A hybrid ensemble feature selection design for candidate biomarkers discovery from transcriptome profiles

Abstract

Discovering disease biomarkers from gene expression data has been greatly advanced by feature selection (FS) methods, especially using ensemble FS (EFS) strategies with perturbation at the data (i.e., homogeneous EFS) or method level (i.e., heterogeneous EFS). Here, we proposed a hybrid EFS design that explores both types of perturbation to disrupt associations of good performance with a single dataset, single algorithm, or a specific combination of both, which is particularly interesting for better reproducibility of genomic biomarkers. We investigated the adequacy of our approach for microarray data in four types of cancer, extensively comparing it with other ensemble and single FS approaches. Five FS algorithms were analyzed: Wx, Symmetrical Uncertainty, Gain Ratio, Characteristic Direction, and ReliefF. We observed that, across distinct datasets, the hybrid and heterogeneous approaches attenuated the large performance variation of most methods without function perturbation. Additionally, the proposed hybrid EFS has superior performance to the heterogeneous EFS. Interestingly, the ranks produced by our method reached greater biological plausibility, with a notably high enrichment for cancer-related genes and pathways. Thus, our experiments suggest the potential of the proposed hybrid EFS design for discovering candidate biomarkers from microarray data. Finally, we provide an open-source framework to support similar analyses in this and other domains, being available as a user-friendly application and a programmable Python package.