In a promising step for cancer precision medicine, we devised a novel strategy that identifies genetic interactions and robustly predicts drug response, spanning 21 targeted therapies and 11 immunotherapy datasets across 10 different cancer types (1/9) bit.ly/38Pzn9m 
To predict treatment outcomes, we defined genetic interaction scores, which capture the likelihood of a given patient responding to a given drug. (2/9)
We predicted treatment response in 17 of the 21 targeted- and 8 of the 11 immuno- therapy datasets, with considerable accuracy (AUC > 0.7). (3/9)
Intriguingly, we found key immune genes such as IFNAR2, IFNGR1, and B2M to be interacting partners of PD1 and CTLA4, and their interactions *effectively* predict response across immune checkpoint therapy datasets. (4/9)
Our exciting results led us to ask a compelling question: how many patients would have likely benefited from prescribed drugs if the treatment choices for patients would have been guided by our computational approach? (5/9)
Strikingly, we found that *70% of patients* in the WINTHER cohort could have likely become responders if they had been matched to therapies guided by our approach. *94% of patients* could have been matched to more effective therapies overall. (6/9)
Notably, the most frequently recommended drugs by our approach in the WINTHER cohort were dasatinib, venetoclax, and cobimetinib. (7/9)
Our approach's robust success highlights the potential for *major improvement* in transcriptomics-guided cancer treatment. We are excited by the possibilities this opens up in the world of precision medicine. (8/9)
Kudos to @joo_sang_lee for leading this work and to all the authors! (9/9)
