Mapping and Targeting Genetic and Physical Interactions at Scale

Biological phenotypes are mediated by a network of functional interactions between genes, proteins, and other biomolecules present in the cell. While high-throughput screening efforts have largely mapped the role of individual genes in controlling phenotypes such as cellular proliferation, interactions between genes/proteins remain largely unmapped and untargeted. In this dissertation, we develop and apply novel screening methodologies to map and exploit interactions between genes/proteins. We use pairwise CRISPR-Cas9 mediated gene knockouts to map the full set of genetic interactions among cyclin-dependent kinases (CDKs) and interacting proteins, identifying several synthetic-lethal and synergistic relationships. We perform single-cell RNA sequencing on the CDK knockout populations, quantifying the cell-cycle effects and cell states mediated by individual CDK proteins. While CDKs are readily druggable via small molecules, many cancer drivers have structures which are not amenable to traditional pharmacological inhibition approaches. To address this challenge, we developed a peptide tiling (PepTile) approach to engineer protein inhibitors of cancer drivers and protein-protein interactions in general. By overexpressing pooled libraries of peptides within cancer cells, we map bioactive protein domains and identify peptides derived from key protein-protein interaction (PPI) interfaces which have strong anti-proliferative effects. We show that these peptides can be modified for extracellular delivery, functioning as anticancer drugs with micromolar IC50s. Finally, we demonstrated the versatility of the PepTile approach to alternative contexts, mining physical interactions to improve delivery of therapeutic payloads in vivo. We show our screening datasets can be used to train predictive models, with applications for future engineering efforts towards targeting and delivery of therapeutic biomolecules