Uncovering cancer dependencies in peptide-interacting protein pockets.

Cancer cells often become dependent on specific molecular functions. As many proteins perform multiple functions mediated by different pockets and interfaces, we hypothesized that we could identify distinct cancer dependencies and therapeutic vulnerabilities by disrupting peptide-binding pockets. To test this hypothesis, we screened a proteome-wide library of 7152 genetically encoded peptides across nine cancer cell lines. We identify common and selective dependencies on peptide-binding pockets and find that gene knockout and peptide-mediated inhibition of pockets often drive divergent phenotypes. For the common-essential gene HCF1, we identify a therapeutic window by using inhibitory peptides with varying affinity. Moreover, peptides targeting TLE1-4 reveal a dependency hidden in genetic screens by homolog redundancy. We also uncover that peptides inhibiting cyclin D drive specific suppression of leukemia cell proliferation and demonstrate that these peptides improve the potency of CDK4/6 inhibitors. Overall, our screening platform facilitates data-driven prioritization of molecular pockets for subsequent therapeutic translation.