“…For PIK3CA and TILs gene signatures, raw intensity files (.CEL) from each microarray were processed using MAS5.0 to generate probe-level intensities and normalized to a median array intensity of 600, transformed to log2 values, and scaled by the expression levels of 1322 breast cancer reference genes within each sample normalized to median values in a reference cohort 35 . Two PIK3CA-related gene signatures were calculated for each microarray: first, a PIK3CA gene signature that measures transcriptional activity associated with PI3KCA mutation 41 , and secondly, a modification of that first signature that includes only the probe sets robust to technical variation and variation due to tumor heterogeneity 42 . Two TIL gene signatures were calculated for each microarray: a gene signature trained on TIL infiltrate (calculated as the mean gene expression of IGKC, CXCL9, CCL5, and TRBC1 minus the mean gene expression of AK2, APPBP2, ATP5MF, DARS1, LDHA, TRIM2, UBE2Z, UGP2, VDAC2, and WIPF2WIPF2), and a signature to predict high TILs after subsequent neoadjuvant chemotherapy 43 .…”