Correlation between aggregated molecular cancer subtypes and selected clinical features

Pancreatic Adenocarcinoma (Primary solid tumor)

23 September 2013 | analyses__2013_09_23

Maintainer Information

Citation Information

Maintained by TCGA GDAC Team (Broad Institute/MD Anderson Cancer Center/Harvard Medical School)

Cite as Broad Institute TCGA Genome Data Analysis Center (2013): Correlation between aggregated molecular cancer subtypes and selected clinical features. Broad Institute of MIT and Harvard. doi:10.7908/C1H130C4

Overview

Introduction

This pipeline computes the correlation between cancer subtypes identified by different molecular patterns and selected clinical features.

Summary

Testing the association between subtypes identified by 8 different clustering approaches and 12 clinical features across 53 patients, one significant finding detected with P value < 0.05 and Q value < 0.25.

4 subtypes identified in current cancer cohort by 'Copy Number Ratio CNMF subtypes'. These subtypes correlate to 'PATHOLOGY.M.STAGE'.
3 subtypes identified in current cancer cohort by 'METHLYATION CNMF'. These subtypes do not correlate to any clinical features.
CNMF clustering analysis on sequencing-based mRNA expression data identified 4 subtypes that do not correlate to any clinical features.
Consensus hierarchical clustering analysis on sequencing-based mRNA expression data identified 3 subtypes that do not correlate to any clinical features.
3 subtypes identified in current cancer cohort by 'MIRSEQ CNMF'. These subtypes do not correlate to any clinical features.
3 subtypes identified in current cancer cohort by 'MIRSEQ CHIERARCHICAL'. These subtypes do not correlate to any clinical features.
3 subtypes identified in current cancer cohort by 'MIRseq Mature CNMF subtypes'. These subtypes do not correlate to any clinical features.
3 subtypes identified in current cancer cohort by 'MIRseq Mature cHierClus subtypes'. These subtypes do not correlate to any clinical features.

Results

Overview of the results

Table 1. Get Full Table Overview of the association between subtypes identified by 8 different clustering approaches and 12 clinical features. Shown in the table are P values (Q values). Thresholded by P value < 0.05 and Q value < 0.25, one significant finding detected.


Clinical Features	Statistical Tests	Copy Number Ratio CNMF subtypes	METHLYATION CNMF	RNAseq CNMF subtypes	RNAseq cHierClus subtypes	MIRSEQ CNMF	MIRSEQ CHIERARCHICAL	MIRseq Mature CNMF subtypes	MIRseq Mature cHierClus subtypes
Time to Death	logrank test	0.172 (1.00)	0.447 (1.00)	0.0897 (1.00)	0.69 (1.00)	0.778 (1.00)	0.49 (1.00)	0.399 (1.00)	0.49 (1.00)
AGE	ANOVA	0.995 (1.00)	0.645 (1.00)	0.759 (1.00)	0.793 (1.00)	0.799 (1.00)	0.565 (1.00)	0.295 (1.00)	0.587 (1.00)
NEOPLASM DISEASESTAGE	Chi-square test	0.047 (1.00)	0.536 (1.00)	0.841 (1.00)	0.461 (1.00)	0.588 (1.00)	0.587 (1.00)	0.73 (1.00)	0.496 (1.00)
PATHOLOGY T STAGE	Fisher's exact test	0.468 (1.00)	0.287 (1.00)	0.915 (1.00)	0.376 (1.00)	1 (1.00)	0.6 (1.00)	0.833 (1.00)	0.712 (1.00)
PATHOLOGY N STAGE	Fisher's exact test	0.174 (1.00)	0.917 (1.00)	0.576 (1.00)	0.785 (1.00)	0.53 (1.00)	0.0637 (1.00)	0.826 (1.00)	0.0332 (1.00)
PATHOLOGY M STAGE	Chi-square test	0.00123 (0.116)	0.142 (1.00)	0.0611 (1.00)	0.0117 (1.00)	0.413 (1.00)	0.366 (1.00)	0.308 (1.00)	0.246 (1.00)
GENDER	Fisher's exact test	0.169 (1.00)	1 (1.00)	0.746 (1.00)	1 (1.00)	0.825 (1.00)	0.724 (1.00)	0.934 (1.00)	0.939 (1.00)
HISTOLOGICAL TYPE	Chi-square test	0.923 (1.00)	0.602 (1.00)	0.273 (1.00)	0.0155 (1.00)	0.638 (1.00)	0.311 (1.00)	0.623 (1.00)	0.331 (1.00)
NUMBERPACKYEARSSMOKED	ANOVA		0.859 (1.00)	0.706 (1.00)	0.85 (1.00)	0.427 (1.00)	0.979 (1.00)	0.945 (1.00)	0.979 (1.00)
YEAROFTOBACCOSMOKINGONSET	ANOVA		0.279 (1.00)	0.497 (1.00)	0.505 (1.00)	0.908 (1.00)	0.859 (1.00)	0.561 (1.00)	0.859 (1.00)
COMPLETENESS OF RESECTION	Chi-square test	0.113 (1.00)	0.0977 (1.00)	0.217 (1.00)	0.0196 (1.00)	0.23 (1.00)	0.298 (1.00)	0.0889 (1.00)	0.183 (1.00)
NUMBER OF LYMPH NODES	ANOVA	0.325 (1.00)	0.0915 (1.00)	0.427 (1.00)	0.392 (1.00)	0.471 (1.00)	0.351 (1.00)	0.623 (1.00)	0.398 (1.00)

Clustering Approach #1: 'Copy Number Ratio CNMF subtypes'

Table S1. Description of clustering approach #1: 'Copy Number Ratio CNMF subtypes'

Cluster Labels	1	2	3	4	5
Number of samples	9	31	5	6	1

'Copy Number Ratio CNMF subtypes' versus 'Time to Death'

P value = 0.172 (logrank test), Q value = 1

Table S2. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	48	17	0.0 - 49.4 (4.9)
subtype1	9	4	0.1 - 17.1 (4.8)
subtype2	29	10	0.1 - 49.4 (5.2)
subtype3	5	3	0.0 - 19.7 (3.9)
subtype4	5	0	0.2 - 12.0 (1.5)

Figure S1. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

'Copy Number Ratio CNMF subtypes' versus 'AGE'

P value = 0.995 (ANOVA), Q value = 1

Table S3. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	51	66.0 (10.7)
subtype1	9	66.2 (11.5)
subtype2	31	66.2 (11.5)
subtype3	5	65.4 (7.1)
subtype4	6	65.0 (9.1)

Figure S2. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #2: 'AGE'

'Copy Number Ratio CNMF subtypes' versus 'NEOPLASM.DISEASESTAGE'

P value = 0.047 (Chi-square test), Q value = 1

Table S4. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

nPatients	STAGE IA	STAGE IB	STAGE IIA	STAGE IIB	STAGE III	STAGE IV
ALL	2	2	6	37	2	2
subtype1	0	0	0	9	0	0
subtype2	2	1	5	21	2	0
subtype3	0	1	1	3	0	0
subtype4	0	0	0	4	0	2

Figure S3. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

'Copy Number Ratio CNMF subtypes' versus 'PATHOLOGY.T.STAGE'

P value = 0.468 (Fisher's exact test), Q value = 1

Table S5. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

nPatients	T1+T2	T3+T4
ALL	5	46
subtype1	0	9
subtype2	4	27
subtype3	1	4
subtype4	0	6

Figure S4. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

'Copy Number Ratio CNMF subtypes' versus 'PATHOLOGY.N.STAGE'

P value = 0.174 (Fisher's exact test), Q value = 1

Table S6. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

nPatients	0	1
ALL	12	39
subtype1	0	9
subtype2	9	22
subtype3	2	3
subtype4	1	5

Figure S5. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

'Copy Number Ratio CNMF subtypes' versus 'PATHOLOGY.M.STAGE'

P value = 0.00123 (Chi-square test), Q value = 0.12

Table S7. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

nPatients	M0	M1	MX
ALL	27	2	22
subtype1	6	0	3
subtype2	15	0	16
subtype3	5	0	0
subtype4	1	2	3

Figure S6. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

'Copy Number Ratio CNMF subtypes' versus 'GENDER'

P value = 0.169 (Fisher's exact test), Q value = 1

Table S8. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	28	23
subtype1	2	7
subtype2	20	11
subtype3	3	2
subtype4	3	3

Figure S7. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #7: 'GENDER'

'Copy Number Ratio CNMF subtypes' versus 'HISTOLOGICAL.TYPE'

P value = 0.923 (Chi-square test), Q value = 1

Table S9. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

nPatients	PANCREAS-ADENOCARCINOMA DUCTAL TYPE	PANCREAS-ADENOCARCINOMA-OTHER SUBTYPE	PANCREAS-COLLOID (MUCINOUS NON-CYSTIC) CARCINOMA
ALL	45	4	1
subtype1	7	1	0
subtype2	28	2	1
subtype3	5	0	0
subtype4	5	1	0

Figure S8. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

'Copy Number Ratio CNMF subtypes' versus 'COMPLETENESS.OF.RESECTION'

P value = 0.113 (Chi-square test), Q value = 1

Table S10. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

nPatients	R0	R1	RX
ALL	28	19	2
subtype1	3	5	1
subtype2	21	8	1
subtype3	3	1	0
subtype4	1	5	0

Figure S9. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

'Copy Number Ratio CNMF subtypes' versus 'NUMBER.OF.LYMPH.NODES'

P value = 0.325 (ANOVA), Q value = 1

Table S11. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

	nPatients	Mean (Std.Dev)
ALL	51	2.8 (2.9)
subtype1	9	4.3 (3.4)
subtype2	31	2.7 (3.1)
subtype3	5	1.8 (1.8)
subtype4	6	2.0 (1.1)

Figure S10. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

Clustering Approach #2: 'METHLYATION CNMF'

Table S12. Description of clustering approach #2: 'METHLYATION CNMF'

Cluster Labels	1	2	3
Number of samples	17	24	12

'METHLYATION CNMF' versus 'Time to Death'

P value = 0.447 (logrank test), Q value = 1

Table S13. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	50	17	0.0 - 49.4 (4.9)
subtype1	16	5	0.0 - 19.7 (3.9)
subtype2	22	5	0.1 - 28.0 (2.0)
subtype3	12	7	0.3 - 49.4 (10.9)

Figure S11. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #1: 'Time to Death'

'METHLYATION CNMF' versus 'AGE'

P value = 0.645 (ANOVA), Q value = 1

Table S14. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	53	65.9 (10.7)
subtype1	17	64.4 (9.5)
subtype2	24	67.4 (9.9)
subtype3	12	65.0 (13.9)

Figure S12. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #2: 'AGE'

'METHLYATION CNMF' versus 'NEOPLASM.DISEASESTAGE'

P value = 0.536 (Chi-square test), Q value = 1

Table S15. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

nPatients	STAGE IA	STAGE IB	STAGE IIA	STAGE IIB	STAGE III	STAGE IV
ALL	2	2	6	39	2	2
subtype1	0	1	2	12	0	2
subtype2	2	1	3	17	1	0
subtype3	0	0	1	10	1	0

Figure S13. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

'METHLYATION CNMF' versus 'PATHOLOGY.T.STAGE'

P value = 0.287 (Fisher's exact test), Q value = 1

Table S16. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

nPatients	T1+T2	T3+T4
ALL	5	48
subtype1	1	16
subtype2	4	20
subtype3	0	12

Figure S14. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

'METHLYATION CNMF' versus 'PATHOLOGY.N.STAGE'

P value = 0.917 (Fisher's exact test), Q value = 1

Table S17. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

nPatients	0	1
ALL	12	41
subtype1	4	13
subtype2	6	18
subtype3	2	10

Figure S15. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

'METHLYATION CNMF' versus 'PATHOLOGY.M.STAGE'

P value = 0.142 (Chi-square test), Q value = 1

Table S18. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

nPatients	M0	M1	MX
ALL	27	2	24
subtype1	9	2	6
subtype2	10	0	14
subtype3	8	0	4

Figure S16. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

'METHLYATION CNMF' versus 'GENDER'

P value = 1 (Fisher's exact test), Q value = 1

Table S19. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	28	25
subtype1	9	8
subtype2	13	11
subtype3	6	6

Figure S17. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #7: 'GENDER'

'METHLYATION CNMF' versus 'HISTOLOGICAL.TYPE'

P value = 0.602 (Chi-square test), Q value = 1

Table S20. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

nPatients	PANCREAS-ADENOCARCINOMA DUCTAL TYPE	PANCREAS-ADENOCARCINOMA-OTHER SUBTYPE	PANCREAS-COLLOID (MUCINOUS NON-CYSTIC) CARCINOMA
ALL	47	4	1
subtype1	14	2	0
subtype2	21	2	1
subtype3	12	0	0

Figure S18. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

'METHLYATION CNMF' versus 'NUMBERPACKYEARSSMOKED'

P value = 0.859 (ANOVA), Q value = 1

Table S21. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

	nPatients	Mean (Std.Dev)
ALL	14	24.2 (15.8)
subtype1	4	22.1 (13.9)
subtype2	8	23.9 (18.5)
subtype3	2	30.0 (14.1)

Figure S19. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

'METHLYATION CNMF' versus 'YEAROFTOBACCOSMOKINGONSET'

P value = 0.279 (ANOVA), Q value = 1

Table S22. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

	nPatients	Mean (Std.Dev)
ALL	11	1971.6 (12.3)
subtype1	3	1975.0 (6.0)
subtype2	6	1967.2 (14.0)
subtype3	2	1980.0 (12.7)

Figure S20. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

'METHLYATION CNMF' versus 'COMPLETENESS.OF.RESECTION'

P value = 0.0977 (Chi-square test), Q value = 1

Table S23. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

nPatients	R0	R1	RX
ALL	28	20	2
subtype1	6	9	1
subtype2	17	5	0
subtype3	5	6	1

Figure S21. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

'METHLYATION CNMF' versus 'NUMBER.OF.LYMPH.NODES'

P value = 0.0915 (ANOVA), Q value = 1

Table S24. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

	nPatients	Mean (Std.Dev)
ALL	53	2.8 (2.9)
subtype1	17	1.9 (1.4)
subtype2	24	2.8 (3.0)
subtype3	12	4.2 (3.8)

Figure S22. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

Clustering Approach #3: 'RNAseq CNMF subtypes'

Table S25. Description of clustering approach #3: 'RNAseq CNMF subtypes'

Cluster Labels	1	2	3	4
Number of samples	17	11	12	10

'RNAseq CNMF subtypes' versus 'Time to Death'

P value = 0.0897 (logrank test), Q value = 1

Table S26. Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	47	16	0.0 - 49.4 (4.8)
subtype1	17	7	0.0 - 19.8 (4.0)
subtype2	11	3	0.2 - 16.0 (5.2)
subtype3	10	5	0.1 - 43.8 (6.5)
subtype4	9	1	0.3 - 49.4 (3.0)

Figure S23. Get High-res Image Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

'RNAseq CNMF subtypes' versus 'AGE'

P value = 0.759 (ANOVA), Q value = 1

Table S27. Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	50	65.3 (10.6)
subtype1	17	64.1 (8.4)
subtype2	11	63.5 (12.3)
subtype3	12	67.5 (11.8)
subtype4	10	66.7 (11.5)

Figure S24. Get High-res Image Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #2: 'AGE'

'RNAseq CNMF subtypes' versus 'NEOPLASM.DISEASESTAGE'

P value = 0.841 (Chi-square test), Q value = 1

Table S28. Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

nPatients	STAGE IA	STAGE IB	STAGE IIA	STAGE IIB	STAGE III	STAGE IV
ALL	2	2	6	36	2	2
subtype1	0	1	2	13	0	1
subtype2	0	0	1	9	1	0
subtype3	1	1	1	8	1	0
subtype4	1	0	2	6	0	1

Figure S25. Get High-res Image Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

'RNAseq CNMF subtypes' versus 'PATHOLOGY.T.STAGE'

P value = 0.915 (Fisher's exact test), Q value = 1

Table S29. Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

nPatients	T1+T2	T3+T4
ALL	5	45
subtype1	1	16
subtype2	1	10
subtype3	2	10
subtype4	1	9

Figure S26. Get High-res Image Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

'RNAseq CNMF subtypes' versus 'PATHOLOGY.N.STAGE'

P value = 0.576 (Fisher's exact test), Q value = 1

Table S30. Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

nPatients	0	1
ALL	12	38
subtype1	4	13
subtype2	1	10
subtype3	4	8
subtype4	3	7

Figure S27. Get High-res Image Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

'RNAseq CNMF subtypes' versus 'PATHOLOGY.M.STAGE'

P value = 0.0611 (Chi-square test), Q value = 1

Table S31. Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

nPatients	M0	M1	MX
ALL	26	2	22
subtype1	12	1	4
subtype2	8	0	3
subtype3	4	0	8
subtype4	2	1	7

Figure S28. Get High-res Image Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

'RNAseq CNMF subtypes' versus 'GENDER'

P value = 0.746 (Fisher's exact test), Q value = 1

Table S32. Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	28	22
subtype1	9	8
subtype2	5	6
subtype3	7	5
subtype4	7	3

Figure S29. Get High-res Image Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #7: 'GENDER'

'RNAseq CNMF subtypes' versus 'HISTOLOGICAL.TYPE'

P value = 0.273 (Chi-square test), Q value = 1

Table S33. Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

nPatients	PANCREAS-ADENOCARCINOMA DUCTAL TYPE	PANCREAS-ADENOCARCINOMA-OTHER SUBTYPE	PANCREAS-COLLOID (MUCINOUS NON-CYSTIC) CARCINOMA
ALL	45	3	1
subtype1	15	1	0
subtype2	11	0	0
subtype3	10	2	0
subtype4	9	0	1

Figure S30. Get High-res Image Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

'RNAseq CNMF subtypes' versus 'NUMBERPACKYEARSSMOKED'

P value = 0.706 (ANOVA), Q value = 1

Table S34. Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

	nPatients	Mean (Std.Dev)
ALL	13	25.0 (16.3)
subtype1	5	23.7 (14.4)
subtype2	4	28.8 (21.8)
subtype3	2	20.5 (27.6)
subtype4	2	25.0 (7.1)

Figure S31. Get High-res Image Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

'RNAseq CNMF subtypes' versus 'YEAROFTOBACCOSMOKINGONSET'

P value = 0.497 (ANOVA), Q value = 1

Table S35. Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

	nPatients	Mean (Std.Dev)
ALL	11	1971.6 (12.3)
subtype1	4	1976.0 (9.0)
subtype2	4	1968.8 (17.4)
subtype3	1	1968.0 (NA)
subtype4	2	1970.5 (14.8)

Figure S32. Get High-res Image Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

'RNAseq CNMF subtypes' versus 'COMPLETENESS.OF.RESECTION'

P value = 0.217 (Chi-square test), Q value = 1

Table S36. Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

nPatients	R0	R1	RX
ALL	27	19	1
subtype1	7	9	0
subtype2	5	5	0
subtype3	9	2	0
subtype4	6	3	1

Figure S33. Get High-res Image Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

'RNAseq CNMF subtypes' versus 'NUMBER.OF.LYMPH.NODES'

P value = 0.427 (ANOVA), Q value = 1

Table S37. Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

	nPatients	Mean (Std.Dev)
ALL	50	2.8 (2.9)
subtype1	17	2.3 (1.6)
subtype2	11	3.9 (3.1)
subtype3	12	3.0 (4.3)
subtype4	10	2.1 (1.8)

Figure S34. Get High-res Image Clustering Approach #3: 'RNAseq CNMF subtypes' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

Clustering Approach #4: 'RNAseq cHierClus subtypes'

Table S38. Description of clustering approach #4: 'RNAseq cHierClus subtypes'

Cluster Labels	1	2	3
Number of samples	5	19	26

'RNAseq cHierClus subtypes' versus 'Time to Death'

P value = 0.69 (logrank test), Q value = 1

Table S39. Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	47	16	0.0 - 49.4 (4.8)
subtype1	5	0	0.3 - 12.1 (1.5)
subtype2	18	7	0.0 - 28.0 (4.9)
subtype3	24	9	0.1 - 49.4 (5.6)

Figure S35. Get High-res Image Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

'RNAseq cHierClus subtypes' versus 'AGE'

P value = 0.793 (ANOVA), Q value = 1

Table S40. Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	50	65.3 (10.6)
subtype1	5	68.4 (10.7)
subtype2	19	65.1 (8.9)
subtype3	26	64.8 (12.0)

Figure S36. Get High-res Image Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #2: 'AGE'

'RNAseq cHierClus subtypes' versus 'NEOPLASM.DISEASESTAGE'

P value = 0.461 (Chi-square test), Q value = 1

Table S41. Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

nPatients	STAGE IA	STAGE IB	STAGE IIA	STAGE IIB	STAGE III	STAGE IV
ALL	2	2	6	36	2	2
subtype1	1	0	1	3	0	0
subtype2	0	1	2	14	0	2
subtype3	1	1	3	19	2	0

Figure S37. Get High-res Image Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

'RNAseq cHierClus subtypes' versus 'PATHOLOGY.T.STAGE'

P value = 0.376 (Fisher's exact test), Q value = 1

Table S42. Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

nPatients	T1+T2	T3+T4
ALL	5	45
subtype1	1	4
subtype2	1	18
subtype3	3	23

Figure S38. Get High-res Image Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

'RNAseq cHierClus subtypes' versus 'PATHOLOGY.N.STAGE'

P value = 0.785 (Fisher's exact test), Q value = 1

Table S43. Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

nPatients	0	1
ALL	12	38
subtype1	2	3
subtype2	4	15
subtype3	6	20

Figure S39. Get High-res Image Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

'RNAseq cHierClus subtypes' versus 'PATHOLOGY.M.STAGE'

P value = 0.0117 (Chi-square test), Q value = 1

Table S44. Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

nPatients	M0	M1	MX
ALL	26	2	22
subtype1	0	0	5
subtype2	13	2	4
subtype3	13	0	13

Figure S40. Get High-res Image Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

'RNAseq cHierClus subtypes' versus 'GENDER'

P value = 1 (Fisher's exact test), Q value = 1

Table S45. Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	28	22
subtype1	3	2
subtype2	11	8
subtype3	14	12

Figure S41. Get High-res Image Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #7: 'GENDER'

'RNAseq cHierClus subtypes' versus 'HISTOLOGICAL.TYPE'

P value = 0.0155 (Chi-square test), Q value = 1

Table S46. Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

nPatients	PANCREAS-ADENOCARCINOMA DUCTAL TYPE	PANCREAS-ADENOCARCINOMA-OTHER SUBTYPE	PANCREAS-COLLOID (MUCINOUS NON-CYSTIC) CARCINOMA
ALL	45	3	1
subtype1	3	1	1
subtype2	18	0	0
subtype3	24	2	0

Figure S42. Get High-res Image Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

'RNAseq cHierClus subtypes' versus 'NUMBERPACKYEARSSMOKED'

P value = 0.85 (ANOVA), Q value = 1

Table S47. Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

	nPatients	Mean (Std.Dev)
ALL	13	25.0 (16.3)
subtype2	7	24.1 (12.1)
subtype3	6	26.0 (21.3)

Figure S43. Get High-res Image Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

'RNAseq cHierClus subtypes' versus 'YEAROFTOBACCOSMOKINGONSET'

P value = 0.505 (ANOVA), Q value = 1

Table S48. Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

	nPatients	Mean (Std.Dev)
ALL	11	1971.6 (12.3)
subtype2	6	1974.2 (10.0)
subtype3	5	1968.6 (15.1)

Figure S44. Get High-res Image Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

'RNAseq cHierClus subtypes' versus 'COMPLETENESS.OF.RESECTION'

P value = 0.0196 (Chi-square test), Q value = 1

Table S49. Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

nPatients	R0	R1	RX
ALL	27	19	1
subtype1	2	2	1
subtype2	8	10	0
subtype3	17	7	0

Figure S45. Get High-res Image Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

'RNAseq cHierClus subtypes' versus 'NUMBER.OF.LYMPH.NODES'

P value = 0.392 (ANOVA), Q value = 1

Table S50. Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

	nPatients	Mean (Std.Dev)
ALL	50	2.8 (2.9)
subtype1	5	1.6 (1.7)
subtype2	19	2.4 (1.6)
subtype3	26	3.3 (3.6)

Figure S46. Get High-res Image Clustering Approach #4: 'RNAseq cHierClus subtypes' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

Clustering Approach #5: 'MIRSEQ CNMF'

Table S51. Description of clustering approach #5: 'MIRSEQ CNMF'

Cluster Labels	1	2	3
Number of samples	25	11	17

'MIRSEQ CNMF' versus 'Time to Death'

P value = 0.778 (logrank test), Q value = 1

Table S52. Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	50	17	0.0 - 49.4 (4.9)
subtype1	24	8	0.0 - 28.0 (4.4)
subtype2	11	2	0.2 - 16.0 (5.2)
subtype3	15	7	0.1 - 49.4 (6.0)

Figure S47. Get High-res Image Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #1: 'Time to Death'

'MIRSEQ CNMF' versus 'AGE'

P value = 0.799 (ANOVA), Q value = 1

Table S53. Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	53	65.9 (10.7)
subtype1	25	66.7 (9.7)
subtype2	11	64.1 (11.1)
subtype3	17	65.9 (12.1)

Figure S48. Get High-res Image Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #2: 'AGE'

'MIRSEQ CNMF' versus 'NEOPLASM.DISEASESTAGE'

P value = 0.588 (Chi-square test), Q value = 1

Table S54. Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

nPatients	STAGE IA	STAGE IB	STAGE IIA	STAGE IIB	STAGE III	STAGE IV
ALL	2	2	6	39	2	2
subtype1	1	1	3	18	0	2
subtype2	0	0	1	10	0	0
subtype3	1	1	2	11	2	0

Figure S49. Get High-res Image Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

'MIRSEQ CNMF' versus 'PATHOLOGY.T.STAGE'

P value = 1 (Fisher's exact test), Q value = 1

Table S55. Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

nPatients	T1+T2	T3+T4
ALL	5	48
subtype1	2	23
subtype2	1	10
subtype3	2	15

Figure S50. Get High-res Image Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

'MIRSEQ CNMF' versus 'PATHOLOGY.N.STAGE'

P value = 0.53 (Fisher's exact test), Q value = 1

Table S56. Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

nPatients	0	1
ALL	12	41
subtype1	6	19
subtype2	1	10
subtype3	5	12

Figure S51. Get High-res Image Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

'MIRSEQ CNMF' versus 'PATHOLOGY.M.STAGE'

P value = 0.413 (Chi-square test), Q value = 1

Table S57. Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

nPatients	M0	M1	MX
ALL	27	2	24
subtype1	13	2	10
subtype2	7	0	4
subtype3	7	0	10

Figure S52. Get High-res Image Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

'MIRSEQ CNMF' versus 'GENDER'

P value = 0.825 (Fisher's exact test), Q value = 1

Table S58. Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	28	25
subtype1	13	12
subtype2	5	6
subtype3	10	7

Figure S53. Get High-res Image Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #7: 'GENDER'

'MIRSEQ CNMF' versus 'HISTOLOGICAL.TYPE'

P value = 0.638 (Chi-square test), Q value = 1

Table S59. Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

nPatients	PANCREAS-ADENOCARCINOMA DUCTAL TYPE	PANCREAS-ADENOCARCINOMA-OTHER SUBTYPE	PANCREAS-COLLOID (MUCINOUS NON-CYSTIC) CARCINOMA
ALL	47	4	1
subtype1	21	2	1
subtype2	11	0	0
subtype3	15	2	0

Figure S54. Get High-res Image Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

'MIRSEQ CNMF' versus 'NUMBERPACKYEARSSMOKED'

P value = 0.427 (ANOVA), Q value = 1

Table S60. Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

	nPatients	Mean (Std.Dev)
ALL	14	24.2 (15.8)
subtype1	7	24.1 (12.1)
subtype2	4	31.5 (18.0)
subtype3	3	15.0 (21.7)

Figure S55. Get High-res Image Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

'MIRSEQ CNMF' versus 'YEAROFTOBACCOSMOKINGONSET'

P value = 0.908 (ANOVA), Q value = 1

Table S61. Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

	nPatients	Mean (Std.Dev)
ALL	11	1971.6 (12.3)
subtype1	6	1974.2 (10.0)
subtype2	3	1972.7 (19.1)
subtype3	2	1962.5 (7.8)

Figure S56. Get High-res Image Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

'MIRSEQ CNMF' versus 'COMPLETENESS.OF.RESECTION'

P value = 0.23 (Chi-square test), Q value = 1

Table S62. Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

nPatients	R0	R1	RX
ALL	28	20	2
subtype1	10	12	2
subtype2	6	4	0
subtype3	12	4	0

Figure S57. Get High-res Image Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

'MIRSEQ CNMF' versus 'NUMBER.OF.LYMPH.NODES'

P value = 0.471 (ANOVA), Q value = 1

Table S63. Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

	nPatients	Mean (Std.Dev)
ALL	53	2.8 (2.9)
subtype1	25	2.4 (1.9)
subtype2	11	3.6 (3.2)
subtype3	17	2.9 (3.8)

Figure S58. Get High-res Image Clustering Approach #5: 'MIRSEQ CNMF' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

Clustering Approach #6: 'MIRSEQ CHIERARCHICAL'

Table S64. Description of clustering approach #6: 'MIRSEQ CHIERARCHICAL'

Cluster Labels	1	2	3
Number of samples	11	19	23

'MIRSEQ CHIERARCHICAL' versus 'Time to Death'

P value = 0.49 (logrank test), Q value = 1

Table S65. Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	50	17	0.0 - 49.4 (4.9)
subtype1	9	5	0.1 - 49.4 (7.6)
subtype2	19	5	0.1 - 19.8 (1.0)
subtype3	22	7	0.0 - 28.0 (4.4)

Figure S59. Get High-res Image Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #1: 'Time to Death'

'MIRSEQ CHIERARCHICAL' versus 'AGE'

P value = 0.565 (ANOVA), Q value = 1

Table S66. Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	53	65.9 (10.7)
subtype1	11	65.5 (14.2)
subtype2	19	64.1 (9.6)
subtype3	23	67.6 (9.8)

Figure S60. Get High-res Image Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #2: 'AGE'

'MIRSEQ CHIERARCHICAL' versus 'NEOPLASM.DISEASESTAGE'

P value = 0.587 (Chi-square test), Q value = 1

Table S67. Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

nPatients	STAGE IA	STAGE IB	STAGE IIA	STAGE IIB	STAGE III	STAGE IV
ALL	2	2	6	39	2	2
subtype1	1	1	2	7	0	0
subtype2	0	0	1	17	1	0
subtype3	1	1	3	15	1	2

Figure S61. Get High-res Image Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

'MIRSEQ CHIERARCHICAL' versus 'PATHOLOGY.T.STAGE'

P value = 0.6 (Fisher's exact test), Q value = 1

Table S68. Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

nPatients	T1+T2	T3+T4
ALL	5	48
subtype1	2	9
subtype2	1	18
subtype3	2	21

Figure S62. Get High-res Image Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

'MIRSEQ CHIERARCHICAL' versus 'PATHOLOGY.N.STAGE'

P value = 0.0637 (Fisher's exact test), Q value = 1

Table S69. Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

nPatients	0	1
ALL	12	41
subtype1	4	7
subtype2	1	18
subtype3	7	16

Figure S63. Get High-res Image Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

'MIRSEQ CHIERARCHICAL' versus 'PATHOLOGY.M.STAGE'

P value = 0.366 (Chi-square test), Q value = 1

Table S70. Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

nPatients	M0	M1	MX
ALL	27	2	24
subtype1	4	0	7
subtype2	11	0	8
subtype3	12	2	9

Figure S64. Get High-res Image Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

'MIRSEQ CHIERARCHICAL' versus 'GENDER'

P value = 0.724 (Fisher's exact test), Q value = 1

Table S71. Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	28	25
subtype1	7	4
subtype2	10	9
subtype3	11	12

Figure S65. Get High-res Image Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #7: 'GENDER'

'MIRSEQ CHIERARCHICAL' versus 'HISTOLOGICAL.TYPE'

P value = 0.311 (Chi-square test), Q value = 1

Table S72. Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

nPatients	PANCREAS-ADENOCARCINOMA DUCTAL TYPE	PANCREAS-ADENOCARCINOMA-OTHER SUBTYPE	PANCREAS-COLLOID (MUCINOUS NON-CYSTIC) CARCINOMA
ALL	47	4	1
subtype1	9	2	0
subtype2	19	0	0
subtype3	19	2	1

Figure S66. Get High-res Image Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

'MIRSEQ CHIERARCHICAL' versus 'NUMBERPACKYEARSSMOKED'

P value = 0.979 (ANOVA), Q value = 1

Table S73. Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

	nPatients	Mean (Std.Dev)
ALL	14	24.2 (15.8)
subtype1	2	20.5 (27.6)
subtype2	6	25.0 (17.9)
subtype3	6	24.8 (13.1)

Figure S67. Get High-res Image Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

'MIRSEQ CHIERARCHICAL' versus 'YEAROFTOBACCOSMOKINGONSET'

P value = 0.859 (ANOVA), Q value = 1

Table S74. Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

	nPatients	Mean (Std.Dev)
ALL	11	1971.6 (12.3)
subtype1	1	1968.0 (NA)
subtype2	5	1972.8 (17.6)
subtype3	5	1971.2 (7.8)

Figure S68. Get High-res Image Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

'MIRSEQ CHIERARCHICAL' versus 'COMPLETENESS.OF.RESECTION'

P value = 0.298 (Chi-square test), Q value = 1

Table S75. Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

nPatients	R0	R1	RX
ALL	28	20	2
subtype1	8	2	0
subtype2	11	6	1
subtype3	9	12	1

Figure S69. Get High-res Image Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

'MIRSEQ CHIERARCHICAL' versus 'NUMBER.OF.LYMPH.NODES'

P value = 0.351 (ANOVA), Q value = 1

Table S76. Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

	nPatients	Mean (Std.Dev)
ALL	53	2.8 (2.9)
subtype1	11	2.8 (4.1)
subtype2	19	3.5 (2.9)
subtype3	23	2.2 (2.1)

Figure S70. Get High-res Image Clustering Approach #6: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

Clustering Approach #7: 'MIRseq Mature CNMF subtypes'

Table S77. Description of clustering approach #7: 'MIRseq Mature CNMF subtypes'

Cluster Labels	1	2	3
Number of samples	21	9	23

'MIRseq Mature CNMF subtypes' versus 'Time to Death'

P value = 0.399 (logrank test), Q value = 1

Table S78. Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	50	17	0.0 - 49.4 (4.9)
subtype1	20	7	0.0 - 28.0 (4.4)
subtype2	9	1	0.1 - 19.8 (10.0)
subtype3	21	9	0.1 - 49.4 (5.2)

Figure S71. Get High-res Image Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

'MIRseq Mature CNMF subtypes' versus 'AGE'

P value = 0.295 (ANOVA), Q value = 1

Table S79. Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	53	65.9 (10.7)
subtype1	21	67.9 (10.1)
subtype2	9	61.2 (8.5)
subtype3	23	65.9 (11.7)

Figure S72. Get High-res Image Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #2: 'AGE'

'MIRseq Mature CNMF subtypes' versus 'NEOPLASM.DISEASESTAGE'

P value = 0.73 (Chi-square test), Q value = 1

Table S80. Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

nPatients	STAGE IA	STAGE IB	STAGE IIA	STAGE IIB	STAGE III	STAGE IV
ALL	2	2	6	39	2	2
subtype1	1	1	2	15	0	2
subtype2	0	0	1	8	0	0
subtype3	1	1	3	16	2	0

Figure S73. Get High-res Image Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

'MIRseq Mature CNMF subtypes' versus 'PATHOLOGY.T.STAGE'

P value = 0.833 (Fisher's exact test), Q value = 1

Table S81. Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

nPatients	T1+T2	T3+T4
ALL	5	48
subtype1	2	19
subtype2	0	9
subtype3	3	20

Figure S74. Get High-res Image Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

'MIRseq Mature CNMF subtypes' versus 'PATHOLOGY.N.STAGE'

P value = 0.826 (Fisher's exact test), Q value = 1

Table S82. Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

nPatients	0	1
ALL	12	41
subtype1	5	16
subtype2	1	8
subtype3	6	17

Figure S75. Get High-res Image Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

'MIRseq Mature CNMF subtypes' versus 'PATHOLOGY.M.STAGE'

P value = 0.308 (Chi-square test), Q value = 1

Table S83. Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

nPatients	M0	M1	MX
ALL	27	2	24
subtype1	12	2	7
subtype2	5	0	4
subtype3	10	0	13

Figure S76. Get High-res Image Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

'MIRseq Mature CNMF subtypes' versus 'GENDER'

P value = 0.934 (Fisher's exact test), Q value = 1

Table S84. Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	28	25
subtype1	10	11
subtype2	5	4
subtype3	13	10

Figure S77. Get High-res Image Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #7: 'GENDER'

'MIRseq Mature CNMF subtypes' versus 'HISTOLOGICAL.TYPE'

P value = 0.623 (Chi-square test), Q value = 1

Table S85. Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

nPatients	PANCREAS-ADENOCARCINOMA DUCTAL TYPE	PANCREAS-ADENOCARCINOMA-OTHER SUBTYPE	PANCREAS-COLLOID (MUCINOUS NON-CYSTIC) CARCINOMA
ALL	47	4	1
subtype1	17	2	1
subtype2	9	0	0
subtype3	21	2	0

Figure S78. Get High-res Image Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

'MIRseq Mature CNMF subtypes' versus 'NUMBERPACKYEARSSMOKED'

P value = 0.945 (ANOVA), Q value = 1

Table S86. Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

	nPatients	Mean (Std.Dev)
ALL	14	24.2 (15.8)
subtype1	6	24.8 (13.1)
subtype2	3	21.3 (7.1)
subtype3	5	25.4 (23.8)

Figure S79. Get High-res Image Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

'MIRseq Mature CNMF subtypes' versus 'YEAROFTOBACCOSMOKINGONSET'

P value = 0.561 (ANOVA), Q value = 1

Table S87. Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

	nPatients	Mean (Std.Dev)
ALL	11	1971.6 (12.3)
subtype1	5	1971.2 (7.8)
subtype2	2	1984.5 (6.4)
subtype3	4	1965.8 (15.8)

Figure S80. Get High-res Image Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

'MIRseq Mature CNMF subtypes' versus 'COMPLETENESS.OF.RESECTION'

P value = 0.0889 (Chi-square test), Q value = 1

Table S88. Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

nPatients	R0	R1	RX
ALL	28	20	2
subtype1	7	12	1
subtype2	6	2	1
subtype3	15	6	0

Figure S81. Get High-res Image Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

'MIRseq Mature CNMF subtypes' versus 'NUMBER.OF.LYMPH.NODES'

P value = 0.623 (ANOVA), Q value = 1

Table S89. Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

	nPatients	Mean (Std.Dev)
ALL	53	2.8 (2.9)
subtype1	21	2.4 (2.0)
subtype2	9	3.6 (3.5)
subtype3	23	2.9 (3.3)

Figure S82. Get High-res Image Clustering Approach #7: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

Clustering Approach #8: 'MIRseq Mature cHierClus subtypes'

Table S90. Description of clustering approach #8: 'MIRseq Mature cHierClus subtypes'

Cluster Labels	1	2	3
Number of samples	12	22	19

'MIRseq Mature cHierClus subtypes' versus 'Time to Death'

P value = 0.49 (logrank test), Q value = 1

Table S91. Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	50	17	0.0 - 49.4 (4.9)
subtype1	10	5	0.1 - 49.4 (6.8)
subtype2	21	7	0.0 - 28.0 (4.8)
subtype3	19	5	0.1 - 19.8 (1.0)

Figure S83. Get High-res Image Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

'MIRseq Mature cHierClus subtypes' versus 'AGE'

P value = 0.587 (ANOVA), Q value = 1

Table S92. Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	53	65.9 (10.7)
subtype1	12	65.8 (13.6)
subtype2	22	67.5 (10.0)
subtype3	19	64.1 (9.6)

Figure S84. Get High-res Image Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #2: 'AGE'

'MIRseq Mature cHierClus subtypes' versus 'NEOPLASM.DISEASESTAGE'

P value = 0.496 (Chi-square test), Q value = 1

Table S93. Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

nPatients	STAGE IA	STAGE IB	STAGE IIA	STAGE IIB	STAGE III	STAGE IV
ALL	2	2	6	39	2	2
subtype1	1	1	2	7	1	0
subtype2	1	1	3	15	0	2
subtype3	0	0	1	17	1	0

Figure S85. Get High-res Image Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #3: 'NEOPLASM.DISEASESTAGE'

'MIRseq Mature cHierClus subtypes' versus 'PATHOLOGY.T.STAGE'

P value = 0.712 (Fisher's exact test), Q value = 1

Table S94. Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

nPatients	T1+T2	T3+T4
ALL	5	48
subtype1	2	10
subtype2	2	20
subtype3	1	18

Figure S86. Get High-res Image Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #4: 'PATHOLOGY.T.STAGE'

'MIRseq Mature cHierClus subtypes' versus 'PATHOLOGY.N.STAGE'

P value = 0.0332 (Fisher's exact test), Q value = 1

Table S95. Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

nPatients	0	1
ALL	12	41
subtype1	5	7
subtype2	6	16
subtype3	1	18

Figure S87. Get High-res Image Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #5: 'PATHOLOGY.N.STAGE'

'MIRseq Mature cHierClus subtypes' versus 'PATHOLOGY.M.STAGE'

P value = 0.246 (Chi-square test), Q value = 1

Table S96. Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

nPatients	M0	M1	MX
ALL	27	2	24
subtype1	4	0	8
subtype2	12	2	8
subtype3	11	0	8

Figure S88. Get High-res Image Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #6: 'PATHOLOGY.M.STAGE'

'MIRseq Mature cHierClus subtypes' versus 'GENDER'

P value = 0.939 (Fisher's exact test), Q value = 1

Table S97. Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	28	25
subtype1	7	5
subtype2	11	11
subtype3	10	9

Figure S89. Get High-res Image Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #7: 'GENDER'

'MIRseq Mature cHierClus subtypes' versus 'HISTOLOGICAL.TYPE'

P value = 0.331 (Chi-square test), Q value = 1

Table S98. Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

nPatients	PANCREAS-ADENOCARCINOMA DUCTAL TYPE	PANCREAS-ADENOCARCINOMA-OTHER SUBTYPE	PANCREAS-COLLOID (MUCINOUS NON-CYSTIC) CARCINOMA
ALL	47	4	1
subtype1	10	2	0
subtype2	18	2	1
subtype3	19	0	0

Figure S90. Get High-res Image Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #8: 'HISTOLOGICAL.TYPE'

'MIRseq Mature cHierClus subtypes' versus 'NUMBERPACKYEARSSMOKED'

P value = 0.979 (ANOVA), Q value = 1

Table S99. Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

	nPatients	Mean (Std.Dev)
ALL	14	24.2 (15.8)
subtype1	2	20.5 (27.6)
subtype2	6	24.8 (13.1)
subtype3	6	25.0 (17.9)

Figure S91. Get High-res Image Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #9: 'NUMBERPACKYEARSSMOKED'

'MIRseq Mature cHierClus subtypes' versus 'YEAROFTOBACCOSMOKINGONSET'

P value = 0.859 (ANOVA), Q value = 1

Table S100. Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

	nPatients	Mean (Std.Dev)
ALL	11	1971.6 (12.3)
subtype1	1	1968.0 (NA)
subtype2	5	1971.2 (7.8)
subtype3	5	1972.8 (17.6)

Figure S92. Get High-res Image Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #10: 'YEAROFTOBACCOSMOKINGONSET'

'MIRseq Mature cHierClus subtypes' versus 'COMPLETENESS.OF.RESECTION'

P value = 0.183 (Chi-square test), Q value = 1

Table S101. Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

nPatients	R0	R1	RX
ALL	28	20	2
subtype1	9	2	0
subtype2	8	12	1
subtype3	11	6	1

Figure S93. Get High-res Image Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #11: 'COMPLETENESS.OF.RESECTION'

'MIRseq Mature cHierClus subtypes' versus 'NUMBER.OF.LYMPH.NODES'

P value = 0.398 (ANOVA), Q value = 1

Table S102. Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

	nPatients	Mean (Std.Dev)
ALL	53	2.8 (2.9)
subtype1	12	2.6 (4.0)
subtype2	22	2.3 (2.1)
subtype3	19	3.5 (2.9)

Figure S94. Get High-res Image Clustering Approach #8: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #12: 'NUMBER.OF.LYMPH.NODES'

Methods & Data

Input

Cluster data file = PAAD-TP.mergedcluster.txt
Clinical data file = PAAD-TP.clin.merged.picked.txt
Number of patients = 53
Number of clustering approaches = 8
Number of selected clinical features = 12
Exclude small clusters that include fewer than K patients, K = 3

Clustering approaches

CNMF clustering

consensus non-negative matrix factorization clustering approach (Brunet et al. 2004)

Consensus hierarchical clustering

Resampling-based clustering method (Monti et al. 2003)

Survival analysis

For survival clinical features, the Kaplan-Meier survival curves of tumors with and without gene mutations were plotted and the statistical significance P values were estimated by logrank test (Bland and Altman 2004) using the 'survdiff' function in R

ANOVA analysis

For continuous numerical clinical features, one-way analysis of variance (Howell 2002) was applied to compare the clinical values between tumor subtypes using 'anova' function in R

Chi-square test

For multi-class clinical features (nominal or ordinal), Chi-square tests (Greenwood and Nikulin 1996) were used to estimate the P values using the 'chisq.test' function in R

Fisher's exact test

For binary clinical features, two-tailed Fisher's exact tests (Fisher 1922) were used to estimate the P values using the 'fisher.test' function in R

Q value calculation

For multiple hypothesis correction, Q value is the False Discovery Rate (FDR) analogue of the P value (Benjamini and Hochberg 1995), defined as the minimum FDR at which the test may be called significant. We used the 'Benjamini and Hochberg' method of 'p.adjust' function in R to convert P values into Q values.

Download Results

In addition to the links below, the full results of the analysis summarized in this report can also be downloaded programmatically using firehose_get, or interactively from either the Broad GDAC website or TCGA Data Coordination Center Portal.

References

[1] Brunet et al., Metagenes and molecular pattern discovery using matrix factorization, PNAS 101(12):4164-9 (2004)

[2] Monti et al., Consensus Clustering: A Resampling-Based Method for Class Discovery and Visualization of Gene Expression Microarray Data, Machine Learning 52(1):91-118 (2003)

[3] Bland and Altman, Statistics notes: The logrank test, BMJ 328(7447):1073 (2004)

[4] Howell, D, Statistical Methods for Psychology. (5th ed.), Duxbury Press:324-5 (2002)

[5] Greenwood and Nikulin, A guide to chi-squared testing, Wiley, New York. ISBN 047155779X (1996)

[6] Fisher, R.A., On the interpretation of chi-square from contingency tables, and the calculation of P, Journal of the Royal Statistical Society 85(1):87-94 (1922)

[7] Benjamini and Hochberg, Controlling the false discovery rate: a practical and powerful approach to multiple testing, Journal of the Royal Statistical Society Series B 59:289-300 (1995)

Made with Nozzle