Compressing Chemistry Reveals Functional Groups
This work addresses the need for better chemical representation methods in drug discovery and bioactivity prediction, offering a novel approach that improves performance but is incremental in its application to existing datasets.
The researchers tackled the problem of evaluating chemical functional groups' utility by introducing an unsupervised learning algorithm based on the Minimum Message Length principle to discover compressing substructures in around three million molecules, resulting in fingerprints that significantly outperform existing representations like MACCS and Morgan in bioactivity prediction tasks.
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.