Search papers, labs, and topics across Lattice.
This paper introduces an additive deep-learning framework that separates the contributions of physicochemical descriptors and molecular graph topology in predicting aqueous solubility, a crucial property in drug discovery. By employing a multilayer perceptron for the chemical branch and a graph neural network for the structural branch, the model allows for a clear decomposition of the influences on solubility predictions. The approach not only enhances predictive accuracy through pretraining and fine-tuning but also provides interpretable insights into the roles of chemical and structural factors in solubility.
The framework reveals distinct contributions of chemical and structural factors to aqueous solubility, enhancing both predictive accuracy and interpretability in drug discovery models.
Aqueous solubility is a key property in early-stage drug discovery, but most predictive models merge physicochemical descriptors and molecular graph information into a single representation, obscuring whether a prediction is driven by global chemistry, molecular structure, or both. We present an additive deep-learning framework that keeps these two sources of information separate throughout training: physicochemical descriptors are encoded by a multilayer perceptron (the chemical branch) and molecular graph topology by a graph neural network (the structural branch), with the two outputs combined only at the prediction stage through an additive model with an optional multiplicative interaction. This design provides a direct decomposition of chemical and structural components that can be examined separately after training. Furthermore, pretraining on the larger AqSolDB dataset and fine-tuning on the smaller BigSolDB2 dataset substantially improve accuracy and reduce run-to-run variations, indicating generalizability of the learned features from the data-rich settings. We further interpret the fitted model using best linear projections of the branch outputs, molecule-level embedding summaries across solubility classes, and atom-level GNNExplainer masks aggregated over functional groups. These analyses show that the chemical branch aligns with familiar physicochemical descriptors, while the structural branch captures graph-topological and functional-group patterns associated with solubility. Across both datasets, the framework attains competitive predictive performance while making the distinct roles of chemical and structural information more transparent.
