Discovery and Development of Novel Chemical Reactivity Through AI-Augmented Experimentation
Abstract
High-throughput experimentation (HTE) facilitates rapid discovery relative to analog experimentation. However, the combinatorial nature of parallel reaction screening leads to an exponential increase in the size of the candidate space. An exhaustive exploration of this space would be desirable but unrealistic, even using HTE. Research in the fields of computationally and AI-assisted chemistry has demonstrated that machine learning (ML, a subfield of AI that identifies patterns in data sets to extract rules for solving a given task) can be utilized to predict outcomes for a set of reaction conditions and select for the most informative experiments to perform. By leveraging AI, the discovery process can be reformulated to avoid the need for an exhaustive search. This strategy reduces the number of needed experiments and increases the rate of discovery. The augmentation of HTE with an active, AI-led approach to exploring chemical space lends itself to the development of a self-driving laboratory, in which a closed feedback loop process where information is shared between experiment planner and automated platform enables the AI to develop more refined hypotheses about subsequent experiments.. Altogether, we expect that this AI-augmented HTE strategy would (1) reduce the capital needed for scientific discovery, (2) facilitate an abbreviated but efficacious exploration of a broad swath of chemical space, and (3) guide the rapid identification and development of optimal reaction conditions after an initial hit. As such, we anticipate that this powerful approach to reaction development will greatly accelerate the discovery of novel reactivity and subsequent reaction optimization. The outlined AI-assisted reaction discovery strategy towards synthetic methodology development will aim to address the broad range of highly challenging chemical problems outlined in Aims 1-4. As such, this research program will target the AI-assisted development of classically difficult but valuable chemical trans-formations, including, but not limited to, polyolefin depolymerization, molecular editing, atomswapping towards heterocyclic scaffolds, nitrogenative ring expansion, and strategies for cross-native CC bond formation. The discovery of the proposed reactions and catalytic structures outlined in Aims 1-4 could facilitate: 1) controlled depolymerization of persistent and ubiquitous polyolefin plastics to generate chemical feedstocks, 2) theconstruction of valuable molecular scaffolds from native and ubiquitous functionalities, and 3) the discovery of new catalytic architectures capable of facilitating challenging and energy-intensive chemical transformations.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 06, 2021
- Source ID
- N000142112138
Entities
People
- David MacMillan
Organizations
- Office of Naval Research
- Trustees of Princeton University
- United States Navy