Reformulating Reactivity Design for Data-Efficient Machine Learning

Machine learning (ML) can deliver rapid and accurate reaction barrier predictions for use in rational reactivity design. However, model training requires large data sets of typically thousands or tens of thousands of barriers that are very expensive to obtain computationally or experimentally. Furthermore, bespoke data sets are required for each region of interest in reaction space as models typically struggle to generalize. We have therefore reformulated the ML barrier prediction problem toward a much more data-efficient process: finding a reaction from a prespecified set with a desired target value. Our reformulation enables the rapid selection of reactions with purpose-specific activation barriers, for example, in the design of reactivity and selectivity in synthesis, catalyst design, toxicology, and covalent drug discovery, requiring just tens of accurately measured barriers. Importantly, our reformulation does not require generalization beyond the domain of the data set at hand, and we show excellent results for the highly toxicologically and synthetically relevant data sets of aza-Michael addition and transition-metal-catalyzed dihydrogen activation, typically requiring less than 20 accurately measured density functional theory (DFT) barriers. Even for incomplete data sets of E2 and SN2 reactions, with high numbers of missing barriers (74% and 56% respectively), our chosen ML search method still requires significantly fewer data points than the hundreds or thousands needed for more conventional uses of ML to predict activation barriers. Finally, we include a case study in which we use our process to guide the optimization of the dihydrogen activation catalyst. Our approach was able to identify a reaction within 1 kcal mol-1 of the target barrier by only having to run 12 DFT reaction barrier calculations, which illustrates the usage and real-world applicability of this reformulation for systems of high synthetic importance.

Fragmentation Method for Computing Quantum Mechanics and Molecular Mechanics Gradients for Force Matching: Validation with Hydration Free Energy Predictions Using Adaptive Force Matching

A fragmentation approach referred to as a simple overlapping region method for force matching (SORForM) is presented. SORForM is designed to enable efficient computation of quantum mechanical (QM) forces for large molecules and is validated in the framework of adaptive force matching (AFM) to develop solute models in water. The SORForM method divides a molecule into overlapping QM regions with each region containing a gradient zone and a buffer zone. The buffer zone ensures that the atoms in the gradient zone have their surroundings unchanged with fragmentation. The performance of the method is validated with mefenamic acid and linalyl acetate by comparing the hydration free energies of AFM models developed with and without SORForM. The AFM hydration free energies are also compared with that of the experiments. The models developed with B3LYP-D3(BJ) and def2-TZVP are in excellent agreement with experiments. Our work shows that PBE-D3(BJ) provides less satisfactory results when compared to B3LYP-D3(BJ). The def2-TZVP basis set is found to greatly improve the agreement with experiments when compared to a double-zeta quality basis set.