PROJECT SUMMARY/ABSTRACT
Genes and variants often act in combination to drive cellular and organismal phenotypes. Mapping these
functional interactions advances our fundamental understanding of biological systems and has broad
applicability to therapeutics development. Gene-gene interactions also likely constitute a considerable
component of the undiscovered genetics underlying human diseases, due to the extensive buffering encoded
in genomes which makes many individual genes appear dispensable. In this regard, we and others have
shown that combinatorial screens, such as those based on CRISPR-Cas systems, are powerful platforms for
mapping synergistic relationships among genes and variants. However, unlike screens based on single-gene
perturbations which are broadly utilized, combinatorial screens have been significantly harder to deploy. Two
fundamental challenges underlying combinatorial CRISPR screens are: 1) the requirement to physically link
multiple perturbagens on the same library element which, in addition to complicating library generation,
prevents different classes of genome and epigenome engineering toolsets from being readily combined; and 2)
analysis of the resulting combinatorial screening data is highly complex, especially in the context of multi-
dimensional phenotypic assays. Furthermore, because the perturbation space scales exponentially with the
number of simultaneous perturbagens, it is critical to be able to computationally infer interactions beyond those
measured experimentally. To address these challenges, we propose to engineer a new screening platform,
CombinX, that auto-tethers individual library elements expressed at the RNA, instead of the DNA, level to
enable massively multiplexed combinatorial screens. Scalability of this platform is thus limited only by cell
culture and sequencing power. We propose to develop the system for both two-way and multi-way (>2
perturbagen) combinatorial screens via application to genetic interaction mapping and cellular reprogramming
respectively. Resulting screening data will be interpreted via new advanced computational methods and
machine learning approaches to systematically determine genetic interactions, as well as to predict interactions
well beyond those that can be covered by direct experimental screens. We anticipate this experimental and
computational platform will have broad applicability in basic science and therapeutics discovery, and that it will
generate widely useful reagents and data.