Technology Development for High-Throughput Functional Genomics
Customized genome-scale gene perturbation libraries: Much of the work we do utilizes genetic screens enabled by novel high-coverage CRISPR/Cas9 libraries (10 sgRNAs/gene) we have developed. The high coverage greatly reduces false positive and false negative results. Our platform allows for easy creation of new library designs, and we use a pooled format that can be used to rapidly screen genome-scale libraries in a few hours to a few weeks, depending on the design. Libraries can then be analyzed by deep sequencing to quantify changes in sgRNA abundance.
Directed evolution using dCas9-targeted somatic hypermutation (CRISPR-X): To investigate protein variants and aid in the identification of new functional mutations, we developed a strategy to re-purpose the somatic hypermutation machinery used in antibody affinity maturation to create targeted, diverse populations of point mutations. Using dCas9 to recruit a hyperactive variant of the deaminase AID, we can create diverse point mutations within an ~100bp window centered on the sgRNA PAM site. These mutant populations can then be subjected to selection to evolve proteins with improved function or to map the sites of drug-protein interactions. For example, by tiling mutations across PSMB5, we could map known and novel mutations that affect binding to the chemotherapeutic bortezomib.
Systematic genetic interaction maps: We have also developed strategies to systematically knock down/knock out pairs of genes. This has facilitated some of the first systematic genetic interaction maps in mammalian cells. Using these maps, we can understand coordinated gene functions and predict new functions for uncharacterized genes. They also allow us to quickly identify synthetic lethal genetic interactions specific to cancer that we have targeted with drug combinations.
Systematic functional measurements of protein domain activities: Working with the Bintu lab, we have recently developed a broadly applicable strategy to use our oligonucleotide-based library platform to create complex libraries encoding short protein domains. This has enabled measurement of transcriptional activation and repression activities for thousands of protein domains, annotation of domains which were previously uncharacterized, and the de novo mapping of functional domains within proteins by ‘tiling’ across entire proteins. We have extended this work to study the activity of transcriptional regulators in diverse genomic contexts, and more recently to conduct deep mutational scanning studies of protein domains to enable deep learning models that can predict repressor function and design variants with specified activities (collaboration with Kundaje lab).