Evaluating the incorporation of Biological Knowledge in multiobjective clustering of gene expression data

Analysing large amounts of gene expression data from biological repositories can uncover new patterns, potentially leading to innovative treatments or diagnostic methods for diseases such as cancer, Alzheimer and Parkinson. In this work, we evaluate the incorporation of two biological knowledge’s data on the multi-objective clustering for cancer gene expression data. We implemented a Non-dominated Sorting Genetic Algorithm II (NSGA-II) to optimise two objectives: gene expression similarity and biological coherence, using Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) as biological knowledge sources. The results show a better performance of the multiobjective clustering when it uses the KEGG knowledge database. It provides evidence that KEGG is a worthy alternative to GO in helping clustering algorithms to find meaningful clusters. These results contribute to creating better computational tools for high-quality analysis in cancer genomics, potentially enhancing the precision of treatment strategies.