ROBERT: Automated Chemical Machine Learning Protocols

David Dalmau, Juan V. Alegre-Requena

Despite the abundance of machine learning toolkits, there is a significant gap in implementing machine learning (ML) protocols within the computational and experimental chemistry communities. This gap hinders the widespread adoption of ML techniques as standard practices, leaving many researchers in the field of chemistry lagging behind. In response to this challenge, we introduce ROBERT, [1] a program designed to automate chemical ML workflows. This tool performs commonly used protocols including data curation, hyperparameter optimization, and generating ML predictors (Fig. 1). It allows users to obtain high-quality ML predictions through either a single command line or user-friendly graphical interface. ROBERT enables chemists to achieve results comparable to those of experts in the field, while upholding the current gold standard of reproducibility and transparency. [2]

Figure 1. Overview of protocols automated by ROBERT.

1. Dalmau, D.; Alegre Requena, J. V., ChemRxiv 2023, DOI: 10.26434/chemrxiv-2023-k994h.
2. Walsh, I.; Fishman, D., Nat. Methods 2021, 18, 1122–1127.

 David Dalmau