Table 1:

Associations between ovarian cancer gene signatures, stromal genes (SG), tumor genes (TG), and percent stromal content.

Published gene signature		# Number of genesa	A) Overlap with SGs and TGs		B) % stroma contentb	C) Breast stroma vs. epitheliumb		D) Prostate % stromal contentb
Study	Signature components		# TGs (%)	# SGs (%)	TCGA (n = 518)	Boersma (n = 34)	Casey (n = 28)	Wang (n = 109)	Wang (n = 136)
AOCS C1 subtype stromal and tumor genes	Stromal genes (SG) Tumor genes (TG)	461 227	0 227	461 0	5.6×10−13	4.3×10−8	7.1×10−8	1.5×10−9	4.9×10−17
AOCS signatures (Tothill et al.)	Up in C1 Down in C1	287 147	0 2 (1)	176 (61)* 2 (1)	3.8×10−11	2.8×10−8	5.1×10−7	4.9×10−8	1.1×10−14
	Poor PFS Good PFS	135 143	0 16 (11)*	103 (76)* 0	8.6×10−13	4.0×10−9	4.1×10−8	1.2×10−9	2.0×10−17
	Poor OS Good OS	147 146	0 23 (16)*	101 (69)* 0	5.3×10−11	4.6×10−9	9.7×10−8	1.7×10−5	6.5×10−13
MGH stroma vs. tumor signature	Tumor-associated stroma	431 81	0 23 (28%)*	125 (29%)* 0	6.5×10−15	1.6×10−5	2.2×10−8	1.6×10−11	3.6×10−17
	Tumor
MGH tissue-specific genes	Tumor-associated stroma	11 28	0 17 (57%)*	5 (42%)* 0	2.1×10−14 0.136	8.1×10−5 0.001 (−)	0.04 1.8×10−8 (−)	1.6×10−6 3.7×10−9 (−)	1.0×10−7 1.2×10−19 (−)
	Tumor
Bentink et al.	Angiogenic Non-angiogenic	74 19	0 0	35 (47)* 1 (5)	1.6×10−9	1.5×10−7	0.93	1.1×10−6	6.8×10−14
Bignotti et al.	Up in metastasis Up in primary tumor	89 36	0 1 (3)	54 (61)* 0	1.2×10−9	4.5×10−7	5.2×10−5	7.2×10−4	7.2×10−8
Spentzos et al.	Unfavorable prognosis Favorable prognosis	73 43	0 4 (9)*	19 (26)* 0	1.7×10−10	1.8×10−5	0.0074 (−)	4.1×10−14	1.0×10−7
Bonome et al.	Poor prognosis Good prognosis	288 272	0 5 (2)	37 (13)* 1	9.4×10−11	3.5×10−6	1.5×10−6	7.7×10−5	3.0×10−11
Biade et al.	Malignant cluster Benign cluster	21 15	4 (27)* 1	0 6 (29)*	1.1×10−12 (−)	9.8×10−8 (−)	5.2×10−6 (−)	6.5×10−10 (−)	7.1×10−13 (−)
Konstantinopoulos et al.	BRCA-like Non-BRCA-like	32 27	0 0	2 (6) 2 (7)	0.7279 (−)	0.0043 (−)	7.6×10−6 (−)	0.34	0.19

Details of published signatures in Table S4A

^*p < 0.001

A) Gene overlap between microdissected ovarian cancer gene signatures and both SGs and TGs of the AOCS microdissected dataset. P-values calculated with Fisher’s exact test. Association between gene signatures and B) the % stromal proportion in TCGA ovarian cancer tumors or microdissected stroma in C) two breast cancer datasets, or D) two prostate cancer datasets. P-values were calculated using B, D) linear regression or C) t-tests. Associations with stroma were mostly positive. Negative associations with stroma are indicated in parentheses (−) next to p-values, and otherwise, associations were positive. No multiple testing correction was used.

^a.Total number of unique gene symbols reported in the specified gene signature from each study.

^b.P-values derived using paired t-test or linear regression.