An Additive MLP-GNN Framework for Characterizing Chemical and Structural Contributions to Aqueous Solubility

2026-07-02

AI Summary

This work addresses the challenge of disentangling the independent contributions of chemical properties and molecular structure in aqueous solubility prediction. The authors propose an additive deep learning framework that employs separate multilayer perceptrons (MLPs) and graph neural networks (GNNs) to encode physicochemical descriptors and molecular graph topology, respectively, fusing their outputs additively to enable the first interpretable decomposition of chemical and structural influences. An optional multiplicative interaction term is introduced to enhance model expressivity. Leveraging a pretraining-finetuning strategy, linear projection-based interpretation, embedding clustering, and GNNExplainer-derived atom masking, the model achieves consistently strong predictive performance on AqSolDB and BigSolDB2, significantly improving accuracy while demonstrating robust generalization and interpretability.

Abstract

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.

Problem

• aqueous solubility
• physicochemical descriptors
• molecular graph
• predictive modeling
• interpretability

Innovation

• Additive MLP-GNN framework
• aqueous solubility prediction
• chemical-structural decomposition
• interpretable deep learning
• graph neural networks