Abstract

Continued improvements in computer hardware and algorithms have allowed digital chemistry to be employed for studying more complex problems. However, performing the various computational protocols required for high-throughput screening, molecular optimization, and machine learning (ML) model training can be tedious and time-consuming. Automating these workflows minimizes errors, enhances reproducibility, and facilitates storing and reusing the associated data.

In this talk, we present AQME, an automated end-to-end workflow software that performs multi-step tasks of computational chemistry, and ROBERT, a program that automates ML-related tasks such as data curation, hyperparameter optimization, and ML predictor generation. The combination of these programs allows for high-quality reaction energy profile studies and ML predictions to be generated with just a few command lines, starting from simple inputs such as databases with SMILES strings.