Multiobjective Feature Selection for Identifying MicroRNA Markers

MicroRNAs (miRNAs) play a pivotal role in gene regulation, inhibiting mRNA translation and emerging as potential biomarkers for diseases, particularly cancer. Analyzing miRNA expression confronts challenges due to their ability to target multiple mRNAs and identical sequences. The intricate link between miRNA expression and cancer adds complexity. However, the smaller size of miRNAs proves advantageous in biomarker discovery for cancer. This chapter introduces a multiobjective clustering approach using a genetic algorithm for miRNA microarray data analysis. Employing non-dominated sorting and the Crowding Distance measure, the genetic algorithm identifies a concise set of nonredundant miRNA markers. Objectives encompass optimizing clustering validity indices, determining encoded miRNA markers, and achieving a high classification performance. By clustering miRNAs and selecting central features, the method ensures nonredundant marker inclusion, showcasing clear advantages over comparative methods and contributing significantly to cancer biomarker discovery. The study enhances insights into miRNA expression data analysis, presenting a refined approach for robust identification of cancer-associated miRNA markers.