TCGA tumor purity correlation analysis and Col1:Col3 ratio calculation

Pearson correlations and corresponding p-values were calculated between ABSOLUTE tumor purity estimates and the gene expression estimates for the 20,531 unique genes included in the RNA-seq dataset for the TCGA BRCA cohort. Q-values were calculated by adjusting the p-values for false discovery rate using the Benjamini-Hochberg procedure.

Expression values for COL1A1 and COL1A2 were normalized by calculating the ratio of their expression to COL3A1 expression from the RNA-seq dataset for the TCGA BRCA cohort. Pearson correlations were then calculated between ABSOLUTE tumor purity and each of the COL1A1:COL3A1 and COL1A2:COL3A1 expression ratios, again correcting for multiple hypothesis testing with the Benjamini-Hochberg procedure. Non-zero expression of COL1A1, COL1A2, and COL3A1 was confirmed within the Col1:Col3 high and Col3:Col1 high classified samples by comparing their expression to two reference transcripts, GPX5 and PUM1. GPX5, primarily expressed within the epididymis⁸⁸ and largely absent from breast cancer samples⁸⁹, was selected as a negative control, while PUM1 was included as a positive control given its repeated identification as a housekeeping gene within breast cancer^90–92 as well as other tissues¹⁴¹.