Prediction and Analysis of Key Genes in Prostate Cancer via MRMR Enhanced Similarity Preserving Criteria and Pathway Enrichment Methods

Informative gene discovery has grown in interest in the field of biomedical research with the advent of high throughput technologies. The filter based methods including the similarity based ReliefF, Fisher Score and Laplace Score criteria have been widely used for gene studies. However, the similarity preserving methods are incapable of filtering redundant features. Recently, the mRMR algorithm has been proposed to maximize relevance and minimize redundancy of gene selection but is computationally expensive. In this study, we enhanced the similarity based methods with mRMR to find representative genes to improve prediction performance and identify essential genes. The selected genes are further analysed to extract enriched Gene Ontology annotations and KEGG pathways to determine key genes and their bio-molecular mechanisms. A screened genes showed significant enrichment in the annotations and signaling pathways related to fatty acid biosynthesis, G-quadruplex DNA/RNA binding and spermatid development. The findings revealed the bio-molecular processes and potential biomarker genes to target for diagnostic and therapeutic treatment of prostate cancer.