Compressing Chemistry Reveals Functional Groups

TL;DR

This paper presents a novel unsupervised learning approach based on the Minimum Message Length principle to identify and evaluate chemical functional groups, improving bioactivity prediction accuracy.

Contribution

It introduces a formal, large-scale assessment method for chemical functional groups using data compression principles, revealing both known and novel functional groups.

Findings

Discovered functional groups align with human-curated groups and include novel patterns.

Dataset-specific functional groups improve bioactivity prediction performance.

Fingerprints based on these groups outperform traditional methods.

Abstract

We introduce the first formal large-scale assessment of the utility of traditional chemical functional groups as used in chemical explanations. Our assessment employs a fundamental principle from computational learning theory: a good explanation of data should also compress the data. We introduce an unsupervised learning algorithm based on the Minimum Message Length (MML) principle that searches for substructures that compress around three million biologically relevant molecules. We demonstrate that the discovered substructures contain most human-curated functional groups as well as novel larger patterns with more specific functions. We also run our algorithm on 24 specific bioactivity prediction datasets to discover dataset-specific functional groups. Fingerprints constructed from dataset-specific functional groups are shown to significantly outperform other fingerprint representations, including the MACCS and Morgan fingerprint, when training ridge regression models on bioactivity regression tasks.