Correlation between molecular cancer subtypes and selected clinical features

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 6 clinical features across 218 patients, 18 significant findings detected with P value < 0.05.

3 subtypes identified in current cancer cohort by 'CN CNMF'. These subtypes correlate to 'AGE'.
3 subtypes identified in current cancer cohort by 'METHLYATION CNMF'. These subtypes correlate to 'AGE' and 'HISTOLOGICAL.TYPE'.
CNMF clustering analysis on RPPA data identified 3 subtypes that correlate to 'AGE', 'HISTOLOGICAL.TYPE', and 'RADIATIONS.RADIATION.REGIMENINDICATION'.
Consensus hierarchical clustering analysis on RPPA data identified 3 subtypes that correlate to 'AGE' and 'HISTOLOGICAL.TYPE'.
CNMF clustering analysis on sequencing-based mRNA expression data identified 4 subtypes that correlate to 'AGE', 'HISTOLOGICAL.TYPE', and 'RADIATIONS.RADIATION.REGIMENINDICATION'.
Consensus hierarchical clustering analysis on sequencing-based mRNA expression data identified 3 subtypes that correlate to 'AGE', 'HISTOLOGICAL.TYPE', and 'RADIATIONS.RADIATION.REGIMENINDICATION'.
CNMF clustering analysis on sequencing-based miR expression data identified 3 subtypes that correlate to 'HISTOLOGICAL.TYPE'.
Consensus hierarchical clustering analysis on sequencing-based miR expression data identified 3 subtypes that correlate to 'AGE', 'HISTOLOGICAL.TYPE', and 'RADIATIONS.RADIATION.REGIMENINDICATION'.

Results

Overview of the results

Table 1. Get Full Table Overview of the association between subtypes identified by 8 different clustering approaches and 6 clinical features. Shown in the table are P values from statistical tests. Thresholded by P value < 0.05, 18 significant findings detected.


Clinical Features	Time to Death	AGE	GENDER	HISTOLOGICAL TYPE	RADIATIONS RADIATION REGIMENINDICATION	RADIATIONEXPOSURE
Statistical Tests	logrank test	ANOVA	Fisher's exact test	Chi-square test	Fisher's exact test	Fisher's exact test
CN CNMF	100	0.0151	0.773	0.15	0.51	0.435
METHLYATION CNMF	100	0.027	0.849	1.8e-15	0.06	1
RPPA CNMF subtypes	100	0.000166	0.906	2.64e-06	0.00958	0.791
RPPA cHierClus subtypes	100	0.001	0.39	1.74e-11	0.331	0.0769
RNAseq CNMF subtypes	100	0.0108	0.734	8.86e-17	0.00383	0.664
RNAseq cHierClus subtypes	100	0.0163	0.453	6.92e-16	0.0217	1
MIRseq CNMF subtypes	100	0.0556	0.335	6.15e-18	0.468	1
MIRseq cHierClus subtypes	100	0.00403	0.365	6.25e-20	0.0388	0.834

Clustering Approach #1: 'CN CNMF'

Table S1. Get Full Table Description of clustering approach #1: 'CN CNMF'

Cluster Labels	1	2	3
Number of samples	19	159	36

'CN CNMF' versus 'Time to Death'

P value = 100 (logrank test)

Table S2. Clustering Approach #1: 'CN CNMF' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	214	1	0.0 - 66.1 (8.1)
subtype1	19	1	0.4 - 65.9 (12.3)
subtype2	159	0	0.2 - 66.1 (8.1)
subtype3	36	0	0.0 - 65.9 (6.9)

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

'CN CNMF' versus 'AGE'

P value = 0.0151 (ANOVA)

Table S3. Clustering Approach #1: 'CN CNMF' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	214	46.5 (15.9)
subtype1	19	56.5 (13.5)
subtype2	159	45.6 (16.4)
subtype3	36	45.2 (12.8)

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

'CN CNMF' versus 'GENDER'

P value = 0.773 (Fisher's exact test)

Table S4. Clustering Approach #1: 'CN CNMF' versus Clinical Feature #3: 'GENDER'

nPatients	FEMALE	MALE
ALL	160	54
subtype1	13	6
subtype2	119	40
subtype3	28	8

Figure S3. Get High-res Image Clustering Approach #1: 'CN CNMF' versus Clinical Feature #3: 'GENDER'

'CN CNMF' versus 'HISTOLOGICAL.TYPE'

P value = 0.15 (Chi-square test)

Table S5. Clustering Approach #1: 'CN CNMF' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

nPatients	OTHER	THYROID PAPILLARY CARCINOMA - CLASSICAL/USUAL	THYROID PAPILLARY CARCINOMA - FOLLICULAR (>= 99% FOLLICULAR PATTERNED)	THYROID PAPILLARY CARCINOMA - TALL CELL (>= 50% TALL CELL FEATURES)
ALL	8	123	61	22
subtype1	2	8	8	1
subtype2	4	98	39	18
subtype3	2	17	14	3

Figure S4. Get High-res Image Clustering Approach #1: 'CN CNMF' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

'CN CNMF' versus 'RADIATIONS.RADIATION.REGIMENINDICATION'

P value = 0.51 (Fisher's exact test)

Table S6. Clustering Approach #1: 'CN CNMF' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

nPatients	NO	YES
ALL	14	200
subtype1	2	17
subtype2	11	148
subtype3	1	35

Figure S5. Get High-res Image Clustering Approach #1: 'CN CNMF' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

'CN CNMF' versus 'RADIATIONEXPOSURE'

P value = 0.435 (Fisher's exact test)

Table S7. Clustering Approach #1: 'CN CNMF' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

nPatients	NO	YES
ALL	177	9
subtype1	17	1
subtype2	127	8
subtype3	33	0

Figure S6. Get High-res Image Clustering Approach #1: 'CN CNMF' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

Clustering Approach #2: 'METHLYATION CNMF'

Table S8. Get Full Table Description of clustering approach #2: 'METHLYATION CNMF'

Cluster Labels	1	2	3
Number of samples	72	33	113

'METHLYATION CNMF' versus 'Time to Death'

P value = 100 (logrank test)

Table S9. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	218	1	0.0 - 66.1 (8.0)
subtype1	72	1	0.0 - 66.1 (7.0)
subtype2	33	0	0.1 - 66.1 (6.0)
subtype3	113	0	0.2 - 66.1 (10.0)

Figure S7. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #1: 'Time to Death'

'METHLYATION CNMF' versus 'AGE'

P value = 0.027 (ANOVA)

Table S10. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	218	46.5 (15.9)
subtype1	72	50.5 (15.9)
subtype2	33	43.5 (14.6)
subtype3	113	44.8 (15.9)

Figure S8. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #2: 'AGE'

'METHLYATION CNMF' versus 'GENDER'

P value = 0.849 (Fisher's exact test)

Table S11. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #3: 'GENDER'

nPatients	FEMALE	MALE
ALL	162	56
subtype1	53	19
subtype2	26	7
subtype3	83	30

Figure S9. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #3: 'GENDER'

'METHLYATION CNMF' versus 'HISTOLOGICAL.TYPE'

P value = 1.8e-15 (Chi-square test)

Table S12. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

nPatients	OTHER	THYROID PAPILLARY CARCINOMA - CLASSICAL/USUAL	THYROID PAPILLARY CARCINOMA - FOLLICULAR (>= 99% FOLLICULAR PATTERNED)	THYROID PAPILLARY CARCINOMA - TALL CELL (>= 50% TALL CELL FEATURES)
ALL	8	124	64	22
subtype1	6	16	47	3
subtype2	1	22	6	4
subtype3	1	86	11	15

Figure S10. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

'METHLYATION CNMF' versus 'RADIATIONS.RADIATION.REGIMENINDICATION'

P value = 0.06 (Fisher's exact test)

Table S13. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

nPatients	NO	YES
ALL	14	204
subtype1	1	71
subtype2	2	31
subtype3	11	102

Figure S11. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

'METHLYATION CNMF' versus 'RADIATIONEXPOSURE'

P value = 1 (Fisher's exact test)

Table S14. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

nPatients	NO	YES
ALL	180	9
subtype1	57	3
subtype2	28	1
subtype3	95	5

Figure S12. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

Clustering Approach #3: 'RPPA CNMF subtypes'

Table S15. Get Full Table Description of clustering approach #3: 'RPPA CNMF subtypes'

Cluster Labels	1	2	3
Number of samples	43	55	57

'RPPA CNMF subtypes' versus 'Time to Death'

P value = 100 (logrank test)

Table S16. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	155	1	0.1 - 66.1 (8.2)
subtype1	43	0	0.3 - 50.5 (8.1)
subtype2	55	0	0.2 - 65.9 (9.3)
subtype3	57	1	0.1 - 66.1 (7.7)

Figure S13. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

'RPPA CNMF subtypes' versus 'AGE'

P value = 0.000166 (ANOVA)

Table S17. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	155	46.6 (16.2)
subtype1	43	50.8 (13.7)
subtype2	55	50.5 (16.2)
subtype3	57	39.6 (15.7)

Figure S14. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #2: 'AGE'

'RPPA CNMF subtypes' versus 'GENDER'

P value = 0.906 (Fisher's exact test)

Table S18. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #3: 'GENDER'

nPatients	FEMALE	MALE
ALL	106	49
subtype1	29	14
subtype2	39	16
subtype3	38	19

Figure S15. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #3: 'GENDER'

'RPPA CNMF subtypes' versus 'HISTOLOGICAL.TYPE'

P value = 2.64e-06 (Chi-square test)

Table S19. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

nPatients	OTHER	THYROID PAPILLARY CARCINOMA - CLASSICAL/USUAL	THYROID PAPILLARY CARCINOMA - FOLLICULAR (>= 99% FOLLICULAR PATTERNED)	THYROID PAPILLARY CARCINOMA - TALL CELL (>= 50% TALL CELL FEATURES)
ALL	7	83	52	13
subtype1	2	11	28	2
subtype2	4	31	11	9
subtype3	1	41	13	2

Figure S16. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

'RPPA CNMF subtypes' versus 'RADIATIONS.RADIATION.REGIMENINDICATION'

P value = 0.00958 (Fisher's exact test)

Table S20. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

nPatients	NO	YES
ALL	13	142
subtype1	0	43
subtype2	9	46
subtype3	4	53

Figure S17. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

'RPPA CNMF subtypes' versus 'RADIATIONEXPOSURE'

P value = 0.791 (Fisher's exact test)

Table S21. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

nPatients	NO	YES
ALL	130	7
subtype1	38	1
subtype2	46	3
subtype3	46	3

Figure S18. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

Clustering Approach #4: 'RPPA cHierClus subtypes'

Table S22. Get Full Table Description of clustering approach #4: 'RPPA cHierClus subtypes'

Cluster Labels	1	2	3
Number of samples	23	66	66

'RPPA cHierClus subtypes' versus 'Time to Death'

P value = 100 (logrank test)

Table S23. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	155	1	0.1 - 66.1 (8.2)
subtype1	23	0	1.1 - 65.9 (8.1)
subtype2	66	0	0.1 - 66.1 (9.2)
subtype3	66	1	0.3 - 65.9 (8.0)

Figure S19. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

'RPPA cHierClus subtypes' versus 'AGE'

P value = 0.001 (ANOVA)

Table S24. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	155	46.6 (16.2)
subtype1	23	53.6 (16.7)
subtype2	66	41.3 (15.0)
subtype3	66	49.4 (15.7)

Figure S20. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #2: 'AGE'

'RPPA cHierClus subtypes' versus 'GENDER'

P value = 0.39 (Fisher's exact test)

Table S25. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #3: 'GENDER'

nPatients	FEMALE	MALE
ALL	106	49
subtype1	17	6
subtype2	48	18
subtype3	41	25

Figure S21. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #3: 'GENDER'

'RPPA cHierClus subtypes' versus 'HISTOLOGICAL.TYPE'

P value = 1.74e-11 (Chi-square test)

Table S26. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

nPatients	OTHER	THYROID PAPILLARY CARCINOMA - CLASSICAL/USUAL	THYROID PAPILLARY CARCINOMA - FOLLICULAR (>= 99% FOLLICULAR PATTERNED)	THYROID PAPILLARY CARCINOMA - TALL CELL (>= 50% TALL CELL FEATURES)
ALL	7	83	52	13
subtype1	5	7	10	1
subtype2	1	53	4	8
subtype3	1	23	38	4

Figure S22. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

'RPPA cHierClus subtypes' versus 'RADIATIONS.RADIATION.REGIMENINDICATION'

P value = 0.331 (Fisher's exact test)

Table S27. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

nPatients	NO	YES
ALL	13	142
subtype1	2	21
subtype2	8	58
subtype3	3	63

Figure S23. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

'RPPA cHierClus subtypes' versus 'RADIATIONEXPOSURE'

P value = 0.0769 (Fisher's exact test)

Table S28. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

nPatients	NO	YES
ALL	130	7
subtype1	19	3
subtype2	52	3
subtype3	59	1

Figure S24. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

Clustering Approach #5: 'RNAseq CNMF subtypes'

Table S29. Get Full Table Description of clustering approach #5: 'RNAseq CNMF subtypes'

Cluster Labels	1	2	3	4
Number of samples	64	22	47	65

'RNAseq CNMF subtypes' versus 'Time to Death'

P value = 100 (logrank test)

Table S30. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	198	1	0.0 - 66.1 (8.1)
subtype1	64	1	0.0 - 65.9 (6.9)
subtype2	22	0	0.1 - 66.1 (5.9)
subtype3	47	0	0.2 - 66.1 (10.3)
subtype4	65	0	0.2 - 65.9 (10.0)

Figure S25. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

'RNAseq CNMF subtypes' versus 'AGE'

P value = 0.0108 (ANOVA)

Table S31. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	198	46.7 (16.1)
subtype1	64	51.5 (16.3)
subtype2	22	44.2 (14.1)
subtype3	47	41.5 (13.6)
subtype4	65	46.7 (17.0)

Figure S26. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #2: 'AGE'

'RNAseq CNMF subtypes' versus 'GENDER'

P value = 0.734 (Fisher's exact test)

Table S32. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #3: 'GENDER'

nPatients	FEMALE	MALE
ALL	146	52
subtype1	45	19
subtype2	18	4
subtype3	36	11
subtype4	47	18

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

'RNAseq CNMF subtypes' versus 'HISTOLOGICAL.TYPE'

P value = 8.86e-17 (Chi-square test)

Table S33. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

nPatients	OTHER	THYROID PAPILLARY CARCINOMA - CLASSICAL/USUAL	THYROID PAPILLARY CARCINOMA - FOLLICULAR (>= 99% FOLLICULAR PATTERNED)	THYROID PAPILLARY CARCINOMA - TALL CELL (>= 50% TALL CELL FEATURES)
ALL	7	113	58	20
subtype1	6	14	43	1
subtype2	0	14	4	4
subtype3	1	37	8	1
subtype4	0	48	3	14

Figure S28. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

'RNAseq CNMF subtypes' versus 'RADIATIONS.RADIATION.REGIMENINDICATION'

P value = 0.00383 (Fisher's exact test)

Table S34. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

nPatients	NO	YES
ALL	12	186
subtype1	1	63
subtype2	0	22
subtype3	1	46
subtype4	10	55

Figure S29. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

'RNAseq CNMF subtypes' versus 'RADIATIONEXPOSURE'

P value = 0.664 (Fisher's exact test)

Table S35. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

nPatients	NO	YES
ALL	162	8
subtype1	52	3
subtype2	18	0
subtype3	37	3
subtype4	55	2

Figure S30. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

Clustering Approach #6: 'RNAseq cHierClus subtypes'

Table S36. Get Full Table Description of clustering approach #6: 'RNAseq cHierClus subtypes'

Cluster Labels	1	2	3
Number of samples	89	34	75

'RNAseq cHierClus subtypes' versus 'Time to Death'

P value = 100 (logrank test)

Table S37. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	198	1	0.0 - 66.1 (8.1)
subtype1	89	0	0.2 - 66.1 (9.3)
subtype2	34	0	0.2 - 66.1 (10.1)
subtype3	75	1	0.0 - 65.9 (6.8)

Figure S31. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

'RNAseq cHierClus subtypes' versus 'AGE'

P value = 0.0163 (ANOVA)

Table S38. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	198	46.7 (16.1)
subtype1	89	46.1 (16.1)
subtype2	34	40.8 (15.1)
subtype3	75	50.2 (15.7)

Figure S32. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #2: 'AGE'

'RNAseq cHierClus subtypes' versus 'GENDER'

P value = 0.453 (Fisher's exact test)

Table S39. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #3: 'GENDER'

nPatients	FEMALE	MALE
ALL	146	52
subtype1	65	24
subtype2	28	6
subtype3	53	22

Figure S33. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #3: 'GENDER'

'RNAseq cHierClus subtypes' versus 'HISTOLOGICAL.TYPE'

P value = 6.92e-16 (Chi-square test)

Table S40. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

nPatients	OTHER	THYROID PAPILLARY CARCINOMA - CLASSICAL/USUAL	THYROID PAPILLARY CARCINOMA - FOLLICULAR (>= 99% FOLLICULAR PATTERNED)	THYROID PAPILLARY CARCINOMA - TALL CELL (>= 50% TALL CELL FEATURES)
ALL	7	113	58	20
subtype1	0	64	8	17
subtype2	1	29	4	0
subtype3	6	20	46	3

Figure S34. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

'RNAseq cHierClus subtypes' versus 'RADIATIONS.RADIATION.REGIMENINDICATION'

P value = 0.0217 (Fisher's exact test)

Table S41. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

nPatients	NO	YES
ALL	12	186
subtype1	10	79
subtype2	1	33
subtype3	1	74

Figure S35. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

'RNAseq cHierClus subtypes' versus 'RADIATIONEXPOSURE'

P value = 1 (Fisher's exact test)

Table S42. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

nPatients	NO	YES
ALL	162	8
subtype1	73	4
subtype2	29	1
subtype3	60	3

Figure S36. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

Clustering Approach #7: 'MIRseq CNMF subtypes'

Table S43. Get Full Table Description of clustering approach #7: 'MIRseq CNMF subtypes'

Cluster Labels	1	2	3
Number of samples	70	84	56

'MIRseq CNMF subtypes' versus 'Time to Death'

P value = 100 (logrank test)

Table S44. Clustering Approach #7: 'MIRseq CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	210	1	0.0 - 66.1 (8.2)
subtype1	70	1	0.3 - 65.9 (7.2)
subtype2	84	0	0.2 - 66.1 (10.1)
subtype3	56	0	0.0 - 66.1 (8.0)

Figure S37. Get High-res Image Clustering Approach #7: 'MIRseq CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

'MIRseq CNMF subtypes' versus 'AGE'

P value = 0.0556 (ANOVA)

Table S45. Clustering Approach #7: 'MIRseq CNMF subtypes' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	210	46.6 (16.0)
subtype1	70	49.6 (16.9)
subtype2	84	43.5 (15.1)
subtype3	56	47.6 (15.4)

Figure S38. Get High-res Image Clustering Approach #7: 'MIRseq CNMF subtypes' versus Clinical Feature #2: 'AGE'

'MIRseq CNMF subtypes' versus 'GENDER'

P value = 0.335 (Fisher's exact test)

Table S46. Clustering Approach #7: 'MIRseq CNMF subtypes' versus Clinical Feature #3: 'GENDER'

nPatients	FEMALE	MALE
ALL	155	55
subtype1	48	22
subtype2	62	22
subtype3	45	11

Figure S39. Get High-res Image Clustering Approach #7: 'MIRseq CNMF subtypes' versus Clinical Feature #3: 'GENDER'

'MIRseq CNMF subtypes' versus 'HISTOLOGICAL.TYPE'

P value = 6.15e-18 (Chi-square test)

Table S47. Clustering Approach #7: 'MIRseq CNMF subtypes' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

nPatients	OTHER	THYROID PAPILLARY CARCINOMA - CLASSICAL/USUAL	THYROID PAPILLARY CARCINOMA - FOLLICULAR (>= 99% FOLLICULAR PATTERNED)	THYROID PAPILLARY CARCINOMA - TALL CELL (>= 50% TALL CELL FEATURES)
ALL	8	116	64	22
subtype1	6	15	48	1
subtype2	2	64	10	8
subtype3	0	37	6	13

Figure S40. Get High-res Image Clustering Approach #7: 'MIRseq CNMF subtypes' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

'MIRseq CNMF subtypes' versus 'RADIATIONS.RADIATION.REGIMENINDICATION'

P value = 0.468 (Fisher's exact test)

Table S48. Clustering Approach #7: 'MIRseq CNMF subtypes' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

nPatients	NO	YES
ALL	14	196
subtype1	3	67
subtype2	8	76
subtype3	3	53

Figure S41. Get High-res Image Clustering Approach #7: 'MIRseq CNMF subtypes' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

'MIRseq CNMF subtypes' versus 'RADIATIONEXPOSURE'

P value = 1 (Fisher's exact test)

Table S49. Clustering Approach #7: 'MIRseq CNMF subtypes' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

nPatients	NO	YES
ALL	175	9
subtype1	59	3
subtype2	67	4
subtype3	49	2

Figure S42. Get High-res Image Clustering Approach #7: 'MIRseq CNMF subtypes' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

Clustering Approach #8: 'MIRseq cHierClus subtypes'

Table S50. Get Full Table Description of clustering approach #8: 'MIRseq cHierClus subtypes'

Cluster Labels	1	2	3
Number of samples	62	63	85

'MIRseq cHierClus subtypes' versus 'Time to Death'

P value = 100 (logrank test)

Table S51. Clustering Approach #8: 'MIRseq cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	210	1	0.0 - 66.1 (8.2)
subtype1	62	0	0.0 - 66.1 (8.0)
subtype2	63	1	0.3 - 65.9 (7.1)
subtype3	85	0	0.2 - 66.1 (10.0)

Figure S43. Get High-res Image Clustering Approach #8: 'MIRseq cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

'MIRseq cHierClus subtypes' versus 'AGE'

P value = 0.00403 (ANOVA)

Table S52. Clustering Approach #8: 'MIRseq cHierClus subtypes' versus Clinical Feature #2: 'AGE'

	nPatients	Mean (Std.Dev)
ALL	210	46.6 (16.0)
subtype1	62	48.4 (16.2)
subtype2	63	50.6 (16.2)
subtype3	85	42.3 (14.6)

Figure S44. Get High-res Image Clustering Approach #8: 'MIRseq cHierClus subtypes' versus Clinical Feature #2: 'AGE'

'MIRseq cHierClus subtypes' versus 'GENDER'

P value = 0.365 (Fisher's exact test)

Table S53. Clustering Approach #8: 'MIRseq cHierClus subtypes' versus Clinical Feature #3: 'GENDER'

nPatients	FEMALE	MALE
ALL	155	55
subtype1	50	12
subtype2	45	18
subtype3	60	25

Figure S45. Get High-res Image Clustering Approach #8: 'MIRseq cHierClus subtypes' versus Clinical Feature #3: 'GENDER'

'MIRseq cHierClus subtypes' versus 'HISTOLOGICAL.TYPE'

P value = 6.25e-20 (Chi-square test)

Table S54. Clustering Approach #8: 'MIRseq cHierClus subtypes' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

nPatients	OTHER	THYROID PAPILLARY CARCINOMA - CLASSICAL/USUAL	THYROID PAPILLARY CARCINOMA - FOLLICULAR (>= 99% FOLLICULAR PATTERNED)	THYROID PAPILLARY CARCINOMA - TALL CELL (>= 50% TALL CELL FEATURES)
ALL	8	116	64	22
subtype1	0	39	7	16
subtype2	6	11	45	1
subtype3	2	66	12	5

Figure S46. Get High-res Image Clustering Approach #8: 'MIRseq cHierClus subtypes' versus Clinical Feature #4: 'HISTOLOGICAL.TYPE'

'MIRseq cHierClus subtypes' versus 'RADIATIONS.RADIATION.REGIMENINDICATION'

P value = 0.0388 (Fisher's exact test)

Table S55. Clustering Approach #8: 'MIRseq cHierClus subtypes' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

nPatients	NO	YES
ALL	14	196
subtype1	3	59
subtype2	1	62
subtype3	10	75

Figure S47. Get High-res Image Clustering Approach #8: 'MIRseq cHierClus subtypes' versus Clinical Feature #5: 'RADIATIONS.RADIATION.REGIMENINDICATION'

'MIRseq cHierClus subtypes' versus 'RADIATIONEXPOSURE'

P value = 0.834 (Fisher's exact test)

Table S56. Clustering Approach #8: 'MIRseq cHierClus subtypes' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

nPatients	NO	YES
ALL	175	9
subtype1	56	2
subtype2	52	3
subtype3	67	4

Figure S48. Get High-res Image Clustering Approach #8: 'MIRseq cHierClus subtypes' versus Clinical Feature #6: 'RADIATIONEXPOSURE'

Methods & Data

Input

Cluster data file = THCA-TP.mergedcluster.txt
Clinical data file = THCA-TP.clin.merged.picked.txt
Number of patients = 218
Number of clustering approaches = 8
Number of selected clinical features = 6
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

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

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

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] 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)

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

Made with Nozzle