Functional Groups Are All You Need for Chemically Interpretable Molecular Property Prediction

Molecular property prediction using deep learning (DL) models has accelerated drug and materials discovery, but the resulting DL models often lack interpretability, hindering their adoption by chemists. This work proposes developing molecule representations using the concept of functional groups (FG) in chemistry and introduces the functional group representation (FGR) framework, a novel approach to encoding molecules based on their fundamental chemical substructures. The proposed framework integrates two types of functional groups: those curated from established chemical knowledge (FG) and those mined from a large molecular corpus using sequential pattern mining (MFG). The resulting FGR framework encodes molecules into a lower-dimensional latent space by leveraging pretraining on a large data set of unlabeled molecules. It is shown that the chemistry-inspired, FGR framework achieves state-of-the-art performance on a diverse range of 33 benchmark data sets spanning physical chemistry, biophysics, quantum mechanics, biological activity, and pharmacokinetics while enabling chemical interpretability. Importantly, the FGR-based representations are intrinsically aligned with established chemical principles, enabling chemists to link predicted properties to specific functional groups directly and facilitating novel insights into structure-property relationships. This work demonstrates that the incorporation of chemistry knowledge leads to chemically interpretable and high-performing DL models for property predictions.