Molecular Discovery Pipelines

Molecular Discovery Pipelines

PI Suhas Mahesh (Schmidt Science Fellow, University of Toronto)
Grad student: Alexander Davis

Traditional methods for molecular discovery rely heavily on trial-and-error experimentation. Some types of molecules have proven particularly resistant to traditional discovery, with one notable example being stable blue light-emitting organic molecules. The discovery of such molecules would drastically improve the lifespan of displays, LED lighting, and lasers. One approach to accelerating discovery is a virtual screen, in which expensive and slow trial-and-error experimentation is replaced with cheap and fast trial-and-error simulation.

Following similar successful approaches in drug discovery and protein engineering, this project develops a rigorous and configurable software framework to marshal cluster compute resources and automate a funnel strategy for virtual screening, in order to simulate millions of candidates without sacrificing the accuracy of the final predictions. The approach combines high-speed models (machine learning, semi-empirical methods) with high-accuracy models (density functional theory) in an automated pipeline. The vision of the framework is to enable screening of millions of molecules per week and manage an evolving corpus of results data over multiple screening iterations, while minimizing downtime required for modifications or data analysis. The virtual screening approach significantly reduces the time and resources required for molecular discovery so that researchers can focus on the most promising leads.

The software is be based on the SSEC dplutils framework and the PI’s vision of the simulation and screening pipelines, and designed to be shared with the broader research community and easily adaptable for use in similar high-throughput experimentation settings. By accelerating the discovery of light-emitting molecules, we can contribute to the development of more efficient, longer-lasting displays, lighting, and lasers. More broadly, our software empowers researchers by providing a convenient and flexible screening workflow for materials discovery.