Geetesh Devineni, Graduate Student, Kostal Lab, GW Department of Chemistry
The many applications of in silico modeling: from polymer chemistry to predicting chemical behavior in the environment to designing safer, biobased agrochemicals
There is growing need to move towards a more sustainable economy to reduce the negative impact of chemical production on both human and environmental health. Presented here are several projects that leverage in silico methods for the advancement of this goal. The bioconcentration factor (BCF) is a key metric in evaluating sustainability and potential toxicity in the aquatic environment; however, experimental determinations of BCF are expensive and time-consuming. Instead, Quantitative Structure Property Relationships (QSPR’s) can be developed to predict BCF using mechanistically relevant physiochemical properties. Appropriately validated, these models can supplement or, pending regulatory approval, replace economically and ethically unfeasible animal testing. Our approach to BCF prediction, as outlined here, improves on existing models by considering ionization of xenobiotics, which affects both bioavailability and transport to storage sites, and by gauging relevant properties from explicit Monte Carlo simulations in the condensed phase. In a different project, we show how machine learning can be leveraged to develop a workflow for iterative, virtual synthesis of novel (greener) pesticides from renewable building blocks. Bio-based building blocks such as fumaric acid, succinic acid, lysine, or ethanol, were mined from literature and combinatorially reacted to form successive generations of products (i.e., a virtual library of tens of thousands of compounds). Mimicking computer-aided drug discovery, existing insecticides were then used to screen the virtual library for potential leads based on mode of action and relevant physicochemical/thermodynamic properties. Identified leads are subsequently optimized using free energy perturbation calculations and validated experimentally to propose new agrochemicals. Uniquely, our workflow allows for the quantification and reconciliation of the intrinsic underlying tradeoffs between selective toxicity (i.e., performance), unintended toxicity and environmental persistence, a feat that is much harder to accomplish when redesigning existing (petroleum-based) pesticides. Lastly, we present on a collaboration with the Boyes’ polymer chemistry group to explore reactivity differences across different monomers, which have promising applications in well defined, functional, aromatic polyamides. Our studies provide mechanistic understanding of experimental outcomes, with the goal to guide design of novel building blocks with desirable characteristics.
