Partial least squares based analysis of pathways in recurrent breast cancer

Breast cancer remains a major health problem even with all the recent technological advancements. Large-scale gene expression analysis has offered great ease for both biological characterization and therapeutic planning of breast cancer. Previous studies mostly used variance/regression analysis, which becomes fundamentally flawed when there are unaccounted array specific factors. Here we aim to investigate the underlying mechanism of breast cancer through partial least squares (PLS) based gene expression profile analysis.

With a gene expression profile data set downloaded from the Gene Expression Omnibus database, we performed PLS based analysis.

We acquire 932 and 771 differentially expressed genes (DEGs) in breast cancer metastasis of estrogen-receptor (ER)-positive and ER-negative patients, respectively. For ER-positive patients, 32 pathways were found to be enriched with DEGs, including immune related pathways, cellular processes and environmental information processing pathways. Survival analysis demonstrated that 18 of them were closely related with non-recurrence rate along time after surgery. For ER negative patients, only three pathways including the folate biosynthesis pathway were enriched with DEGs and none of them overlapped with those of ER positive patients. Only the cholinergic synapse pathway was significantly associated with the non-recurrence rate according to the survival analysis.

Our findings shed light on pathways involved in breast cancer relapse with the hope to give some theoretical supports for further therapeutic study.