Block HSIC Lasso: model-free biomarker detection for ultra-high dimensional data

• DOI: 10.1093/bioinformatics/btz333
• Publication Date: 2019-05-09T11:20:03Z
• Authors: Héctor Climente-G...; Chloé-Agathe Azen...; Samuel Kaski; Makoto Yamada
• Keywords: Lasso; Python; Biological data

Abstract Motivation Finding non-linear relationships between biomolecules and a biological outcome is computationally expensive statistically challenging. Existing methods have important drawbacks, including among others lack of parsimony, non-convexity computational overhead. Here we propose block HSIC Lasso, feature selector that does not present the previous drawbacks. Results We compare Lasso to other state-of-the-art selection techniques in both synthetic real data, experiments over three common types genomic data: gene-expression microarrays, single-cell RNA sequencing genome-wide association studies. In all cases, observe features selected by retain more information about underlying biology than those techniques. As proof concept, applied experiment on mouse hippocampus. discovered many genes linked past brain development function are involved differences neurons. Availability implementation Block implemented Python 2/3 package pyHSICLasso, available PyPI. Source code GitHub (https://github.com/riken-aip/pyHSICLasso). Supplementary data at Bioinformatics online.