Grant: Discovery Grants Program - Individual

In the forthcoming proposal, I will highlight recent developments in the field of electronic structure theory and molecular structure optimization using computational modelling from my group. The proposal will detail how we plan to build on this foundation using novel electron pair methodologies and modern machine learning approaches. Specifically, we seek to extend our so--called Single Electron Pair Distribution Analysis Model (SEPDA) to a broader swath of relevant applications, including as a quantitative analysis tool in Physical Organic Chemistry. In our SEPDA modelling, we have the unique ability to isolate specific Lewis pairs within a molecular structure and produce the quantum mechanical electron pair distribution of said Lewis pair. By employing computational analysis techniques to probe both position and momentum space, we produce an information rich distribution that we have shown to be quantitative in predicting a myriad of useful properties including bond strengths and even related pKas. The horizon of our unique modelling techniques include a broad swath of applications such as the analysis of reaction mechanism, the rapid prediction of localized molecular properties, the characterization of molecular and electronic structure, to name just a few. In parallel, we also plan on forging new applications of our novel TSNet machine learning technology. Previously, we have shown that using a unique blend of transfer learning with Siamese deep neural networks, one has the ability to accurately predict transition state structures from the knowledge of just a reactant and product complex. This technique has a wide array of valuable near--term applications in reaction modelling and optimization. In particular, we will propose the extension of our techniques to applications of biochemical reaction mechanisms. Since transition states are critical to understanding (and intervening in) protein function, rapid prediction of complex transition states using novel machine learning technologies will afford a major advance in biochemical modelling for drug discovery, among other important verticals.