PgmNr Y3161: Unbiased functional annotation of compound libraries using yeast chemical genomics.

Authors:
Sheena C. Li 1 ; Scott W. Simpkins 2 ; Justin Nelson 2 ; Jeff S. Piotrowski 3 ; Hamid Safizadeh 2 ; Karen Kubo 4 ; Nikko Torres 6 ; Grant Brown 6 ; Yoshikazu Ohya 4 ; Ming-Wei Wang 5 ; Minoru Yoshida 1 ; Chad L. Myers 2 ; Charles M. Boone 1,5

Institutes
1) RIKEN Center for Sustainable Resource Science, Wako, Saitama, JP; 2) Computer Science and Engineering, University of Minnesota, MN, USA; 3) Yumanity Therapeutics, Cambridge, MA, USA; 4) University of Tokyo, Kashiwa, Chiba, Japan; 5) The National Center for Drug Screening, Shanghai, China; 6) Donnelly Centre for Cellular and Biomolecular Research, University of Toronto, ON, Canada.

Keyword: Human diseases/Drug Discovery
Abstract:

To elucidate the molecular mechanism of bioactive compounds in an unbiased manner, we used chemical genomics in budding yeast to systematically screen very large and diverse compound collections, with the goal of linking unknown compounds to their cellular targets.

A high-throughput pipeline using various barcoded yeast mutant collections in a drug hypersensitive genetic background was developed. We designed a diagnostic pool of 310 non-essential deletion mutants to minimize the amount of compound required for screening and to maximize dynamic range. In addition, we constructed temperature-sensitive and heterozygous diploid mutant collections for all essential genes to obtain higher resolution chemical genetic information. Chemical genetic signatures for thousands of compounds were efficiently and rapidly generated via highly multiplexed next generation sequencing of DNA barcodes in the mutant strains. Targets were predicted using two strategies: For the diagnostic non-essential pool and the essential temperature-sensitive collection, we compared chemical genetic signatures with genome-wide synthetic lethal data to predict compound functionality at the biological process level. We complemented this analysis with a drug-induced haploinsufficiency approach to determine precise gene targets using the heterozygous diploid collection.

We applied this system to screen eight diverse compound libraries. Using the diagnostic pool approach, we assessed over 18,000 compounds for target specificity, and identified high confidence target process predictions for ~2000 unique compounds. Functional diversity of different compound libraries was characterized by mapping predicted targets onto the global yeast genetic interaction network. We then used the set of compounds with high confidence predictions for further chemical genetic screens with temperature-sensitive and heterozygous diploid collections for essential genes. Validation studies have confirmed predictions at the biological process level and also at the specific gene target level, including compounds with multiple predicted modes of action.