Abstract 6894: Predict, perturb, and process: A systems biology pipeline for identifying synthetic lethality.

Abstract The synthetic lethality principle holds great promise for cancer therapy, but identifying synthetic lethals presents enormous challenges. Synthetic lethality, where loss of two genes causes cell death, but loss of either gene alone does not, is difficult to assay because combinatorial perturbation technologies are inefficient and because the search space is well beyond the capacity of current experimental systems: 19,000 protein coding genes offer more than 180 million candidate gene pairs. To address this, we combine network-driven predictive models with the In4mer CRISPR/Cas12a combinatorial knockout platform to conduct tractable pairwise knockout screens with high probability of identifying synthetic lethals.Even with these advances, effective methods for quantifying GIs in these screens remain limited. To address this gap, we developed GRAPE (Genetic interaction Regression Analysis of Pairwise Effects), a novel regression-based method for analyzing GIs in all-by-all gene knockout (KO) library designs. GRAPE infers single-gene KO fitness from multiplex CRISPR arrays, predicts combinatorial gene KO fitness, and identifies synthetic lethals that deviate from this expectation. Since no gold-standard exists for benchmarking, we built a simulation framework to systematically evaluate GRAPE, demonstrating improved computational efficiency, adaptability, and precision-recall performance over existing methods.Using the network-driven experimental design, In4mer screening, and GRAPE analysis, we screened all pairwise combinations of 206 genes involved in receptor tyrosine kinase (RTK) signaling, all pairs of 167 genes in DNA damage response (DDR) pathways, and over 4,000 paralog pairs across 12 diverse cancer cell lines. Our DDR results closely align with findings reported by other studies, while our RTK network provides novel insight into ER-mediated protein modification and oncogenic signaling dependencies. Overall, our efforts confirm that we can predict and detect both global and background-specific genetic interactions, advancing the state of the art in functional genomics and cancer target finding. Citation Format: Juihsuan Chou, Chenchu Lin, Iulia Veronica Gheorghe, Sabriyeh Alibai, Subin Kim, Lori Wilson, Xingdi Ma, Junjie Chen, Glen Traver Hart. Predict, perturb, and process: A systems biology pipeline for identifying synthetic lethality abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 6894.