Predicting Pathologic Complete Response to neoadjuvant chemotherapy in breast cancer using Sparse Logistic Regression

Abstract

We utilised Sparse Logistic Regression (SLR) to build two sparse and interpretable predictors. The first one (SLR-65) was based on a signature consisting of the top 65 probe sets (59 genes) differentially expressed between Pathologic Complete Response (PCR) and Residual Disease (RD) cases, and the second one (SLR-Notch) was based on the genes involved in the Notch singling related pathways (113 genes). The two predictors produced better predictions than the predictor in a previous study. The SLR-65 selected 16 informative genes and the SLR-Notch selected 12 informative genes.