Predicting gene essentiality in Caenorhabditis elegans by feature engineering and machine-learning

• DOI: 10.1016/j.csbj.2020.05.008
• Publication Date: 2020-05-15T21:32:11Z
• Authors: Tulio L. Campos; Pasi K. Korhonen; Paul W. Sternberg; Robin B. Gasser; Neil D. Young
• Keywords: Model Organism

Defining genes that are essential for life has major implications understanding critical biological processes and mechanisms. Although have been identified characterised experimentally using functional genomic tools, it is challenging to predict with confidence such from molecular phenomic data sets computational methods. Using extensive available the model organism Caenorhabditis elegans, we constructed here a machine-learning (ML)-based workflow prediction of on genome-wide scale. We strong predictors showed trained ML models consistently achieve highly-accurate classifications. Complementary analyses revealed an association between chromosomal location. Our findings reveal in C. elegans tend be located or near centre autosomal chromosomes; positively correlated low single nucleotide polymorphim (SNP) densities epigenetic markers promoter regions; involved protein processing; transcribed most cells; enriched reproductive tissues targets small RNAs bound argonaut CSR-1. Based these results, hypothesise interplay RNA pathways germline, transcription-based memory; this hypothesis warrants testing. From technical perspective, further work needed evaluate whether present ML-based approach will applicable other metazoans (including Drosophila melanogaster) which comprehensive (i.e. genomic, transcriptomic, proteomic, variomic, phenomic) available.