Inferring Genetic Co-dependencies To Identify New Vulnerabilities In Cancer
- Author(s): Matthew, Thomas James
- Advisor(s): Stuart, Joshua
- et al.
Translation of cancer genomic data into cancer therapies and companion
diagnostics remains a primary challenge in personalized medicine. Much of this
challenge is due to the difficulty of identifying genetic co-dependencies that lead to
clinically actionable drug targets. Targeting many of the known essential gene
products are not always selectively efficacious because these targets may be
common to both malignant and benign cells. However, essential genes that are
associated with particular genomic alterations in cancer cells, like those from
synthetic lethality, can potentially provide a source of tumor-specific drug targets.
To help aid novel drug discovery, I developed a computational approach called
CLOvE, a multi-omic approach that identifies co-dependencies in pairs of genes.
These co-dependencies are inferred from context-dependent changes in
expression, where CLOvE assigns high scores to those genes with the greatest
compensatory change in expression. These scores may suggest synthetic lethal
interactions, which may uncover clinically actionable essential genes. These
methods were developed in CCLE cell lines and validated with RNAi and CRISPR
viability data. CLOvE identifies meaningful expression changes, assigns high scores
to known essentials, reveals known synthetic lethal connections, and implicates
many possible new connections. This approach could provide a tool to accelerate
both target discovery and biomarker discovery, to develop drugs suitable for a
specific cancer, and identify and stratify patients who may benefit from these