Data-Science Guided Study and Design of Phosphoranyl Radical-Mediated Reactions

The discovery of new pharmaceutical compounds relies on the ability to synthesize increasingly complex scaffolds with desirable biological and pharmacokinetic properties. To foster continued growth in this area, new transformations to access small organic molecules are required. Methods which form new C–N bonds are highly valuable, as over 50% of the top two hundred drug compounds contain at least one N-atom. Recently, the Doyle group has demonstrated that sulfonamides can be activated for hydroamination reactions using photoredox/phosphine catalysis through α-scission of an intermediate phosphoranyl radical. While phosphoranyl radicals have been used to enable a variety of transformations through β-scission, transformations proceeding through α-scission are underdeveloped despite providing an approach to form C–N or C–heteroatom bonds. To promote the development of new transformations relying on phosphoranyl radicals, a better understanding of the factors controlling α- versus β-scission is required. We propose leveraging data science to build a comprehensive, predictive model explaining phosphoranyl radical α- versus β-scission selectivity. This model will employ data collected from a model reaction to train predictive machine learning algorithms, overcoming limitations in our current mechanistic understanding of this process. In addition, we will leverage phosphine activation of N-nucleophiles for a novel alkene carboamination reaction based on a photoredox, phosphine, and nickel tandem catalysis. This unique approach relies on the ability of phosphine catalysis to form N-centered radicals and the established reactivity of nickel catalysis towards C-centered radicals to enable a desirable carboamination reaction without the requirement of pre-functionalized amines or stoichiometric reagents. In addition, this approach is amenable to carboamination with N-heterocycles. Overall, the proposed research is comprised of two aims: (1) develop predictive machine learning algorithms to determine factors influencing the selectivity of α- and β-scission of phosphoranyl radicals, and (2) develop a photoredox, phosphine, and nickel catalysis mediated carboamination reaction. These aims will be explored concomitantly with data collected in aim 1 being applicable to, but not necessary for, the completion of aim 2. Aim 1 will contribute significantly to our understanding of phosphoranyl radical reactivity, promoting the design and development of new transformations relying on this powerful mechanistic step. The transformation developed in aim 2 will address challenges in currently available carboamination reactions to generate desirable C–N and C–C bonds and will provide a framework for future reactions combining photoredox, phosphine and nickel catalysis. Overall, this work represents a significant advance in the field of phosphoranyl radical reactivity and is expected to have longstanding impacts on future applications of these approaches to pharmaceutically relevant transformations.