Multi-scale approach for identification of context-specific fitness genes and mechanisms from genome-wide CRISPR/Cas9 screens

周治瑗 Chih-Yuan Chou1, 李容羽 Jung-Yu Lee1 , 楊進木 Jinn-Moon Yang1

1Institute of Bioinformatics and Systems Biology, National Yang Ming Chiao Tung University

Genome-wide CRISPR/Cas9 screens provide insight into potential diagnostic and therapeutic targets. It is critical to distinguish core and context-specific fitness genes for reducing cellular toxicity and  identifying novel disease-specific biomarkers. However, most approaches require large numbers of samples, and the identification and interpretation of both fitness genes is still incomplete. By integrating transcriptomic expressions and protein-protein interactions, we developed a binary classification model to predict that whether a gene is a context-specific gene or not through single gene- and community-based biological regulatory pathways and biological processes regardless of sample size. We predicted POLR2B, SNRPC, and PSMD12 as core fitness genes involving RNA polymerase, spliceosome, and proteasome respectively, which are related to genetic information processing such as transcription, protein folding, sorting and degradation. We identified breast cancer-specific fitness genes that regulate pathways such as D-Arginine and D-ornithine metabolism, Hippo signaling pathway, and MET receptor activation. Meanwhile, we calculated the contributions of pathway features to the classification of core or breast cancer-specific fitness genes to prioritize the dependencies of regulated pathways. As a result, we identified gene A as a breast cancer-specific fitness gene and it was involved in as D-Arginine and D-ornithine metabolism, leading to decreased survival and reduction of anti-tumor activity. Therefore, our model was able to predict and illustrate disease-specific biomarkers and mechanisms involved, which is a powerful strategy for unveiling more potential therapeutics.