Using Genetic Algorithms and Sparse Logistic Regression to Find Gene Signatures for Chemosensitivity Prediction in Breast Cancer

Abstract

Various gene signatures of chemosensitivity in breast cancer have been identified. When used to build predictors of have chemosensitivity, many of them have their prediction accuracy around 80%. Identifying gene signatures to build high accuracy such predictors is a prerequisite for their clinical tests and applications. To elucidate the importance of each individual gene in a signature is another pressing need before such signature could be tested in clinical settings. In this study, Genetic Algorithms (GAs) and Sparse Logistic Regression (SLR) were employed to identify two signatures. The first had 28 probe sets selected by GA from the top 65 probe sets that were highly overexpressed between pathologic compete response (pCR) and residual disease (RD) and was used to build a SLR predictor of pCR (SLR-28). The second had 86 probe sets (Notch-86) selected by GA from Notch signaling pathway and was used to develop a SLR predictor of pCR (SLR-Notch-86). These two predictors tested on a training set (n=81) and validation set (n=52) had very precise predictions measured by accuracy, specificity, sensitivity, positive predictive value and negative predictive value with their corresponding P value all zero. Furthermore, these two predictors discovered 12 important genes in the 28 probe set signature and 14 important genes in the Notch-86 signature. Our two signatures produced superior performance over a signature in a previous study, demonstrating the potential of GA and SLR in identifying robust gene signatures in chemo response prediction in breast cancer.