Correlation between aggregated 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 10 different clustering approaches and 14 clinical features across 528 patients, 43 significant findings 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 'YEARS_TO_BIRTH', 'GENDER', and 'YEAR_OF_TOBACCO_SMOKING_ONSET'.
7 subtypes identified in current cancer cohort by 'METHLYATION CNMF'. These subtypes correlate to 'Time to Death', 'YEARS_TO_BIRTH', 'PATHOLOGY_T_STAGE', 'PATHOLOGY_N_STAGE', 'GENDER', 'HISTOLOGICAL_TYPE', 'NUMBER_PACK_YEARS_SMOKED', 'NUMBER_OF_LYMPH_NODES', and 'RACE'.
CNMF clustering analysis on RPPA data identified 5 subtypes that correlate to 'GENDER'.
Consensus hierarchical clustering analysis on RPPA data identified 3 subtypes that correlate to 'PATHOLOGIC_STAGE', 'PATHOLOGY_N_STAGE', and 'NUMBER_OF_LYMPH_NODES'.
CNMF clustering analysis on sequencing-based mRNA expression data identified 4 subtypes that correlate to 'PATHOLOGY_T_STAGE', 'GENDER', 'RADIATION_THERAPY', and 'HISTOLOGICAL_TYPE'.
Consensus hierarchical clustering analysis on sequencing-based mRNA expression data identified 4 subtypes that correlate to 'Time to Death', 'YEARS_TO_BIRTH', 'PATHOLOGY_T_STAGE', 'GENDER', 'RADIATION_THERAPY', 'HISTOLOGICAL_TYPE', 'NUMBER_PACK_YEARS_SMOKED', and 'NUMBER_OF_LYMPH_NODES'.
5 subtypes identified in current cancer cohort by 'MIRSEQ CNMF'. These subtypes correlate to 'Time to Death', 'GENDER', 'RADIATION_THERAPY', 'HISTOLOGICAL_TYPE', and 'RACE'.
3 subtypes identified in current cancer cohort by 'MIRSEQ CHIERARCHICAL'. These subtypes correlate to 'Time to Death', 'PATHOLOGY_T_STAGE', 'GENDER', 'RADIATION_THERAPY', 'NUMBER_OF_LYMPH_NODES', and 'RACE'.
3 subtypes identified in current cancer cohort by 'MIRseq Mature CNMF subtypes'. These subtypes correlate to 'RADIATION_THERAPY' and 'NUMBER_PACK_YEARS_SMOKED'.
3 subtypes identified in current cancer cohort by 'MIRseq Mature cHierClus subtypes'. These subtypes correlate to 'PATHOLOGY_T_STAGE' and 'RADIATION_THERAPY'.

Results

Overview of the results

Table 1. Get Full Table Overview of the association between subtypes identified by 10 different clustering approaches and 14 clinical features. Shown in the table are P values (Q values). Thresholded by P value < 0.05 and Q value < 0.25, 43 significant findings detected.


Clinical Features	Statistical Tests	Copy Number Ratio CNMF subtypes	METHLYATION CNMF	RPPA CNMF subtypes	RPPA cHierClus subtypes	RNAseq CNMF subtypes	RNAseq cHierClus subtypes	MIRSEQ CNMF	MIRSEQ CHIERARCHICAL	MIRseq Mature CNMF subtypes	MIRseq Mature cHierClus subtypes
Time to Death	logrank test	0.19 (0.311)	0.0121 (0.0548)	0.167 (0.293)	0.144 (0.273)	0.155 (0.281)	0.00537 (0.0301)	0.00483 (0.0282)	0.028 (0.0979)	0.656 (0.747)	0.0582 (0.169)
YEARS TO BIRTH	Kruskal-Wallis (anova)	0.0182 (0.0708)	1.08e-08 (1.51e-06)	0.134 (0.265)	0.0759 (0.193)	0.595 (0.702)	0.00439 (0.0279)	0.287 (0.398)	0.0733 (0.192)	0.811 (0.881)	0.739 (0.822)
PATHOLOGIC STAGE	Fisher's exact test	0.185 (0.311)	0.523 (0.654)	0.266 (0.383)	0.029 (0.0991)	0.0742 (0.192)	0.0986 (0.226)	0.134 (0.265)	0.0591 (0.169)	0.833 (0.89)	0.0822 (0.2)
PATHOLOGY T STAGE	Fisher's exact test	0.213 (0.331)	0.00091 (0.00986)	0.13 (0.264)	0.191 (0.311)	0.0264 (0.0974)	3e-05 (7e-04)	0.122 (0.256)	0.00165 (0.0154)	0.825 (0.888)	0.0179 (0.0708)
PATHOLOGY N STAGE	Fisher's exact test	0.059 (0.169)	0.0142 (0.062)	0.406 (0.532)	2e-05 (0.00056)	0.268 (0.383)	0.123 (0.256)	0.612 (0.714)	0.0861 (0.204)	0.686 (0.774)	0.951 (0.994)
PATHOLOGY M STAGE	Fisher's exact test	0.337 (0.453)	0.561 (0.683)			1 (1.00)	0.26 (0.379)	0.273 (0.386)	0.278 (0.389)	1 (1.00)	0.633 (0.732)
GENDER	Fisher's exact test	0.0304 (0.101)	1e-05 (0.00035)	0.00047 (0.00658)	0.215 (0.331)	4e-05 (8e-04)	7e-05 (0.00122)	0.00198 (0.0163)	0.0108 (0.0521)	0.359 (0.479)	0.122 (0.256)
RADIATION THERAPY	Fisher's exact test	0.461 (0.587)	0.0564 (0.169)	0.205 (0.326)	0.757 (0.834)	0.00061 (0.00776)	0.00243 (0.0179)	0.024 (0.0907)	0.0107 (0.0521)	0.0316 (0.103)	0.00162 (0.0154)
HISTOLOGICAL TYPE	Fisher's exact test	0.0828 (0.2)	0.0174 (0.0708)			0.00234 (0.0179)	0.00025 (0.00389)	0.00467 (0.0282)	0.238 (0.359)	0.191 (0.311)	0.159 (0.282)
NUMBER PACK YEARS SMOKED	Kruskal-Wallis (anova)	0.0617 (0.173)	0.00942 (0.0507)	0.0724 (0.192)	0.586 (0.701)	0.154 (0.281)	0.00186 (0.0163)	0.583 (0.701)	0.182 (0.311)	0.000916 (0.00986)	0.556 (0.683)
YEAR OF TOBACCO SMOKING ONSET	Kruskal-Wallis (anova)	0.00311 (0.0218)	0.245 (0.361)	0.324 (0.445)	0.223 (0.34)	0.839 (0.89)	0.199 (0.32)	0.0559 (0.169)	0.0804 (0.2)	0.114 (0.256)	0.118 (0.256)
NUMBER OF LYMPH NODES	Kruskal-Wallis (anova)	0.117 (0.256)	5.56e-06 (0.000259)	0.209 (0.328)	4.76e-06 (0.000259)	0.14 (0.269)	0.0102 (0.0521)	0.157 (0.281)	0.00425 (0.0279)	0.0929 (0.217)	0.597 (0.702)
RACE	Fisher's exact test	0.149 (0.278)	0.0182 (0.0708)	0.073 (0.192)	0.428 (0.555)	0.059 (0.169)	0.126 (0.259)	0.0119 (0.0548)	0.0272 (0.0976)	0.243 (0.361)	0.181 (0.311)
ETHNICITY	Fisher's exact test	0.14 (0.269)	0.332 (0.451)	0.697 (0.781)	0.544 (0.674)	0.648 (0.744)	0.774 (0.847)	0.438 (0.562)	0.377 (0.498)	0.931 (0.98)	0.486 (0.612)

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
Number of samples	203	128	145	46

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

P value = 0.19 (logrank test), Q value = 0.31

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

	nPatients	nDeath	Duration Range (Median), Month
ALL	520	219	0.1 - 211.0 (21.1)
subtype1	202	95	0.5 - 180.2 (19.1)
subtype2	128	50	0.1 - 153.9 (24.9)
subtype3	144	54	0.1 - 211.0 (22.2)
subtype4	46	20	3.1 - 140.8 (19.0)

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 'YEARS_TO_BIRTH'

P value = 0.0182 (Kruskal-Wallis (anova)), Q value = 0.071

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

	nPatients	Mean (Std.Dev)
ALL	521	60.9 (11.9)
subtype1	203	59.3 (12.3)
subtype2	128	60.2 (10.5)
subtype3	144	62.6 (12.1)
subtype4	46	64.5 (12.0)

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

'Copy Number Ratio CNMF subtypes' versus 'PATHOLOGIC_STAGE'

P value = 0.185 (Fisher's exact test), Q value = 0.31

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

nPatients	STAGE I	STAGE II	STAGE III	STAGE IVA	STAGE IVB	STAGE IVC
ALL	27	76	82	252	12	1
subtype1	12	27	28	106	9	0
subtype2	4	16	16	65	2	1
subtype3	9	25	25	62	1	0
subtype4	2	8	13	19	0	0

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

'Copy Number Ratio CNMF subtypes' versus 'PATHOLOGY_T_STAGE'

P value = 0.213 (Fisher's exact test), Q value = 0.33

Table S5. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

nPatients	T0+T1	T2	T3	T4
ALL	50	137	100	173
subtype1	19	53	39	74
subtype2	10	28	29	42
subtype3	18	40	19	47
subtype4	3	16	13	10

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 = 0.059 (Fisher's exact test), Q value = 0.17

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

nPatients	N0	N1	N2	N3
ALL	179	68	168	8
subtype1	65	25	79	5
subtype2	40	14	42	2
subtype3	61	18	33	1
subtype4	13	11	14	0

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 = 0.337 (Fisher's exact test), Q value = 0.45

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

nPatients	0	1
ALL	190	1
subtype1	68	0
subtype2	41	1
subtype3	59	0
subtype4	22	0

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 = 0.0304 (Fisher's exact test), Q value = 0.1

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

nPatients	FEMALE	MALE
ALL	140	382
subtype1	49	154
subtype2	27	101
subtype3	45	100
subtype4	19	27

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 'RADIATION_THERAPY'

P value = 0.461 (Fisher's exact test), Q value = 0.59

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

nPatients	NO	YES
ALL	162	301
subtype1	63	114
subtype2	32	78
subtype3	49	82
subtype4	18	27

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

'Copy Number Ratio CNMF subtypes' versus 'HISTOLOGICAL_TYPE'

P value = 0.0828 (Fisher's exact test), Q value = 0.2

Table S10. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

nPatients	HEAD & NECK SQUAMOUS CELL CARCINOMA	HEAD & NECK SQUAMOUS CELL CARCINOMA BASALOID TYPE	HEAD & NECK SQUAMOUS CELL CARCINOMA, SPINDLE CELL VARIANT
ALL	511	10	1
subtype1	202	1	0
subtype2	122	5	1
subtype3	141	4	0
subtype4	46	0	0

Figure S9. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

'Copy Number Ratio CNMF subtypes' versus 'NUMBER_PACK_YEARS_SMOKED'

P value = 0.0617 (Kruskal-Wallis (anova)), Q value = 0.17

Table S11. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

	nPatients	Mean (Std.Dev)
ALL	293	45.9 (35.3)
subtype1	111	47.5 (34.6)
subtype2	82	49.8 (34.3)
subtype3	76	42.7 (39.7)
subtype4	24	35.2 (25.1)

Figure S10. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

'Copy Number Ratio CNMF subtypes' versus 'YEAR_OF_TOBACCO_SMOKING_ONSET'

P value = 0.00311 (Kruskal-Wallis (anova)), Q value = 0.022

Table S12. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

	nPatients	Mean (Std.Dev)
ALL	277	1967.4 (12.8)
subtype1	105	1969.6 (12.8)
subtype2	73	1964.6 (11.6)
subtype3	73	1965.3 (13.4)
subtype4	26	1972.1 (12.3)

Figure S11. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

'Copy Number Ratio CNMF subtypes' versus 'NUMBER_OF_LYMPH_NODES'

P value = 0.117 (Kruskal-Wallis (anova)), Q value = 0.26

Table S13. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

	nPatients	Mean (Std.Dev)
ALL	408	2.2 (4.3)
subtype1	171	2.6 (4.9)
subtype2	97	2.3 (5.1)
subtype3	101	1.6 (2.6)
subtype4	39	1.5 (2.0)

Figure S12. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

'Copy Number Ratio CNMF subtypes' versus 'RACE'

P value = 0.149 (Fisher's exact test), Q value = 0.28

Table S14. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #13: 'RACE'

nPatients	AMERICAN INDIAN OR ALASKA NATIVE	ASIAN	BLACK OR AFRICAN AMERICAN	WHITE
ALL	2	11	48	446
subtype1	1	6	24	164
subtype2	0	0	14	112
subtype3	1	4	8	129
subtype4	0	1	2	41

Figure S13. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #13: 'RACE'

'Copy Number Ratio CNMF subtypes' versus 'ETHNICITY'

P value = 0.14 (Fisher's exact test), Q value = 0.27

Table S15. Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #14: 'ETHNICITY'

nPatients	HISPANIC OR LATINO	NOT HISPANIC OR LATINO
ALL	26	459
subtype1	14	176
subtype2	3	113
subtype3	5	130
subtype4	4	40

Figure S14. Get High-res Image Clustering Approach #1: 'Copy Number Ratio CNMF subtypes' versus Clinical Feature #14: 'ETHNICITY'

Clustering Approach #2: 'METHLYATION CNMF'

Table S16. Description of clustering approach #2: 'METHLYATION CNMF'

Cluster Labels	1	2	3	4	5	6	7
Number of samples	87	87	104	91	78	75	6

'METHLYATION CNMF' versus 'Time to Death'

P value = 0.0121 (logrank test), Q value = 0.055

Table S17. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	526	223	0.1 - 211.0 (21.2)
subtype1	85	37	2.3 - 211.0 (18.5)
subtype2	87	35	0.1 - 153.9 (24.4)
subtype3	104	49	0.4 - 169.4 (20.9)
subtype4	91	47	0.4 - 139.4 (20.6)
subtype5	78	19	0.1 - 105.9 (23.7)
subtype6	75	31	0.5 - 180.2 (19.0)
subtype7	6	5	11.8 - 35.8 (16.5)

Figure S15. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #1: 'Time to Death'

'METHLYATION CNMF' versus 'YEARS_TO_BIRTH'

P value = 1.08e-08 (Kruskal-Wallis (anova)), Q value = 1.5e-06

Table S18. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

	nPatients	Mean (Std.Dev)
ALL	527	60.9 (11.9)
subtype1	86	59.8 (12.7)
subtype2	87	58.7 (10.8)
subtype3	104	66.2 (11.6)
subtype4	91	64.3 (10.3)
subtype5	78	57.6 (9.6)
subtype6	75	56.4 (13.4)
subtype7	6	63.7 (9.7)

Figure S16. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

'METHLYATION CNMF' versus 'PATHOLOGIC_STAGE'

P value = 0.523 (Fisher's exact test), Q value = 0.65

Table S19. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

nPatients	STAGE I	STAGE II	STAGE III	STAGE IVA	STAGE IVB	STAGE IVC
ALL	27	77	82	257	12	1
subtype1	10	12	13	47	0	0
subtype2	3	16	13	42	3	1
subtype3	4	20	17	52	4	0
subtype4	2	8	13	52	3	0
subtype5	2	9	11	25	1	0
subtype6	5	12	14	36	1	0
subtype7	1	0	1	3	0	0

Figure S17. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

'METHLYATION CNMF' versus 'PATHOLOGY_T_STAGE'

P value = 0.00091 (Fisher's exact test), Q value = 0.0099

Table S20. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

nPatients	T0+T1	T2	T3	T4
ALL	50	140	101	175
subtype1	16	25	14	28
subtype2	9	20	16	34
subtype3	5	27	21	45
subtype4	4	19	19	38
subtype5	10	23	11	8
subtype6	5	26	17	20
subtype7	1	0	3	2

Figure S18. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

'METHLYATION CNMF' versus 'PATHOLOGY_N_STAGE'

P value = 0.0142 (Fisher's exact test), Q value = 0.062

Table S21. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

nPatients	N0	N1	N2	N3
ALL	180	68	172	8
subtype1	33	14	32	0
subtype2	41	12	21	1
subtype3	44	16	23	3
subtype4	23	8	39	3
subtype5	14	9	23	1
subtype6	24	9	30	0
subtype7	1	0	4	0

Figure S19. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

'METHLYATION CNMF' versus 'PATHOLOGY_M_STAGE'

P value = 0.561 (Fisher's exact test), Q value = 0.68

Table S22. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

nPatients	0	1
ALL	191	1
subtype1	44	0
subtype2	35	1
subtype3	40	0
subtype4	20	0
subtype5	21	0
subtype6	28	0
subtype7	3	0

Figure S20. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

'METHLYATION CNMF' versus 'GENDER'

P value = 1e-05 (Fisher's exact test), Q value = 0.00035

Table S23. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	142	386
subtype1	30	57
subtype2	14	73
subtype3	44	60
subtype4	17	74
subtype5	9	69
subtype6	27	48
subtype7	1	5

Figure S21. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #7: 'GENDER'

'METHLYATION CNMF' versus 'RADIATION_THERAPY'

P value = 0.0564 (Fisher's exact test), Q value = 0.17

Table S24. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #8: 'RADIATION_THERAPY'

nPatients	NO	YES
ALL	163	303
subtype1	35	45
subtype2	27	52
subtype3	37	58
subtype4	20	49
subtype5	16	55
subtype6	24	42
subtype7	4	2

Figure S22. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #8: 'RADIATION_THERAPY'

'METHLYATION CNMF' versus 'HISTOLOGICAL_TYPE'

P value = 0.0174 (Fisher's exact test), Q value = 0.071

Table S25. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

nPatients	HEAD & NECK SQUAMOUS CELL CARCINOMA	HEAD & NECK SQUAMOUS CELL CARCINOMA BASALOID TYPE	HEAD & NECK SQUAMOUS CELL CARCINOMA, SPINDLE CELL VARIANT
ALL	517	10	1
subtype1	87	0	0
subtype2	84	3	0
subtype3	104	0	0
subtype4	89	2	0
subtype5	73	5	0
subtype6	74	0	1
subtype7	6	0	0

Figure S23. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

'METHLYATION CNMF' versus 'NUMBER_PACK_YEARS_SMOKED'

P value = 0.00942 (Kruskal-Wallis (anova)), Q value = 0.051

Table S26. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

	nPatients	Mean (Std.Dev)
ALL	298	45.8 (35.2)
subtype1	53	48.0 (44.0)
subtype2	57	49.5 (26.4)
subtype3	53	42.5 (41.9)
subtype4	63	51.1 (30.1)
subtype5	42	39.8 (37.0)
subtype6	28	37.3 (25.8)
subtype7	2	44.0 (5.7)

Figure S24. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

'METHLYATION CNMF' versus 'YEAR_OF_TOBACCO_SMOKING_ONSET'

P value = 0.245 (Kruskal-Wallis (anova)), Q value = 0.36

Table S27. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

	nPatients	Mean (Std.Dev)
ALL	282	1967.3 (12.7)
subtype1	52	1968.0 (13.6)
subtype2	51	1965.9 (11.4)
subtype3	50	1964.8 (12.8)
subtype4	59	1967.1 (12.7)
subtype5	40	1968.8 (12.7)
subtype6	28	1972.1 (13.1)
subtype7	2	1955.5 (6.4)

Figure S25. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

'METHLYATION CNMF' versus 'NUMBER_OF_LYMPH_NODES'

P value = 5.56e-06 (Kruskal-Wallis (anova)), Q value = 0.00026

Table S28. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

	nPatients	Mean (Std.Dev)
ALL	413	2.2 (4.3)
subtype1	71	2.3 (3.3)
subtype2	75	1.0 (2.1)
subtype3	87	1.4 (2.2)
subtype4	71	4.5 (8.3)
subtype5	40	2.2 (2.8)
subtype6	64	1.7 (2.1)
subtype7	5	4.0 (3.5)

Figure S26. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

'METHLYATION CNMF' versus 'RACE'

P value = 0.0182 (Fisher's exact test), Q value = 0.071

Table S29. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #13: 'RACE'

nPatients	AMERICAN INDIAN OR ALASKA NATIVE	ASIAN	BLACK OR AFRICAN AMERICAN	WHITE
ALL	2	11	48	452
subtype1	1	3	5	73
subtype2	0	0	13	70
subtype3	0	4	9	90
subtype4	0	1	15	74
subtype5	0	0	3	75
subtype6	1	3	3	65
subtype7	0	0	0	5

Figure S27. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #13: 'RACE'

'METHLYATION CNMF' versus 'ETHNICITY'

P value = 0.332 (Fisher's exact test), Q value = 0.45

Table S30. Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #14: 'ETHNICITY'

nPatients	HISPANIC OR LATINO	NOT HISPANIC OR LATINO
ALL	26	465
subtype1	3	78
subtype2	2	74
subtype3	4	96
subtype4	5	79
subtype5	4	69
subtype6	7	64
subtype7	1	5

Figure S28. Get High-res Image Clustering Approach #2: 'METHLYATION CNMF' versus Clinical Feature #14: 'ETHNICITY'

Clustering Approach #3: 'RPPA CNMF subtypes'

Table S31. Description of clustering approach #3: 'RPPA CNMF subtypes'

Cluster Labels	1	2	3	4	5
Number of samples	51	30	33	29	69

'RPPA CNMF subtypes' versus 'Time to Death'

P value = 0.167 (logrank test), Q value = 0.29

Table S32. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	212	128	0.1 - 211.0 (23.1)
subtype1	51	33	2.5 - 172.7 (19.0)
subtype2	30	21	3.5 - 139.4 (22.9)
subtype3	33	16	3.5 - 129.2 (35.9)
subtype4	29	14	4.0 - 156.5 (26.4)
subtype5	69	44	0.1 - 211.0 (19.8)

Figure S29. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

'RPPA CNMF subtypes' versus 'YEARS_TO_BIRTH'

P value = 0.134 (Kruskal-Wallis (anova)), Q value = 0.26

Table S33. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

	nPatients	Mean (Std.Dev)
ALL	212	62.1 (12.2)
subtype1	51	59.8 (12.6)
subtype2	30	61.2 (9.1)
subtype3	33	63.8 (9.8)
subtype4	29	64.1 (14.4)
subtype5	69	62.6 (13.2)

Figure S30. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

'RPPA CNMF subtypes' versus 'PATHOLOGIC_STAGE'

P value = 0.266 (Fisher's exact test), Q value = 0.38

Table S34. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

nPatients	STAGE I	STAGE II	STAGE III	STAGE IVA	STAGE IVB
ALL	9	39	31	117	4
subtype1	1	16	7	25	1
subtype2	0	2	3	19	0
subtype3	2	3	8	18	1
subtype4	2	7	5	13	0
subtype5	4	11	8	42	2

Figure S31. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

'RPPA CNMF subtypes' versus 'PATHOLOGY_T_STAGE'

P value = 0.13 (Fisher's exact test), Q value = 0.26

Table S35. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

nPatients	T0+T1	T2	T3	T4
ALL	13	59	53	79
subtype1	2	20	11	17
subtype2	1	4	7	13
subtype3	2	4	13	13
subtype4	2	12	3	11
subtype5	6	19	19	25

Figure S32. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

'RPPA CNMF subtypes' versus 'PATHOLOGY_N_STAGE'

P value = 0.406 (Fisher's exact test), Q value = 0.53

Table S36. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

nPatients	N0	N1	N2	N3
ALL	72	21	79	4
subtype1	21	3	16	1
subtype2	6	2	15	0
subtype3	12	4	13	1
subtype4	14	4	7	0
subtype5	19	8	28	2

Figure S33. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

'RPPA CNMF subtypes' versus 'GENDER'

P value = 0.00047 (Fisher's exact test), Q value = 0.0066

Table S37. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	62	150
subtype1	11	40
subtype2	7	23
subtype3	2	31
subtype4	13	16
subtype5	29	40

Figure S34. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #7: 'GENDER'

'RPPA CNMF subtypes' versus 'RADIATION_THERAPY'

P value = 0.205 (Fisher's exact test), Q value = 0.33

Table S38. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #8: 'RADIATION_THERAPY'

nPatients	NO	YES
ALL	73	87
subtype1	18	24
subtype2	7	18
subtype3	9	13
subtype4	13	10
subtype5	26	22

Figure S35. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #8: 'RADIATION_THERAPY'

'RPPA CNMF subtypes' versus 'NUMBER_PACK_YEARS_SMOKED'

P value = 0.0724 (Kruskal-Wallis (anova)), Q value = 0.19

Table S39. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

	nPatients	Mean (Std.Dev)
ALL	110	49.5 (37.5)
subtype1	24	47.6 (28.3)
subtype2	15	43.3 (22.7)
subtype3	22	61.6 (32.0)
subtype4	16	51.5 (70.0)
subtype5	33	44.6 (29.7)

Figure S36. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

'RPPA CNMF subtypes' versus 'YEAR_OF_TOBACCO_SMOKING_ONSET'

P value = 0.324 (Kruskal-Wallis (anova)), Q value = 0.45

Table S40. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

	nPatients	Mean (Std.Dev)
ALL	117	1964.9 (12.3)
subtype1	28	1968.0 (11.8)
subtype2	16	1966.6 (10.1)
subtype3	20	1961.7 (9.6)
subtype4	17	1964.5 (17.5)
subtype5	36	1963.6 (12.1)

Figure S37. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

'RPPA CNMF subtypes' versus 'NUMBER_OF_LYMPH_NODES'

P value = 0.209 (Kruskal-Wallis (anova)), Q value = 0.33

Table S41. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

	nPatients	Mean (Std.Dev)
ALL	170	2.9 (5.1)
subtype1	39	2.7 (4.3)
subtype2	24	3.6 (4.6)
subtype3	29	2.4 (5.6)
subtype4	24	1.4 (1.8)
subtype5	54	3.6 (6.4)

Figure S38. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

'RPPA CNMF subtypes' versus 'RACE'

P value = 0.073 (Fisher's exact test), Q value = 0.19

Table S42. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #13: 'RACE'

nPatients	AMERICAN INDIAN OR ALASKA NATIVE	ASIAN	BLACK OR AFRICAN AMERICAN	WHITE
ALL	1	3	19	185
subtype1	1	0	7	42
subtype2	0	1	4	25
subtype3	0	1	5	26
subtype4	0	0	0	29
subtype5	0	1	3	63

Figure S39. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #13: 'RACE'

'RPPA CNMF subtypes' versus 'ETHNICITY'

P value = 0.697 (Fisher's exact test), Q value = 0.78

Table S43. Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #14: 'ETHNICITY'

nPatients	HISPANIC OR LATINO	NOT HISPANIC OR LATINO
ALL	13	190
subtype1	4	44
subtype2	3	25
subtype3	1	30
subtype4	2	27
subtype5	3	64

Figure S40. Get High-res Image Clustering Approach #3: 'RPPA CNMF subtypes' versus Clinical Feature #14: 'ETHNICITY'

Clustering Approach #4: 'RPPA cHierClus subtypes'

Table S44. Description of clustering approach #4: 'RPPA cHierClus subtypes'

Cluster Labels	1	2	3
Number of samples	78	85	49

'RPPA cHierClus subtypes' versus 'Time to Death'

P value = 0.144 (logrank test), Q value = 0.27

Table S45. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	212	128	0.1 - 211.0 (23.1)
subtype1	78	46	2.5 - 172.7 (22.6)
subtype2	85	59	0.1 - 211.0 (21.5)
subtype3	49	23	2.1 - 111.2 (31.4)

Figure S41. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

'RPPA cHierClus subtypes' versus 'YEARS_TO_BIRTH'

P value = 0.0759 (Kruskal-Wallis (anova)), Q value = 0.19

Table S46. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

	nPatients	Mean (Std.Dev)
ALL	212	62.1 (12.2)
subtype1	78	61.2 (13.2)
subtype2	85	64.3 (11.8)
subtype3	49	59.9 (10.8)

Figure S42. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

'RPPA cHierClus subtypes' versus 'PATHOLOGIC_STAGE'

P value = 0.029 (Fisher's exact test), Q value = 0.099

Table S47. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

nPatients	STAGE I	STAGE II	STAGE III	STAGE IVA	STAGE IVB
ALL	9	39	31	117	4
subtype1	3	18	12	41	1
subtype2	2	11	7	57	2
subtype3	4	10	12	19	1

Figure S43. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

'RPPA cHierClus subtypes' versus 'PATHOLOGY_T_STAGE'

P value = 0.191 (Fisher's exact test), Q value = 0.31

Table S48. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

nPatients	T0+T1	T2	T3	T4
ALL	13	59	53	79
subtype1	5	26	12	33
subtype2	4	19	27	31
subtype3	4	14	14	15

Figure S44. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

'RPPA cHierClus subtypes' versus 'PATHOLOGY_N_STAGE'

P value = 2e-05 (Fisher's exact test), Q value = 0.00056

Table S49. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

nPatients	N0	N1	N2	N3
ALL	72	21	79	4
subtype1	29	9	26	1
subtype2	15	5	45	2
subtype3	28	7	8	1

Figure S45. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

'RPPA cHierClus subtypes' versus 'GENDER'

P value = 0.215 (Fisher's exact test), Q value = 0.33

Table S50. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	62	150
subtype1	22	56
subtype2	21	64
subtype3	19	30

Figure S46. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #7: 'GENDER'

'RPPA cHierClus subtypes' versus 'RADIATION_THERAPY'

P value = 0.757 (Fisher's exact test), Q value = 0.83

Table S51. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #8: 'RADIATION_THERAPY'

nPatients	NO	YES
ALL	73	87
subtype1	30	33
subtype2	25	35
subtype3	18	19

Figure S47. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #8: 'RADIATION_THERAPY'

'RPPA cHierClus subtypes' versus 'NUMBER_PACK_YEARS_SMOKED'

P value = 0.586 (Kruskal-Wallis (anova)), Q value = 0.7

Table S52. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

	nPatients	Mean (Std.Dev)
ALL	110	49.5 (37.5)
subtype1	38	52.6 (48.6)
subtype2	45	52.0 (34.6)
subtype3	27	40.9 (20.5)

Figure S48. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

'RPPA cHierClus subtypes' versus 'YEAR_OF_TOBACCO_SMOKING_ONSET'

P value = 0.223 (Kruskal-Wallis (anova)), Q value = 0.34

Table S53. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

	nPatients	Mean (Std.Dev)
ALL	117	1964.9 (12.3)
subtype1	43	1966.1 (14.7)
subtype2	50	1962.6 (11.2)
subtype3	24	1967.2 (9.4)

Figure S49. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

'RPPA cHierClus subtypes' versus 'NUMBER_OF_LYMPH_NODES'

P value = 4.76e-06 (Kruskal-Wallis (anova)), Q value = 0.00026

Table S54. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

	nPatients	Mean (Std.Dev)
ALL	170	2.9 (5.1)
subtype1	62	2.5 (3.8)
subtype2	70	4.4 (6.8)
subtype3	38	0.7 (1.3)

Figure S50. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

'RPPA cHierClus subtypes' versus 'RACE'

P value = 0.428 (Fisher's exact test), Q value = 0.55

Table S55. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #13: 'RACE'

nPatients	AMERICAN INDIAN OR ALASKA NATIVE	ASIAN	BLACK OR AFRICAN AMERICAN	WHITE
ALL	1	3	19	185
subtype1	1	0	10	66
subtype2	0	2	6	74
subtype3	0	1	3	45

Figure S51. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #13: 'RACE'

'RPPA cHierClus subtypes' versus 'ETHNICITY'

P value = 0.544 (Fisher's exact test), Q value = 0.67

Table S56. Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #14: 'ETHNICITY'

nPatients	HISPANIC OR LATINO	NOT HISPANIC OR LATINO
ALL	13	190
subtype1	7	69
subtype2	4	75
subtype3	2	46

Figure S52. Get High-res Image Clustering Approach #4: 'RPPA cHierClus subtypes' versus Clinical Feature #14: 'ETHNICITY'

Clustering Approach #5: 'RNAseq CNMF subtypes'

Table S57. Description of clustering approach #5: 'RNAseq CNMF subtypes'

Cluster Labels	1	2	3	4
Number of samples	136	114	172	98

'RNAseq CNMF subtypes' versus 'Time to Death'

P value = 0.155 (logrank test), Q value = 0.28

Table S58. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	518	220	0.1 - 211.0 (21.4)
subtype1	136	66	0.5 - 211.0 (19.0)
subtype2	113	49	0.4 - 169.4 (21.0)
subtype3	172	59	0.1 - 139.4 (23.2)
subtype4	97	46	2.1 - 180.2 (23.1)

Figure S53. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

'RNAseq CNMF subtypes' versus 'YEARS_TO_BIRTH'

P value = 0.595 (Kruskal-Wallis (anova)), Q value = 0.7

Table S59. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

	nPatients	Mean (Std.Dev)
ALL	519	60.9 (11.9)
subtype1	136	61.4 (11.9)
subtype2	114	61.5 (13.3)
subtype3	172	60.3 (10.3)
subtype4	97	60.4 (12.7)

Figure S54. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

'RNAseq CNMF subtypes' versus 'PATHOLOGIC_STAGE'

P value = 0.0742 (Fisher's exact test), Q value = 0.19

Table S60. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

nPatients	STAGE I	STAGE II	STAGE III	STAGE IVA	STAGE IVB	STAGE IVC
ALL	27	74	81	253	12	1
subtype1	5	17	17	74	8	0
subtype2	8	20	21	57	0	0
subtype3	4	20	27	75	3	1
subtype4	10	17	16	47	1	0

Figure S55. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

'RNAseq CNMF subtypes' versus 'PATHOLOGY_T_STAGE'

P value = 0.0264 (Fisher's exact test), Q value = 0.097

Table S61. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

nPatients	T0+T1	T2	T3	T4
ALL	49	136	99	174
subtype1	7	37	17	61
subtype2	11	30	29	38
subtype3	17	37	34	48
subtype4	14	32	19	27

Figure S56. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

'RNAseq CNMF subtypes' versus 'PATHOLOGY_N_STAGE'

P value = 0.268 (Fisher's exact test), Q value = 0.38

Table S62. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

nPatients	N0	N1	N2	N3
ALL	176	67	169	8
subtype1	42	17	48	5
subtype2	49	15	32	0
subtype3	52	20	51	3
subtype4	33	15	38	0

Figure S57. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

'RNAseq CNMF subtypes' versus 'PATHOLOGY_M_STAGE'

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

Table S63. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

nPatients	0	1
ALL	186	1
subtype1	46	0
subtype2	50	0
subtype3	55	1
subtype4	35	0

Figure S58. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

'RNAseq CNMF subtypes' versus 'GENDER'

P value = 4e-05 (Fisher's exact test), Q value = 8e-04

Table S64. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	136	384
subtype1	32	104
subtype2	47	67
subtype3	29	143
subtype4	28	70

Figure S59. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #7: 'GENDER'

'RNAseq CNMF subtypes' versus 'RADIATION_THERAPY'

P value = 0.00061 (Fisher's exact test), Q value = 0.0078

Table S65. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #8: 'RADIATION_THERAPY'

nPatients	NO	YES
ALL	161	297
subtype1	39	80
subtype2	39	65
subtype3	38	111
subtype4	45	41

Figure S60. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #8: 'RADIATION_THERAPY'

'RNAseq CNMF subtypes' versus 'HISTOLOGICAL_TYPE'

P value = 0.00234 (Fisher's exact test), Q value = 0.018

Table S66. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

nPatients	HEAD & NECK SQUAMOUS CELL CARCINOMA	HEAD & NECK SQUAMOUS CELL CARCINOMA BASALOID TYPE	HEAD & NECK SQUAMOUS CELL CARCINOMA, SPINDLE CELL VARIANT
ALL	509	10	1
subtype1	134	1	1
subtype2	114	0	0
subtype3	163	9	0
subtype4	98	0	0

Figure S61. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

'RNAseq CNMF subtypes' versus 'NUMBER_PACK_YEARS_SMOKED'

P value = 0.154 (Kruskal-Wallis (anova)), Q value = 0.28

Table S67. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

	nPatients	Mean (Std.Dev)
ALL	296	45.8 (35.3)
subtype1	76	48.6 (42.5)
subtype2	59	36.6 (20.0)
subtype3	110	47.7 (29.9)
subtype4	51	48.2 (45.9)

Figure S62. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

'RNAseq CNMF subtypes' versus 'YEAR_OF_TOBACCO_SMOKING_ONSET'

P value = 0.839 (Kruskal-Wallis (anova)), Q value = 0.89

Table S68. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

	nPatients	Mean (Std.Dev)
ALL	281	1967.4 (12.7)
subtype1	74	1967.4 (13.0)
subtype2	55	1968.3 (13.7)
subtype3	97	1966.2 (12.2)
subtype4	55	1968.5 (12.4)

Figure S63. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

'RNAseq CNMF subtypes' versus 'NUMBER_OF_LYMPH_NODES'

P value = 0.14 (Kruskal-Wallis (anova)), Q value = 0.27

Table S69. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

	nPatients	Mean (Std.Dev)
ALL	405	2.2 (4.3)
subtype1	109	2.7 (5.2)
subtype2	95	1.6 (2.4)
subtype3	114	1.9 (3.6)
subtype4	87	2.7 (5.3)

Figure S64. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

'RNAseq CNMF subtypes' versus 'RACE'

P value = 0.059 (Fisher's exact test), Q value = 0.17

Table S70. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #13: 'RACE'

nPatients	AMERICAN INDIAN OR ALASKA NATIVE	ASIAN	BLACK OR AFRICAN AMERICAN	WHITE
ALL	2	11	48	444
subtype1	1	2	18	109
subtype2	0	6	7	98
subtype3	0	2	18	148
subtype4	1	1	5	89

Figure S65. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #13: 'RACE'

'RNAseq CNMF subtypes' versus 'ETHNICITY'

P value = 0.648 (Fisher's exact test), Q value = 0.74

Table S71. Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #14: 'ETHNICITY'

nPatients	HISPANIC OR LATINO	NOT HISPANIC OR LATINO
ALL	25	458
subtype1	8	114
subtype2	5	105
subtype3	6	153
subtype4	6	86

Figure S66. Get High-res Image Clustering Approach #5: 'RNAseq CNMF subtypes' versus Clinical Feature #14: 'ETHNICITY'

Clustering Approach #6: 'RNAseq cHierClus subtypes'

Table S72. Description of clustering approach #6: 'RNAseq cHierClus subtypes'

Cluster Labels	1	2	3	4
Number of samples	179	104	172	65

'RNAseq cHierClus subtypes' versus 'Time to Death'

P value = 0.00537 (logrank test), Q value = 0.03

Table S73. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	518	220	0.1 - 211.0 (21.4)
subtype1	177	89	0.4 - 180.2 (21.2)
subtype2	104	47	0.1 - 129.2 (25.4)
subtype3	172	72	0.4 - 211.0 (18.5)
subtype4	65	12	0.1 - 139.4 (23.4)

Figure S67. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

'RNAseq cHierClus subtypes' versus 'YEARS_TO_BIRTH'

P value = 0.00439 (Kruskal-Wallis (anova)), Q value = 0.028

Table S74. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

	nPatients	Mean (Std.Dev)
ALL	519	60.9 (11.9)
subtype1	178	59.2 (13.0)
subtype2	104	61.9 (9.8)
subtype3	172	63.0 (12.0)
subtype4	65	58.1 (10.3)

Figure S68. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

'RNAseq cHierClus subtypes' versus 'PATHOLOGIC_STAGE'

P value = 0.0986 (Fisher's exact test), Q value = 0.23

Table S75. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

nPatients	STAGE I	STAGE II	STAGE III	STAGE IVA	STAGE IVB	STAGE IVC
ALL	27	74	81	253	12	1
subtype1	17	25	32	88	4	0
subtype2	0	15	18	54	3	1
subtype3	6	29	25	90	4	0
subtype4	4	5	6	21	1	0

Figure S69. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

'RNAseq cHierClus subtypes' versus 'PATHOLOGY_T_STAGE'

P value = 3e-05 (Fisher's exact test), Q value = 7e-04

Table S76. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

nPatients	T0+T1	T2	T3	T4
ALL	49	136	99	174
subtype1	25	54	42	47
subtype2	5	23	24	41
subtype3	9	44	25	78
subtype4	10	15	8	8

Figure S70. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

'RNAseq cHierClus subtypes' versus 'PATHOLOGY_N_STAGE'

P value = 0.123 (Fisher's exact test), Q value = 0.26

Table S77. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

nPatients	N0	N1	N2	N3
ALL	176	67	169	8
subtype1	58	25	71	3
subtype2	38	12	32	3
subtype3	70	25	46	1
subtype4	10	5	20	1

Figure S71. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

'RNAseq cHierClus subtypes' versus 'PATHOLOGY_M_STAGE'

P value = 0.26 (Fisher's exact test), Q value = 0.38

Table S78. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

nPatients	0	1
ALL	186	1
subtype1	68	0
subtype2	28	1
subtype3	70	0
subtype4	20	0

Figure S72. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

'RNAseq cHierClus subtypes' versus 'GENDER'

P value = 7e-05 (Fisher's exact test), Q value = 0.0012

Table S79. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	136	384
subtype1	53	126
subtype2	23	81
subtype3	56	116
subtype4	4	61

Figure S73. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #7: 'GENDER'

'RNAseq cHierClus subtypes' versus 'RADIATION_THERAPY'

P value = 0.00243 (Fisher's exact test), Q value = 0.018

Table S80. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #8: 'RADIATION_THERAPY'

nPatients	NO	YES
ALL	161	297
subtype1	68	92
subtype2	29	55
subtype3	54	98
subtype4	10	52

Figure S74. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #8: 'RADIATION_THERAPY'

'RNAseq cHierClus subtypes' versus 'HISTOLOGICAL_TYPE'

P value = 0.00025 (Fisher's exact test), Q value = 0.0039

Table S81. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

nPatients	HEAD & NECK SQUAMOUS CELL CARCINOMA	HEAD & NECK SQUAMOUS CELL CARCINOMA BASALOID TYPE	HEAD & NECK SQUAMOUS CELL CARCINOMA, SPINDLE CELL VARIANT
ALL	509	10	1
subtype1	178	1	0
subtype2	100	4	0
subtype3	171	0	1
subtype4	60	5	0

Figure S75. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

'RNAseq cHierClus subtypes' versus 'NUMBER_PACK_YEARS_SMOKED'

P value = 0.00186 (Kruskal-Wallis (anova)), Q value = 0.016

Table S82. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

	nPatients	Mean (Std.Dev)
ALL	296	45.8 (35.3)
subtype1	96	42.9 (36.9)
subtype2	72	54.1 (28.8)
subtype3	92	42.1 (35.2)
subtype4	36	46.6 (41.1)

Figure S76. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

'RNAseq cHierClus subtypes' versus 'YEAR_OF_TOBACCO_SMOKING_ONSET'

P value = 0.199 (Kruskal-Wallis (anova)), Q value = 0.32

Table S83. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

	nPatients	Mean (Std.Dev)
ALL	281	1967.4 (12.7)
subtype1	98	1969.0 (13.1)
subtype2	65	1964.7 (10.5)
subtype3	85	1968.2 (13.2)
subtype4	33	1965.6 (13.7)

Figure S77. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

'RNAseq cHierClus subtypes' versus 'NUMBER_OF_LYMPH_NODES'

P value = 0.0102 (Kruskal-Wallis (anova)), Q value = 0.052

Table S84. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

	nPatients	Mean (Std.Dev)
ALL	405	2.2 (4.3)
subtype1	155	2.4 (4.3)
subtype2	84	2.4 (6.1)
subtype3	136	1.8 (3.1)
subtype4	30	2.7 (3.1)

Figure S78. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

'RNAseq cHierClus subtypes' versus 'RACE'

P value = 0.126 (Fisher's exact test), Q value = 0.26

Table S85. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #13: 'RACE'

nPatients	AMERICAN INDIAN OR ALASKA NATIVE	ASIAN	BLACK OR AFRICAN AMERICAN	WHITE
ALL	2	11	48	444
subtype1	1	8	11	153
subtype2	0	1	15	84
subtype3	1	2	17	147
subtype4	0	0	5	60

Figure S79. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #13: 'RACE'

'RNAseq cHierClus subtypes' versus 'ETHNICITY'

P value = 0.774 (Fisher's exact test), Q value = 0.85

Table S86. Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #14: 'ETHNICITY'

nPatients	HISPANIC OR LATINO	NOT HISPANIC OR LATINO
ALL	25	458
subtype1	10	159
subtype2	3	89
subtype3	8	152
subtype4	4	58

Figure S80. Get High-res Image Clustering Approach #6: 'RNAseq cHierClus subtypes' versus Clinical Feature #14: 'ETHNICITY'

Clustering Approach #7: 'MIRSEQ CNMF'

Table S87. Description of clustering approach #7: 'MIRSEQ CNMF'

Cluster Labels	1	2	3	4	5
Number of samples	81	140	132	91	79

'MIRSEQ CNMF' versus 'Time to Death'

P value = 0.00483 (logrank test), Q value = 0.028

Table S88. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	521	221	0.1 - 211.0 (21.2)
subtype1	81	40	0.4 - 153.9 (19.1)
subtype2	140	50	0.4 - 169.4 (19.4)
subtype3	132	41	0.1 - 139.4 (27.4)
subtype4	91	55	0.1 - 211.0 (20.5)
subtype5	77	35	2.1 - 180.2 (21.5)

Figure S81. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #1: 'Time to Death'

'MIRSEQ CNMF' versus 'YEARS_TO_BIRTH'

P value = 0.287 (Kruskal-Wallis (anova)), Q value = 0.4

Table S89. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

	nPatients	Mean (Std.Dev)
ALL	522	60.9 (12.0)
subtype1	81	61.2 (12.5)
subtype2	140	60.4 (12.7)
subtype3	132	60.7 (10.1)
subtype4	91	63.3 (13.4)
subtype5	78	59.4 (11.1)

Figure S82. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

'MIRSEQ CNMF' versus 'PATHOLOGIC_STAGE'

P value = 0.134 (Fisher's exact test), Q value = 0.26

Table S90. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

nPatients	STAGE I	STAGE II	STAGE III	STAGE IVA	STAGE IVB	STAGE IVC
ALL	27	76	79	257	12	1
subtype1	0	7	13	45	4	1
subtype2	10	22	26	66	2	0
subtype3	3	15	13	63	2	0
subtype4	7	18	16	44	2	0
subtype5	7	14	11	39	2	0

Figure S83. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

'MIRSEQ CNMF' versus 'PATHOLOGY_T_STAGE'

P value = 0.122 (Fisher's exact test), Q value = 0.26

Table S91. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

nPatients	T0+T1	T2	T3	T4
ALL	49	138	99	175
subtype1	4	12	16	40
subtype2	13	38	31	45
subtype3	11	35	21	34
subtype4	9	29	16	33
subtype5	12	24	15	23

Figure S84. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

'MIRSEQ CNMF' versus 'PATHOLOGY_N_STAGE'

P value = 0.612 (Fisher's exact test), Q value = 0.71

Table S92. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

nPatients	N0	N1	N2	N3
ALL	176	68	171	8
subtype1	32	7	28	2
subtype2	55	22	40	2
subtype3	33	12	46	2
subtype4	28	15	28	1
subtype5	28	12	29	1

Figure S85. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

'MIRSEQ CNMF' versus 'PATHOLOGY_M_STAGE'

P value = 0.273 (Fisher's exact test), Q value = 0.39

Table S93. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

nPatients	0	1
ALL	189	1
subtype1	26	1
subtype2	68	0
subtype3	42	0
subtype4	25	0
subtype5	28	0

Figure S86. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

'MIRSEQ CNMF' versus 'GENDER'

P value = 0.00198 (Fisher's exact test), Q value = 0.016

Table S94. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	141	382
subtype1	22	59
subtype2	48	92
subtype3	19	113
subtype4	26	65
subtype5	26	53

Figure S87. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #7: 'GENDER'

'MIRSEQ CNMF' versus 'RADIATION_THERAPY'

P value = 0.024 (Fisher's exact test), Q value = 0.091

Table S95. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #8: 'RADIATION_THERAPY'

nPatients	NO	YES
ALL	161	300
subtype1	29	43
subtype2	44	87
subtype3	29	90
subtype4	27	42
subtype5	32	38

Figure S88. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #8: 'RADIATION_THERAPY'

'MIRSEQ CNMF' versus 'HISTOLOGICAL_TYPE'

P value = 0.00467 (Fisher's exact test), Q value = 0.028

Table S96. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

nPatients	HEAD & NECK SQUAMOUS CELL CARCINOMA	HEAD & NECK SQUAMOUS CELL CARCINOMA BASALOID TYPE	HEAD & NECK SQUAMOUS CELL CARCINOMA, SPINDLE CELL VARIANT
ALL	512	10	1
subtype1	78	2	1
subtype2	140	0	0
subtype3	125	7	0
subtype4	91	0	0
subtype5	78	1	0

Figure S89. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

'MIRSEQ CNMF' versus 'NUMBER_PACK_YEARS_SMOKED'

P value = 0.583 (Kruskal-Wallis (anova)), Q value = 0.7

Table S97. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

	nPatients	Mean (Std.Dev)
ALL	295	45.7 (35.2)
subtype1	46	45.1 (24.8)
subtype2	81	45.5 (37.4)
subtype3	80	48.3 (35.3)
subtype4	41	49.5 (47.8)
subtype5	47	39.0 (26.1)

Figure S90. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

'MIRSEQ CNMF' versus 'YEAR_OF_TOBACCO_SMOKING_ONSET'

P value = 0.0559 (Kruskal-Wallis (anova)), Q value = 0.17

Table S98. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

	nPatients	Mean (Std.Dev)
ALL	279	1967.2 (12.7)
subtype1	38	1963.9 (10.3)
subtype2	72	1970.6 (12.6)
subtype3	75	1966.3 (13.1)
subtype4	48	1966.1 (15.1)
subtype5	46	1967.3 (10.7)

Figure S91. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

'MIRSEQ CNMF' versus 'NUMBER_OF_LYMPH_NODES'

P value = 0.157 (Kruskal-Wallis (anova)), Q value = 0.28

Table S99. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

	nPatients	Mean (Std.Dev)
ALL	410	2.2 (4.3)
subtype1	68	2.8 (6.7)
subtype2	119	1.6 (2.6)
subtype3	84	2.2 (2.6)
subtype4	72	1.9 (3.0)
subtype5	67	3.0 (6.0)

Figure S92. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

'MIRSEQ CNMF' versus 'RACE'

P value = 0.0119 (Fisher's exact test), Q value = 0.055

Table S100. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #13: 'RACE'

nPatients	AMERICAN INDIAN OR ALASKA NATIVE	ASIAN	BLACK OR AFRICAN AMERICAN	WHITE
ALL	2	11	47	448
subtype1	0	1	12	64
subtype2	0	9	11	116
subtype3	1	0	13	117
subtype4	0	1	5	82
subtype5	1	0	6	69

Figure S93. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #13: 'RACE'

'MIRSEQ CNMF' versus 'ETHNICITY'

P value = 0.438 (Fisher's exact test), Q value = 0.56

Table S101. Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #14: 'ETHNICITY'

nPatients	HISPANIC OR LATINO	NOT HISPANIC OR LATINO
ALL	25	461
subtype1	2	69
subtype2	7	129
subtype3	4	118
subtype4	6	78
subtype5	6	67

Figure S94. Get High-res Image Clustering Approach #7: 'MIRSEQ CNMF' versus Clinical Feature #14: 'ETHNICITY'

Clustering Approach #8: 'MIRSEQ CHIERARCHICAL'

Table S102. Description of clustering approach #8: 'MIRSEQ CHIERARCHICAL'

Cluster Labels	1	2	3
Number of samples	193	144	186

'MIRSEQ CHIERARCHICAL' versus 'Time to Death'

P value = 0.028 (logrank test), Q value = 0.098

Table S103. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	521	221	0.1 - 211.0 (21.2)
subtype1	191	97	0.4 - 180.2 (20.9)
subtype2	144	59	0.1 - 153.9 (20.6)
subtype3	186	65	0.1 - 211.0 (22.7)

Figure S95. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #1: 'Time to Death'

'MIRSEQ CHIERARCHICAL' versus 'YEARS_TO_BIRTH'

P value = 0.0733 (Kruskal-Wallis (anova)), Q value = 0.19

Table S104. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

	nPatients	Mean (Std.Dev)
ALL	522	60.9 (12.0)
subtype1	192	59.6 (13.5)
subtype2	144	62.8 (11.1)
subtype3	186	60.8 (10.8)

Figure S96. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

'MIRSEQ CHIERARCHICAL' versus 'PATHOLOGIC_STAGE'

P value = 0.0591 (Fisher's exact test), Q value = 0.17

Table S105. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

nPatients	STAGE I	STAGE II	STAGE III	STAGE IVA	STAGE IVB	STAGE IVC
ALL	27	76	79	257	12	1
subtype1	19	31	31	94	5	0
subtype2	3	18	20	80	5	1
subtype3	5	27	28	83	2	0

Figure S97. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

'MIRSEQ CHIERARCHICAL' versus 'PATHOLOGY_T_STAGE'

P value = 0.00165 (Fisher's exact test), Q value = 0.015

Table S106. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

nPatients	T0+T1	T2	T3	T4
ALL	49	138	99	175
subtype1	27	58	41	55
subtype2	5	30	27	66
subtype3	17	50	31	54

Figure S98. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

'MIRSEQ CHIERARCHICAL' versus 'PATHOLOGY_N_STAGE'

P value = 0.0861 (Fisher's exact test), Q value = 0.2

Table S107. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

nPatients	N0	N1	N2	N3
ALL	176	68	171	8
subtype1	67	25	73	3
subtype2	61	15	40	4
subtype3	48	28	58	1

Figure S99. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

'MIRSEQ CHIERARCHICAL' versus 'PATHOLOGY_M_STAGE'

P value = 0.278 (Fisher's exact test), Q value = 0.39

Table S108. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

nPatients	0	1
ALL	189	1
subtype1	74	0
subtype2	51	1
subtype3	64	0

Figure S100. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

'MIRSEQ CHIERARCHICAL' versus 'GENDER'

P value = 0.0108 (Fisher's exact test), Q value = 0.052

Table S109. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	141	382
subtype1	58	135
subtype2	47	97
subtype3	36	150

Figure S101. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #7: 'GENDER'

'MIRSEQ CHIERARCHICAL' versus 'RADIATION_THERAPY'

P value = 0.0107 (Fisher's exact test), Q value = 0.052

Table S110. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #8: 'RADIATION_THERAPY'

nPatients	NO	YES
ALL	161	300
subtype1	74	96
subtype2	39	82
subtype3	48	122

Figure S102. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #8: 'RADIATION_THERAPY'

'MIRSEQ CHIERARCHICAL' versus 'HISTOLOGICAL_TYPE'

P value = 0.238 (Fisher's exact test), Q value = 0.36

Table S111. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

nPatients	HEAD & NECK SQUAMOUS CELL CARCINOMA	HEAD & NECK SQUAMOUS CELL CARCINOMA BASALOID TYPE	HEAD & NECK SQUAMOUS CELL CARCINOMA, SPINDLE CELL VARIANT
ALL	512	10	1
subtype1	191	2	0
subtype2	141	2	1
subtype3	180	6	0

Figure S103. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

'MIRSEQ CHIERARCHICAL' versus 'NUMBER_PACK_YEARS_SMOKED'

P value = 0.182 (Kruskal-Wallis (anova)), Q value = 0.31

Table S112. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

	nPatients	Mean (Std.Dev)
ALL	295	45.7 (35.2)
subtype1	96	44.0 (36.8)
subtype2	77	49.9 (30.1)
subtype3	122	44.3 (36.8)

Figure S104. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

'MIRSEQ CHIERARCHICAL' versus 'YEAR_OF_TOBACCO_SMOKING_ONSET'

P value = 0.0804 (Kruskal-Wallis (anova)), Q value = 0.2

Table S113. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

	nPatients	Mean (Std.Dev)
ALL	279	1967.2 (12.7)
subtype1	99	1969.4 (13.2)
subtype2	72	1964.6 (10.5)
subtype3	108	1966.9 (13.4)

Figure S105. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

'MIRSEQ CHIERARCHICAL' versus 'NUMBER_OF_LYMPH_NODES'

P value = 0.00425 (Kruskal-Wallis (anova)), Q value = 0.028

Table S114. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

	nPatients	Mean (Std.Dev)
ALL	410	2.2 (4.3)
subtype1	167	2.4 (4.2)
subtype2	119	1.6 (3.5)
subtype3	124	2.5 (4.9)

Figure S106. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

'MIRSEQ CHIERARCHICAL' versus 'RACE'

P value = 0.0272 (Fisher's exact test), Q value = 0.098

Table S115. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #13: 'RACE'

nPatients	AMERICAN INDIAN OR ALASKA NATIVE	ASIAN	BLACK OR AFRICAN AMERICAN	WHITE
ALL	2	11	47	448
subtype1	1	8	12	165
subtype2	0	2	20	118
subtype3	1	1	15	165

Figure S107. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #13: 'RACE'

'MIRSEQ CHIERARCHICAL' versus 'ETHNICITY'

P value = 0.377 (Fisher's exact test), Q value = 0.5

Table S116. Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #14: 'ETHNICITY'

nPatients	HISPANIC OR LATINO	NOT HISPANIC OR LATINO
ALL	25	461
subtype1	12	169
subtype2	4	128
subtype3	9	164

Figure S108. Get High-res Image Clustering Approach #8: 'MIRSEQ CHIERARCHICAL' versus Clinical Feature #14: 'ETHNICITY'

Clustering Approach #9: 'MIRseq Mature CNMF subtypes'

Table S117. Description of clustering approach #9: 'MIRseq Mature CNMF subtypes'

Cluster Labels	1	2	3
Number of samples	144	189	144

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

P value = 0.656 (logrank test), Q value = 0.75

Table S118. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	475	197	0.1 - 211.0 (21.2)
subtype1	144	71	0.1 - 211.0 (22.2)
subtype2	187	67	0.1 - 180.2 (19.7)
subtype3	144	59	0.4 - 139.4 (24.9)

Figure S109. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #1: 'Time to Death'

'MIRseq Mature CNMF subtypes' versus 'YEARS_TO_BIRTH'

P value = 0.811 (Kruskal-Wallis (anova)), Q value = 0.88

Table S119. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

	nPatients	Mean (Std.Dev)
ALL	476	61.0 (11.8)
subtype1	144	61.4 (11.7)
subtype2	188	61.2 (12.0)
subtype3	144	60.4 (11.5)

Figure S110. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

'MIRseq Mature CNMF subtypes' versus 'PATHOLOGIC_STAGE'

P value = 0.833 (Fisher's exact test), Q value = 0.89

Table S120. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

nPatients	STAGE I	STAGE II	STAGE III	STAGE IVA	STAGE IVB	STAGE IVC
ALL	25	72	74	230	10	1
subtype1	9	25	23	78	3	0
subtype2	8	26	34	87	6	1
subtype3	8	21	17	65	1	0

Figure S111. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

'MIRseq Mature CNMF subtypes' versus 'PATHOLOGY_T_STAGE'

P value = 0.825 (Fisher's exact test), Q value = 0.89

Table S121. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

nPatients	T0+T1	T2	T3	T4
ALL	46	125	90	158
subtype1	14	42	29	53
subtype2	20	55	33	60
subtype3	12	28	28	45

Figure S112. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

'MIRseq Mature CNMF subtypes' versus 'PATHOLOGY_N_STAGE'

P value = 0.686 (Fisher's exact test), Q value = 0.77

Table S122. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

nPatients	N0	N1	N2	N3
ALL	161	65	154	6
subtype1	48	21	58	1
subtype2	68	28	56	4
subtype3	45	16	40	1

Figure S113. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

'MIRseq Mature CNMF subtypes' versus 'PATHOLOGY_M_STAGE'

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

Table S123. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

nPatients	0	1
ALL	175	1
subtype1	45	0
subtype2	99	1
subtype3	31	0

Figure S114. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

'MIRseq Mature CNMF subtypes' versus 'GENDER'

P value = 0.359 (Fisher's exact test), Q value = 0.48

Table S124. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	130	347
subtype1	43	101
subtype2	54	135
subtype3	33	111

Figure S115. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #7: 'GENDER'

'MIRseq Mature CNMF subtypes' versus 'RADIATION_THERAPY'

P value = 0.0316 (Fisher's exact test), Q value = 0.1

Table S125. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #8: 'RADIATION_THERAPY'

nPatients	NO	YES
ALL	146	275
subtype1	53	69
subtype2	51	127
subtype3	42	79

Figure S116. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #8: 'RADIATION_THERAPY'

'MIRseq Mature CNMF subtypes' versus 'HISTOLOGICAL_TYPE'

P value = 0.191 (Fisher's exact test), Q value = 0.31

Table S126. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

nPatients	HEAD & NECK SQUAMOUS CELL CARCINOMA	HEAD & NECK SQUAMOUS CELL CARCINOMA BASALOID TYPE	HEAD & NECK SQUAMOUS CELL CARCINOMA, SPINDLE CELL VARIANT
ALL	466	10	1
subtype1	142	2	0
subtype2	181	7	1
subtype3	143	1	0

Figure S117. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

'MIRseq Mature CNMF subtypes' versus 'NUMBER_PACK_YEARS_SMOKED'

P value = 0.000916 (Kruskal-Wallis (anova)), Q value = 0.0099

Table S127. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

	nPatients	Mean (Std.Dev)
ALL	262	45.9 (36.2)
subtype1	71	43.4 (28.9)
subtype2	104	37.2 (22.5)
subtype3	87	58.3 (49.4)

Figure S118. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

'MIRseq Mature CNMF subtypes' versus 'YEAR_OF_TOBACCO_SMOKING_ONSET'

P value = 0.114 (Kruskal-Wallis (anova)), Q value = 0.26

Table S128. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

	nPatients	Mean (Std.Dev)
ALL	255	1967.1 (12.9)
subtype1	77	1966.6 (12.4)
subtype2	86	1969.1 (13.4)
subtype3	92	1965.8 (12.7)

Figure S119. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

'MIRseq Mature CNMF subtypes' versus 'NUMBER_OF_LYMPH_NODES'

P value = 0.0929 (Kruskal-Wallis (anova)), Q value = 0.22

Table S129. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

	nPatients	Mean (Std.Dev)
ALL	376	2.2 (4.3)
subtype1	124	2.9 (6.1)
subtype2	155	1.5 (2.1)
subtype3	97	2.2 (4.1)

Figure S120. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

'MIRseq Mature CNMF subtypes' versus 'RACE'

P value = 0.243 (Fisher's exact test), Q value = 0.36

Table S130. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #13: 'RACE'

nPatients	AMERICAN INDIAN OR ALASKA NATIVE	ASIAN	BLACK OR AFRICAN AMERICAN	WHITE
ALL	2	11	44	405
subtype1	0	1	14	123
subtype2	2	8	15	156
subtype3	0	2	15	126

Figure S121. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #13: 'RACE'

'MIRseq Mature CNMF subtypes' versus 'ETHNICITY'

P value = 0.931 (Fisher's exact test), Q value = 0.98

Table S131. Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #14: 'ETHNICITY'

nPatients	HISPANIC OR LATINO	NOT HISPANIC OR LATINO
ALL	24	417
subtype1	8	127
subtype2	9	161
subtype3	7	129

Figure S122. Get High-res Image Clustering Approach #9: 'MIRseq Mature CNMF subtypes' versus Clinical Feature #14: 'ETHNICITY'

Clustering Approach #10: 'MIRseq Mature cHierClus subtypes'

Table S132. Description of clustering approach #10: 'MIRseq Mature cHierClus subtypes'

Cluster Labels	1	2	3
Number of samples	174	104	199

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

P value = 0.0582 (logrank test), Q value = 0.17

Table S133. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

	nPatients	nDeath	Duration Range (Median), Month
ALL	475	197	0.1 - 211.0 (21.2)
subtype1	172	83	0.4 - 180.2 (21.6)
subtype2	104	45	0.1 - 211.0 (18.0)
subtype3	199	69	0.1 - 172.7 (23.8)

Figure S123. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #1: 'Time to Death'

'MIRseq Mature cHierClus subtypes' versus 'YEARS_TO_BIRTH'

P value = 0.739 (Kruskal-Wallis (anova)), Q value = 0.82

Table S134. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

	nPatients	Mean (Std.Dev)
ALL	476	61.0 (11.8)
subtype1	173	60.3 (13.5)
subtype2	104	61.9 (11.9)
subtype3	199	61.2 (9.9)

Figure S124. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #2: 'YEARS_TO_BIRTH'

'MIRseq Mature cHierClus subtypes' versus 'PATHOLOGIC_STAGE'

P value = 0.0822 (Fisher's exact test), Q value = 0.2

Table S135. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

nPatients	STAGE I	STAGE II	STAGE III	STAGE IVA	STAGE IVB	STAGE IVC
ALL	25	72	74	230	10	1
subtype1	16	32	32	79	4	0
subtype2	2	13	13	61	4	0
subtype3	7	27	29	90	2	1

Figure S125. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #3: 'PATHOLOGIC_STAGE'

'MIRseq Mature cHierClus subtypes' versus 'PATHOLOGY_T_STAGE'

P value = 0.0179 (Fisher's exact test), Q value = 0.071

Table S136. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

nPatients	T0+T1	T2	T3	T4
ALL	46	125	90	158
subtype1	24	54	39	46
subtype2	6	23	17	49
subtype3	16	48	34	63

Figure S126. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #4: 'PATHOLOGY_T_STAGE'

'MIRseq Mature cHierClus subtypes' versus 'PATHOLOGY_N_STAGE'

P value = 0.951 (Fisher's exact test), Q value = 0.99

Table S137. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

nPatients	N0	N1	N2	N3
ALL	161	65	154	6
subtype1	64	24	61	2
subtype2	37	13	38	2
subtype3	60	28	55	2

Figure S127. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #5: 'PATHOLOGY_N_STAGE'

'MIRseq Mature cHierClus subtypes' versus 'PATHOLOGY_M_STAGE'

P value = 0.633 (Fisher's exact test), Q value = 0.73

Table S138. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

nPatients	0	1
ALL	175	1
subtype1	65	0
subtype2	54	0
subtype3	56	1

Figure S128. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #6: 'PATHOLOGY_M_STAGE'

'MIRseq Mature cHierClus subtypes' versus 'GENDER'

P value = 0.122 (Fisher's exact test), Q value = 0.26

Table S139. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #7: 'GENDER'

nPatients	FEMALE	MALE
ALL	130	347
subtype1	57	117
subtype2	26	78
subtype3	47	152

Figure S129. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #7: 'GENDER'

'MIRseq Mature cHierClus subtypes' versus 'RADIATION_THERAPY'

P value = 0.00162 (Fisher's exact test), Q value = 0.015

Table S140. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #8: 'RADIATION_THERAPY'

nPatients	NO	YES
ALL	146	275
subtype1	69	81
subtype2	26	70
subtype3	51	124

Figure S130. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #8: 'RADIATION_THERAPY'

'MIRseq Mature cHierClus subtypes' versus 'HISTOLOGICAL_TYPE'

P value = 0.159 (Fisher's exact test), Q value = 0.28

Table S141. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

nPatients	HEAD & NECK SQUAMOUS CELL CARCINOMA	HEAD & NECK SQUAMOUS CELL CARCINOMA BASALOID TYPE	HEAD & NECK SQUAMOUS CELL CARCINOMA, SPINDLE CELL VARIANT
ALL	466	10	1
subtype1	172	2	0
subtype2	102	1	1
subtype3	192	7	0

Figure S131. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #9: 'HISTOLOGICAL_TYPE'

'MIRseq Mature cHierClus subtypes' versus 'NUMBER_PACK_YEARS_SMOKED'

P value = 0.556 (Kruskal-Wallis (anova)), Q value = 0.68

Table S142. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

	nPatients	Mean (Std.Dev)
ALL	262	45.9 (36.2)
subtype1	86	44.8 (38.8)
subtype2	54	45.8 (42.2)
subtype3	122	46.7 (31.5)

Figure S132. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #10: 'NUMBER_PACK_YEARS_SMOKED'

'MIRseq Mature cHierClus subtypes' versus 'YEAR_OF_TOBACCO_SMOKING_ONSET'

P value = 0.118 (Kruskal-Wallis (anova)), Q value = 0.26

Table S143. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

	nPatients	Mean (Std.Dev)
ALL	255	1967.1 (12.9)
subtype1	94	1968.7 (13.5)
subtype2	41	1968.9 (13.3)
subtype3	120	1965.4 (12.1)

Figure S133. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #11: 'YEAR_OF_TOBACCO_SMOKING_ONSET'

'MIRseq Mature cHierClus subtypes' versus 'NUMBER_OF_LYMPH_NODES'

P value = 0.597 (Kruskal-Wallis (anova)), Q value = 0.7

Table S144. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

	nPatients	Mean (Std.Dev)
ALL	376	2.2 (4.3)
subtype1	148	2.1 (4.1)
subtype2	87	2.2 (3.4)
subtype3	141	2.2 (5.0)

Figure S134. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #12: 'NUMBER_OF_LYMPH_NODES'

'MIRseq Mature cHierClus subtypes' versus 'RACE'

P value = 0.181 (Fisher's exact test), Q value = 0.31

Table S145. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #13: 'RACE'

nPatients	AMERICAN INDIAN OR ALASKA NATIVE	ASIAN	BLACK OR AFRICAN AMERICAN	WHITE
ALL	2	11	44	405
subtype1	1	6	10	151
subtype2	1	1	13	84
subtype3	0	4	21	170

Figure S135. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #13: 'RACE'

'MIRseq Mature cHierClus subtypes' versus 'ETHNICITY'

P value = 0.486 (Fisher's exact test), Q value = 0.61

Table S146. Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #14: 'ETHNICITY'

nPatients	HISPANIC OR LATINO	NOT HISPANIC OR LATINO
ALL	24	417
subtype1	12	155
subtype2	4	91
subtype3	8	171

Figure S136. Get High-res Image Clustering Approach #10: 'MIRseq Mature cHierClus subtypes' versus Clinical Feature #14: 'ETHNICITY'

Methods & Data

Input

Cluster data file = /xchip/cga/gdac-prod/tcga-gdac/jobResults/GDAC_mergedClustering/HNSC-TP/22541004/HNSC-TP.mergedcluster.txt
Clinical data file = /xchip/cga/gdac-prod/tcga-gdac/jobResults/Append_Data/HNSC-TP/22506490/HNSC-TP.merged_data.txt
Number of patients = 528
Number of clustering approaches = 10
Number of selected clinical features = 14
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

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.

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] 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)

[5] 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