Most candidates for new cancer drugs ultimately prove to be disappointing — because of a lack of efficacy, the emergence of drug resistance, or side effects. This high failure rate illustrates the need for better approaches for identifying the most promising therapeutic contenders and predicting which ones are most likely to be successful in the clinic. Microscopy-based screens, which employ staining and imaging of cells, are an effective approach in screening for potential cancer drugs. With current technologies, however, it is difficult to identify exactly which molecules in a cancer cell the most promising compounds are targeting. In this project, researchers at the University of California, San Francisco in the laboratory of Professors Steven Altschuler and Lani Wu are developing a new platform called PhenoTIP (for Phenotypic Target Identification Platform) that identifies novel cancer drug candidates and narrows in on their mode of action. This platform combines the power and scalability of genome-wide screens with the observation of responses at the single-cell level. Machine learning is used to analyze cellular images and identify the most promising compounds and their targets.
This project aims to first establish PhenoTIP as a valuable approach for cancer drug discovery. Preliminary studies have suggested that this method can successfully combine in situ sequencing and phenotypic profiling. The researchers are also using CRISPR-Cas9 to develop reagents and cell lines as well as to match single-cell phenotypes with specific genetic alterations. To establish the platform’s efficacy in measuring dose-response, morphological chemical-genetic interactions, and more, the team is evaluating PhenoTIP’s performance on compounds with both known and unknown targets. When completed, this project will support broader deployment of PhenoTip by demonstrating its efficacy and usefulness in overcoming the challenge of deconvoluting phenotypic screens.