We present the first computational strategy leveraging genetic interactions successfully predicting patients response to many different targeted and immunotherapy treatments based on their tumors transcriptome. Sharing the details here- biorxiv.org/content/10.110… 
Current precision oncology is mainly performed by targeting actionable mutations & has a limited patient coverage. Here we present the first approach for systematically guiding patients treatment based on vulnerabilities identified across the whole exome.
We stratified responders in 17 out of 21 different targeted therapies cohorts with an AUC>0.7 across different cancer types better than previously published transcriptome-based signatures.
We stratified responders in 8 out of 11 different immunotherapy therapies cohorts (again AUC>0.7) across a range of cancer types, checkpoint targets and inhibitors.
Overall, through a retrospective analysis of 32 clinical trials of >1500 patients, treated with targeted and immune- therapies, our framework is predictive of response in 25 out of 32 cases (~80%) across 8 different cancer types and 13 therapies.
*We emphasize that the treatment outcome info was never used in learning the response predictions and a fixed parameter set was used, thus reducing the well-known risk of over-fitting and the lack of generalizability.*
Evaluating this approach retrospectively in a recent multi-arm basket clinical trial (WINTHER), we show that the fraction of patients benefitting from transcriptomic-based treatments could be markedly increased from 15% to 85% by targeting GI-based vulnerabilities in their tumors
Our genetic interaction score is also associated with the clinical response observed across different cancer types to checkpoint immunotherapies.
In summary, this is the first computational approach to obtain considerable predictive performance across many different targeted and immunotherapy datasets.
Combined with present precision oncology approaches, our strategy provides the first systematic framework for analyzing prospective transcriptomics-based clinical trials.
This work has been led by @joo_sang_lee & Eytan Ruppin at Cancer Data Science at @theNCI.
