Compressing Chemistry Reveals Functional Groups

Clinical Snapshot

Key Findings

• P Value: Not reported

• Effect Size: Significant outperformance over MACCS and Morgan fingerprints

• Primary Outcome: Discovery of substructures that compress molecular data and contain most human-curated functional groups plus novel larger patterns

• Nnt Or Sensitivity: Not applicable - computational methodology study

• Confidence Interval: Not reported

Clinical Application

High feasibility for implementation in computational chemistry workflows Relevant to Australian pharmaceutical research institutions and TGA drug evaluation processes requiring chemical structure analysis Researchers in pharmaceutical development, medicinal chemistry, and computational drug discovery

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 compression of data should reveal a good explanation. 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 data sets to discover data set-specific functional groups. Fingerprints constructed from data set-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.

References

1. 1.Sharma, R., & King, R. D. (2026). Compressing chemistry reveals functional groups. Journal of Chemical Information and Modeling. https://doi.org/10.1021/acs.jcim.5c02917