(8d) A Big Data Approach to Predict the Kinetics of Organic Chemistry Reactions

Computing reaction rate constants ab initio quickly becomes unfeasible as the system size grows. In this talk we will discuss our ongoing work which aims to explore and employ machine learning algorithms to predict reactivity in the ground state. Our current efforts focus on creating a large reaction rate constant dataset for organic chemistry reactions to train supervised machine learning algorithms to predict quantum rate constants. The dataset is generated by extending our previous model dataset of over 1.5 million examples [1] to a set of known classes of chemical reactions such as e.g. nuclear substitution. Furthermore, dataset examples now include reactant partition functions computed using the rigid body and harmonic oscillator approximation at the DFT level of theory. With this dataset we are currently training ML algorithms to predict quantum reaction rate constants at various temperatures. In the future, we plan to extend this work further to include features to describe reactivity on catalytic surfaces.

[1] E. Komp and S. Valleau, “Machine Learning Quantum Reaction Rate Constants,” J. Phys. Chem. A, 124:8607-8613, 2020.