●
●
●●
●
●●●
●●
●●
●●●
● ●●●
●●
●●
●●
●● 
●●●
●
●●●●
●●
●●
●
●
● ● ●
●
● ●●
●●
●● 
●
●●
●
●● ●●●●
● 
●●
●
●
●●
●
●●
●●
●●
●
● ● 
●
●
●●
● ● 
●
● ● 
●● 
●
●
●●● 
● ● 
●●
●●
●
●●●●
●
●
● ●●●●
●
●
●●●
●●
●●
●●
●●●
●
●●●
●●●
●●
●●
● 
●
●●
●
●
●
●
●●
●●
●●
●
●
● ●
● 
●
●
●●
●
●●
●
●●
●
●●
●●
●●
●●
●
●●●
●● ● ●
●●
●
●●
●●
●
●●●
● 
●
●●●
●
●●
●
●
●
●●
●
●●
●
●
●●
●●●
●
●●
●●
● ● ● ● 
●
●●
●
●●
● 
● 
●
●●
●
●●
●●●
●●
●●
●
●●●
●●
●●●
●
●●
●●
●
●●
●●
●●●
●
●
●
●●
●●●
●
●
●●
●
●
●●
●
●
●●
●●
● 
●●
●●
●●●
●●●
●●
●● 
●●
●
●●●
●●
●●●
●●
●●●
●●
●
●
●●
●●● ●
●●
● 
● ●
● 
●● ●●
●●
● ●
● 
●
●
●●● 
●●
●
● 
● 
● ● ●●
● ●
●
●
● ●●●
●
●●
●●
●
●●
●●●
●
●●
●●
●
●●
● 
●
●●● ● ●●● 
● 
●●●
●● 
●
● ●●
●●
●
●●●●●
●●●●●
●
●●●
●●●
●●●
● 
●
●●
● 
● 
●
●●
●●
●●
●
●●
●●● ●
●
●●
● ●●
●●
●●●
●●
●
●● 
●
●●●
●●
●● ●
●
●●
●● ●● 
●
●● 
●
●●
●●
●●●
●●●●
●●●●●
●●●
●
●
●● 
●● ● 
●●
●●
●
●
●
●●
●●
●
●●
●●
●●
●
●● ●
●● ●
●●●●
●●●
●
●●
●●
●
●● ● 
●
●●
●●
● 
● 
● ● 
●
● ● 
●●
●●●
● ●● 
●
●●
●●●
●●
●
●
●
●
●●
●●
●●●
●●●
● ●
●●
●● ●●
●● ● 
●
●●
●●
●●●●
●
●
●●●
●
●●
●●
●●
●
●
●
●
●●●
●●●
●●
●●●
●●
●
●
●●●●
●●
●
●●●
●●
●● ●
●●
●
● 
●
● ●
●● 
● ●
●● ●● 
●
●
●
●● ●●
●● 
●●●
●
●●
●●
●
●
● ●
●●●●
●●
● ● 
●
●
●●
●●●
●
●●
●●
●●●
●
●●●
●●
●
●●
●
●●
●
●●
●●
●●
●●
●
●●●
●●
●●
●●
●
●
●●●
●
●●
●●● ●
●
● 
●
●●
●●
●●
●●
●
●●●
●
●●
●●●
●●
●● ● 
●●
●
●●
●●
●●
●●●●
●
●●
●●●●
●
●●
●●
●●
●
●●
●●●
●●●
●●●
●
●
●●
●●●
●●
●
●●
●●
●●
●
●●●
● ●●
●●
●
●●
●●
●●●
●●● 
●●
●
●●
●●
●●
●
●
●●
●●●
●
●
●●
●
●●●
●
●●●
●
●
●●
●●
●●
●●●●
●
●
●●●
●
●
●●●●
● 
●
●
●
●
●●
●●●
●●●
●
●●
●
●●
●
●
●●
●
●
●●
●●
●●
●●
●
●
●
●
● ●
●●
●
●●
●
● 
●●
●
●
●●
●
●
●
●●●
●
●
●●●
●●
● ● 
●●
●
●●●
● ●
●
●●●
●
●
●●
●
●●●
●●
●
●●●
●
● 
●●
●●
●●
●
●●
●
●●●● ●
●●
● 
●
●●
●
●
●●
●●●
●●
●●●
●●●
●
●●
●
●
●
●●
●
●●
●●
●
●●
●
● ● 
●●
●
●●
●
●● ●
●●
● 
●
●●●●
●●
● ●
●
●
●●
●
●●
●
●
●●
●● 
●●
●●●●●
●
●●●
● 
●●●
●●●
● 
●●●
●●
●
●
●
●●●●
●
●●
●●● ●
●●
●●● 
● ●●
●●
●
●●● ●●●
●
●●
●●
●●
●● 
●●●●
●
●●
●●●● 
●
●
●
●
●●
●●
●
●●
●
●●
●●
●●
●●
●
●
●
●
●●
●●
●●
●●
●●
●
●
●
●●
●●
●
●
●●●
●●
●
●
●●●
●
●
●
●●● 
●● ●● ●
●
●●●●
●●
●
●●
●
●●●●
●
●●
●
●●
●● ●
●
●●
●
●●
●
●●
●
●●
●
●● 
●
●●
●●
●●
●
●
●●●
●
●
●●●●
●
●
●
●
●●●
●●
●●
●●
●
●●●
●●
●
●●
●
●
●●
●
●
●
● ●
● 
●
●●●
●
●●
●● ●
●
●●●
●
●●
●●●
●
●●
●●
● 
●●
●●
●●
●●●
●●●●
●●●●●●
●●●
●● 
●●
●●
●●●●
● ● 
●
●●
●
●
●●
●
● ●● 
●
●●
●
●● 
●
●
●
●
●● ● 
● 
●●
●●●
●●
●●●
●●
●
●
●●
●
●●
●●●●
●
●
●●
●●
●
●●●●
●●
●
●
●●
●●
●
● ●●
● 
●
●
●
●● ●
●
●●●●●
● 
●●
●
●●
●
●●●
●●●
●
●
●
●●
●●
●●
●
●●
●●
●●●
●
●
●●
●
●●
●
●
●●●●
●●
●●
●
●
●●
●
●●●
●
●●●
●
●
●●●●
●●●
●
●●
●
●
●●●●
●●
●
●
●●
●
●●
●●
●●
●
●●
●
●●
●
●●●●
●●
●●
●
●
●●●
●●●●
●●●
●●
●
●●●
●
●●
●●
●
●
●
●
●
●●●
●
●●
●
●
●● ●●
●●
●●
●
●●
●
●●●●●
●●●
●
●●●
●●●
●
●
●●●
●●
●
●●●
●
●
●
●
●
●●
●
●●●
●●
●
●
●
●●
●
●●
●
●●
●
●●
●●
●●
●
●●
●●
●●●
●●
●●
●● 
●●
●●
●●
●●●
●
●●●
●
●●●
●●●●
●●
●●●
●●
●
●●
●
●●●
●
●
●
●●
●●●●
●
●●
●
●●●
●●
●
●●
●
●●
●●
●
●
●●●
●
●
●●
●
●
●●
●●
●
●
●●
●
●●
●
●●
●●
●●●
●
●●
●●
●●
●●
●● ● 
●
●●●
●●
●●
●
●
●●
●●
●●
●●●●
●●
●●
● ●●
● 
●
●●●
●
●
●
●●
●●●
●●
●●●
●●
●
●●
●●
●●
●●● ●●
●
●
●●●
●●
●● ●
● 
●
●
●
●●
●
●●
●●
●●
●
●
●●●
●
●●
●
●●
●●
●●
●
● 
●● ●●
●●●
● 
●
●●
●●●●●●
● 
● 
●●●
●●
●●
●
●
● 
● ●
●●
●
●●
●
●
● 
● ●● 
●
●●
●
●●
●
●●
●●●
●●●
●
● ● 
●●
●
● ● ● 
● ● 
● ● ● ● ● ● ● 
● ●● ● 
● 
● ● ● 
● ● ● ● ● 
● ● 
● 
● ● ● 
● 
● ● ● ● 
● 
● 
● 
● 
●●● 
● 
● 
● 
● 
● ● 
● ●● 
● 
● 
● ● ● 
● ● ● 
●● ● ● ●● ● ● ●●●
● 
● ●●●
● 
● ● 
● ●●
●
●● 
●
●
● ● ● 
●
●●● ●●●
●
● ●
● ● 
● ● 
● ● ● ● ● ● 
● ●● ●●
●●
●●
● 
● 
●
● 
●
● ●
● ●
● ●● 
● 
●
●●● 
● 
● ●
● ● ● ● ●●●
●
● ● ● ●● ● ● ● ● ● ● ●
●●
● ● ●●
●
● ●
● ●●
● ● ●● ● ● 
●● ● 
● ●
●● ●● 
●
●● 
● 
● ● 
● ●● 
● ● ● ● ● ● ● ●
●●
● ●
● 
● 
● ● ●
●● ● 
● ● ●●
●● 
● 
● ● ●● ● 
●
● ● 
●● 
● ●
● ● ● ● 
● ●
●
● ●
● 
● 
● ● ● ● ● ● 
● ● 
● ●
●● ●● ● ● 
●
●● ● 
● 
● ● ● 
● ● 
● 
● ● ● 
● ●● 
● 
● 
● ● ●● 
● ● ●
●
●● 
● 
● ●● 
● 
●● ● 
● ● ● 
●● ● ● 
● ● 
● ●
● 
● ● ● 
● 
● ● ● ●
● ●● 
● 
● ●●
●● 
● 
●● 
● ●
● ● 
● 
●
● 
●
●
● ● 
● ● 
● ● 
●●● 
● ●
●●● ● 
● ● 
● ● 
● ● ●
● ●● 
● ●● 
●
● ● ●● 
●
●● ● 
●● 
● 
● 
●●
● 
● 
● 
● 
●
● 
●
● ● 
● ● ● 
● 
● ● 
● ● 
● ● ● ● ● ● 
●●
● ● ●● 
● ● 
● 
●● ● ● 
●
● ● 
● ● 
● 
● 
● 
● ● ● 
● ● ● ● 
● ● 
●
●●
● ● ● 
● 
●
● ● ● ●
●● ● ● ● ●
● ● 
● 
● ● 
● ● 
● 
● 
●
● ●● 
●
● ●● 
● 
●● ● 
● ● ●
● ●
●● ● ● 
●● 
● 
● 
● ● ●● ● 
● 
●
● 
●
●
● ● 
●
●● ●●● 
●●
●
●●
● ● ● 
●
●
●●
●● ● 
●
● 
● ● ● 
● 
● ● ● 
● 
● ● ● 
● 
● ●●● 
● ● 
● ● 
●
●
●●
● ●
● 
●●●
● ● ● 
● ● 
● ● 
●●
● ● ● 
●
●● ● ● 
● 
●● ●● 
● ● ● 
● ●●
● 
●
●●
●● 
●
● 
● ● 
●● ●
● 
●
●●
●● ● 
● ●
● 
● ● 
● ● 
● ● 
●●
● ● 
●
●●
● ● 
●
● ● 
● ● ● ●
●● 
● ● 
●● 
● ●● 
●
● ●
● 
●
●● ● ● 
● ●
● 
●
● ● 
● ● 
● ● 
●●
● ● ● ● 
●
●●
●
● ● 
● ● 
● 
● ● 
●●● ● 
●●
●●
● ● 
●●●
● ● 
● ● 
● ● 
●
● ●
●● ● ●
●● ● ● 
● ● 
●
●
●
●
●
●● ● ●● 
●
●
●
●
●
● ● ● 
● ● 
●
●
● ● 
●
● ● 
●
● ● ●● 
● ●
● 
● ● 
●● ● 
● ● 
● 
● ● 
● ● ● 
● 
● 
● ● ● 
●
● ●
●● 
● ● 
●
● ●
● ● 
●
●
●
● ●
● 
● ● 
● ● 
●
●● ● 
●● 
● ●● 
●● ● ● 
● ●● ● 
●●●
● ●
●
●●
●● 
● ●
●● ●
● 
● 
● 
● ● 
● ● 
● ●
● 
● ● 
● ● 
● ● 
● ● 
●
●
●
●
● ● 
●
● 
● ● 
●
●
● 
●●
●
● 
● ●
●
● ●● 
● ● 
● ● 
● ● 
● ● 
●
●
●●
●● ●●
● ● ● 
●
●●
● ● ● 
● ● 
●
●●
● 
● 
● ● 
●
●● ● 
●
● 
●
●
●
●
●
● ● 
●● 
●
● ● ●● ●● ● ●●● ●● ●
●●● 
● 
● ●●
● ● 
● 
● ● 
● 
●
●●
●● ●● 
●●
●
t-SNE 3 
a 
−6 −4 −2
 0
 2
 4 
−
 4 
●
●
●
● 
● 
●
●
● 
● 
● 
● 
● 
● 
● 
● 
●
●
● 
● 
● 
●
●
● 
●●
●
●
●
● 
● 
●
● 
● 
● 
●
● 
● 
●
●
● 
●
●
●
●
● 
●
● 
●
●
●
● 
● 
● 
●
●
● 
●
● 
● 
●
● 
●
● 
●
●
●
●
● 
● 
● 
● 
● 
●
● 
● 
●
● 
● 
●
● ●
● 
● 
● 
●
● 
● 
● 
●● 
●
●
● 
●
● ● 
●
●
● 
●
●
●
● 
●
● 
●
●
●
● 
● 
● 
● 
●
●
●
● 
●
● 
●
● 
●●
● 
● 
● 
●
● 
● 
●
●
●
● 
● 
●
●
●
● 
● 
●
● 
● ● 
●
●
● 
● 
●●
●
●●
●●
● 
● 
● 
●
● 
●
●
●
● 
●
● 
● 
● 
● 
●
● 
● 
●
●
●
● 
● 
●
● 
●
●
● 
●● 
●
●
● 
●
● 
● 
●
● 
●
● 
●
● 
●
● 
● 
● 
● 
●
●
●
● 
●
● 
●
● 
● 
● 
●
●
● 
● 
● 
● 
● 
●
● 
● 
●
● 
●
● 
● 
●
● 
● 
●
●
●
●
● 
● 
●
● 
● 
●
● 
●
● 
● 
● 
● 
●
● 
● ● 
●
●
● 
● 
● 
●●
●● 
●
● 
● 
● 
● ● 
● 
● 
● 
● ● 
●
●
● 
●
●
● 
● 
● 
● 
● 
●
● 
● 
● 
●
●●
● 
● 
● 
● 
●
●
● 
●
●
●●
●
● 
● 
●● 
● 
●
● 
●
●
●
● 
● 
● 
● 
●
● 
●
● 
● 
● 
● 
● 
● 
● 
● 
● 
● 
●
●
● 
● 
●
●
●●
● 
● 
● 
●
●
● 
● 
● 
●● 
● 
● 
● 
● 
● 
●
●
● 
●●●
●
●
● 
●
● 
● 
● 
● 
●
●
● 
● 
● 
● 
● 
● 
● 
●
● 
●
●●●
●
● 
● 
● 
●●
● 
● 
● 
● 
● 
t-SNE 1 
t-SNE 2 
t-SNE 2 
c 
DMRs identified on the training set (plasma cfDNA)
DNA methylation data from training and validation sets (plasma cfDNA) (n=388) 
● 
Healthy
donors 
●
● 
●
●
● 
●●
●
● 
● 
● 
●
●
● 
●
●
● 
●
●
● 
● 
● 
●
●
● 
●
●
● 
●
● 
● 
● 
●
●
●
● 
●
● 
● 
●
●
● 
●
●● 
●
● 
●
●
●●
●
● 
●
●
●
●
●
● 
●
●
● 
●● 
● 
● 
●● 
●
●
●
●
●●
●
●●
●
●
●
● 
●
●●
● 
● 
● 
●
●
● 
●
●
● 
●
● 
● 
● 
● 
●
● 
●
●
●
● 
●
● 
● 
● 
● 
● 
PDAC 
LUC 
● 
●
● 
●
●
● 
● 
●
● 
● 
● 
●
●
●
● 
●
● 
●
●
● 
●
● 
●
● 
● 
● 
●
● 
●
● 
●●
● 
●
●● 
● 
● 
●
●● 
● 
● 
●
● 
●
●●
●
●
●
● 
●
●
● 
●
●●
●●
● 
● 
●
●
●
● 
●
●● 
● 
●
● 
● 
●●
●
●
● 
●
● 
●
● 
●
●
● 
●
● 
● 
●
● 
● 
●
●
● 
●
● 
● 
●
● 
● 
●
●●
●
●
●
● 
● 
●
●
●
● 
● 
●●
●
● 
● 
●
●
● 
●
●
● 
●●
●
● 
●
●
● 
● 
●
● 
● 
● 
● 
● 
● 
●
● 
●● 
● 
●
●
● 
● 
●
●
●
● 
●
● 
●
●
● 
●
● 
● 
●
●
● 
●
●
●
● 
●
● 
●●
●
●
●
●
●
●
●
● 
●
●●
●
● 
● 
● 
● 
●
●●
● 
●
● 
●
● 
●● 
● 
●● 
●
●● 
● 
BRCA
● 
● 
●
● 
●
●
●
●
● 
●
●●
● 
●
● 
● 
● 
● 
●
● 
●
● 
●
● 
● 
●
● 
● 
●●
●
● 
●
●
●
●
● 
●●
● 
●
●
●
●● 
●
●● 
● 
●
● 
● 
● 
t-SNE 3
● 
● 
t-SNE 3
● 
●
●
●
●
● 
●
● 
● 
●
●
● 
● 
●
● 
●
●
●
●
● 
●
● 
● 
● 
●
●●
● 
CRC
●
● 
●
● 
●
● 
●
●
● 
●●
● ●
●● 
● 
●
●
●
●
●
● 
●
● 
● 
●
● 
● 
●
● 
●
●
● 
●
● 
● 
●
● 
●
● 
● 
● 
●
● 
● 
●
●
● 
●
●
● 
●●
● 
●
● 
●●
●●●
●
● 
● ●●
● 
AML
● 
●
● 
●●
● 
● 
●
● 
● 
●
● 
●
●
●
● 
●● 
● 
●
● 
●
● 
●
●● 
● 
●
● 
● 
● 
●
●● 
●
● 
●
● 
● 
● 
●●●
● 
● 
●
● 
● 
● 
●
●
● 
●
● 
● 
RCC 
BLCA 
● 
●
● 
●
●
● ● 
●
●● 
●
●
● 
● 
●● 
●
●
●
●● 
●
●
● 
●
● 
●●
●
●
●
●
●● 
●
● 
● 
● 
●
●
● 
● ●
● ●
● ● 
●
● 
● 
t-SNE 1 
t-SNE 1 
t-SNE 2 
b 
DMRs identified on the training set (plasma cfDNA)
DNA methylation data from TCGA (tissue) (n=4,032) 
●
● 
● 
●● 
●●●●
● 
●● 
Blood 
● ●●●●●●●●●●●●●● ● ●●●●●●●●●● 
● ●●●●●●●●●●●● ●●● 
●●● ●●● ●● ●● ●●● ●●● ●● ● ● 
PDAC 
● 
● 
●
●● 
● 
LUC 
15−10 −5  0  5  10 15 − 
− 
●
●●
●
●
●
●
●
●
●●●●
●●
●●
●
●●●
●●
●
●●
●●
●
●
●●●●
●
●●
●●
●
●●●
●
●
●
●
●
●●
●
●●●●
●●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
● 
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
● 
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●●
● 
●
●
●
●
●
●
●●●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
● 
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
● 
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
● 
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
● 
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
● 
●
●
●
●
●
● 
●
●
●
●
● 
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
● 
●
●
●
●
●
●
●
●
●
●
●●
●
● 
●
●
●
●
● 
●
●
●
●●
●
●
●
●
●
●
●●
●
●
● 
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
● 
●
●
● 
●
●
●
●●
●
●
●
●
●
● 
●
●
●
● 
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
● 
●
●
●
●
●
●
●
●●
●●
●
●
●●●
●
●
●
●●
●●
●
●
●
● 
●
●
●
●
●
●
●
●●
●●
●
●
● 
●
●
●
●●
●
●
●
●●
●
●
●
●●
● 
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
● 
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●●
●
●●
● 
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●●
● 
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●●
●
● 
●●
●
●
●
●
●
●
●
●
●
●● 
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●● 
●
●
●●
●
●●
●
● 
●
●
●
●
●
●
●
● 
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
● 
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
● 
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●●
●●
●
● 
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●●
●
●●
● 
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●●
●●
●●
●
● 
●
●
●● 
●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
● 
●
●
●
●●●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
● 
●
●
●
●
●
●
●
●
● 
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
● 
●
●
●
●
●
●
● 
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
● 
●
●
●
●● 
●
●
● 
●
●
●
●
● 
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
● 
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●●
●
●
●
●
●
●
●
● 
●
●
●
●●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●● 
●
●
●
●
●
●
●
● 
●●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
● 
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
● 
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
● 
●
●
●●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
● 
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
● 
●
●
●● 
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
● 
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●●●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
● 
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
● 
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
● 
●●
●
●
●
●
●
●
●
●
●●
●
●
●●●
●●●●
●
●
●
●
●
●
● 
●
●
●
●
●
● 
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
● 
●
●
●
●
●
●
●●
●
●
●
● 
●
●
● 
●
● 
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●● 
●
●
●
●
●
● 
●
●●
●
●
● 
●
●
●
●
●
●●
●
●
●
●
●●
● 
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
● 
●
●
● ● 
●
●
●
● 
●
●
● 
●
●
●●● 
●
●
●●
●
t-SNE 2 
t-SNE 1 
t-SNE 3 
t-SNE 2 
● 
● 
●
● 
●
● 
●
●
●
● 
●
● 
●
● 
●
●
● 
● 
●
● 
●
●
●
●
● 
● 
●
● 
●
●
● 
●
●
● 
● 
●
t-SNE 3
● 
● 
t-SNE 3
● 
● 
● 
●
●
● 
●
● 
● 
●
● 
●
●
● 
● 
●
●
● 
●●
● ●●
● 
●●
●●
●
● 
●
●
● ● 
●
● 
●●
●
● 
● 
● 
●
● 
● 
●
● 
●
● 
● 
●● ●●● 
●●● ● ● ●● ●●●●●●●●●●●●● ● 
BRCA
● 
●●
● 
●
● 
● 
● 
●
● 
●
●● 
● 
●
● 
●
● 
● 
● 
● 
● 
● 
●
●●
● 
●
●
● 
●
● 
● 
● 
● 
●●● 
● 
● 
●
●●
●
● 
● ● 
● 
●
●
● 
● 
● 
●● 
●
● 
●● 
● 
●
● 
●
● 
● 
● 
●
●
● 
●● 
● 
● 
●
● 
●
●
● 
● 
●
●● 
● 
●
● 
●●
●
●
● 
● 
●●
●●
●
● 
●●
●●● 
● 
● 
● 
●
● 
●
●
● 
●
● 
● 
● 
●
● 
● 
● 
● 
● 
●
●
● 
●
● 
●●
● 
● 
●●
●
●●
● 
●
● 
● 
●
● 
●
● 
●
● 
●
●
●
● 
●
●● 
●
●
●●
● 
● 
●
● 
● 
● 
●
●
● 
●
● 
CRC 
AML 
● 
●
● 
●
●
●●
● 
● 
●●
● 
●●
●●
● 
● 
● 
●
●
● 
●
● 
● 
●
●
● 
●●●
●
●
● 
●
●
●●
● 
● 
●
● 
● 
● 
● 
● 
● 
● 
● 
●●
●● 
●
● 
●
● 
●
● 
● 
● 
● 
● 
● 
● 
● 
● 
● 
● 
● 
● 
●
●
● 
●
●●
●
●
●
● 
●
●
●
●
●●
●
●
●
● 
●
●
●●
● 
● 
●
● 
● 
● 
●
● 
● 
●
●
●
● 
●
●
●
● 
●
● 
●
●
● 
● ● 
● 
●
●
● 
●
● 
●
● 
●
●
● 
●
● 
● 
●
●
●●
● 
● 
●
●
● 
● 
●● 
●
●●
●
● 
●
● 
●●
● 
●
●● 
● 
● 
●
●
●
● 
● 
● 
●
●
●
●
●●
● 
●
● 
● 
● 
● 
●●
● 
● 
● 
● 
●
●
● 
●●
●
●
●
● 
●
●
● 
●
●
● 
● 
●●
● 
● 
● 
●● 
●
●●
●
● 
● 
●● 
●
●
●●
●
● 
●
● 
●
●
●
● 
● 
●●
● 
● 
●
●
●
● 
● 
● 
● 
●
● 
● 
●●●
● 
●
● 
● 
● 
●
●●
● 
●
● 
●
●● 
● 
●● 
●●
● 
● 
● 
●●
● 
● 
●
● 
● 
●
● 
●
● 
●●
● 
●
●
● 
●
●
● 
● 
●
●
●
●
● 
●
● 
●
●
●
●
●
●
●
●
● 
● 
●
● 
● ● 
●
● 
●●
● 
●
● 
●
●
● 
●
●●
●
●
●
● 
● 
RCC 
BLCA 
●
● 
●
●
● 
● 
●●
●●
●
●● 
●
● 
●
●
●● 
●●●
●
●
●●
● 
● 
●
● 
●●
●
●
●
● 
●
●● 
● 
●
●
●
● 
● 
●
● 
●
●●
● 
● 
●
● 
●
● 
●
●
● 
● 
● 
●●
● 
● 
● 
● 
● 
● 
●
● 
● 
●
● 
● 
●
●●
●
●
● 
●●
●● 
●
● 
●
●● 
●
●●
● 
● 
●
● 
●
● 
● 
●
●
●
● 
● 
●●● 
● 
●
●●
●
● 
● 
●
● 
● 
● 
● 
● 
● 
●
●
●
●
●
● 
●
●
●
●
● 
●
●
● 
●
●
●
● 
●
●
● 
●
● 
●
● 
●●
●
● ●
●
● 
●
●
●
●
● 
●
●
● ●
● 
●●●
●
● 
t-SNE 1 
t-SNE 1 
t-SNE 2 
DMRs identified on the training set (plasma cfDNA)
DNA methylation data from COSMIC (cancer cell lines) (n=400) 
60−40−20  0  20 40 60 
●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
●●●● 
● 
●
●
●
●
● 
●
●●
●
●
●
●●
●
● 
●
●
●
●●
●●
●
●
●
●
●
●
● 
●
●
●●
● ●
● ●
●
●
●
●
●
●
●
● 
●
● 
● 
●
●●
●
●
●
●
●
●
●●
●
●
●
●
● ●
●
● 
●
●
●
●
●
●
●
● 
● 
●
●
●
●
●
●●●●
●
●
●
●
● 
●
●
●●●
●
● 
● 
●
●
●
●
●
●
●
●
●
●
●
● 
●
●
●
● 
●
●
●
●
●
●
●
●
● 
●
●
●
●
● 
●
●
●
●
● 
● 
●
● 
● 
●
●
●
●
●
●
●
●
●
● 
● 
●
●
●
●
●
●
● 
●
● 
● 
●
● 
● 
●
● 
●
●
●
● ● 
●●
●
●
●
●
● 
● 
●
●
● 
●
●
● 
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
● 
●
●
●
●
●
●
● 
● 
●
●
● 
●
●
● 
● 
●
●
●
●
●
●● 
●
● 
●
●
● 
●
●
●
● ● 
● 
● 
●
●
●
●
●
●
●
● 
●
●
●●
● ●
●
●
●●
●
●
●
●
●
●●●
●● 
●
●●
●
●●
●
●
●●
●
●
●
● 
●
●● 
●
●●
●
●
●●●
● 
●
●
●
● 
●
● 
●●
● 
●
●● ●
●●
●●●●
●●
● 
●
●
●
●
●●
●● 
●
●
●
● 
●
●
●
●
●●
●
●
●
●
●
●●● 
●
●
●
●
●●●
●
●
●
●
● 
●
●
●●
●
●●
●
t-SNE 2 
t-SNE 1 
t-SNE 3 
t-SNE 2 
● 
● ●
●
●
● 
● 
● 
● 
● 
t-SNE 3 
Blood 
PDAC 
LUC 
● 
●
●
●
● 
● 
●
●
● 
● 
●
● 
●
● 
● 
●
● 
●● 
●●●
●●● 
●●
●
●●
●
●
●
● 
● 
●
●
● 
●●
● 
● 
● 
●
●● 
●
●
●
●
● ● 
●
●
● 
●
●
● 
●
●
●●
●
● 
● 
t-SNE 3
●
● 
● 
●
●●
●
●
● 
● 
●
●● 
● 
● 
●●
● 
● 
● 
● 
● 
●
●
●●
● 
●●
●
● 
●
● 
● ●
●
●
● 
● ● 
●●
●
●
●
●
●
●
● 
● 
BRCA 
CRC 
AML 
● 
● 
● 
● 
●
●
●●
● 
●● 
●
● 
●
●
●
● 
●
●
● 
●
●
● 
●
●● 
●
● 
● 
●
●●
●
● 
● 
● 
●●
● 
●
●● 
● 
●
●● 
●
● ● 
●
● 
● 
●
● 
● 
●
●
●
● 
●
● 
● 
●
● 
● 
●
●
● 
●
●
●
●
●
● ●
●
● 
●
● 
●● 
● 
●● 
●
● 
● 
●
●
● 
●●
● 
● 
● 
RCC 
BLCA 
●● 
●
●
● 
● 
● ●
● 
● 
●
● 
● 
● 
●
● 
t-SNE 1 
t-SNE 1 
t-SNE 2