On Improving Graph Neural Networks for QSAR by Pre-training on Extended-Connectivity Fingerprints

2026-05-11

AI Summary

This work addresses the unstable out-of-distribution (OOD) generalization of molecular graph neural networks in quantitative structure–activity relationship (QSAR) tasks. The authors propose a pretraining strategy based on extended-connectivity fingerprints (ECFP), which enhances the representational capacity of graph neural networks by predicting ECFP substructure information. The approach is systematically evaluated on both in-distribution and OOD QSAR benchmarks. Experiments on six Biogen datasets demonstrate that the method significantly outperforms existing baselines on five datasets, yielding notable improvements in OOD generalization; however, its benefits are limited in tasks involving highly heterogeneous compounds or complex endpoints. The study also uncovers how substructure-level data leakage influences pretraining efficacy, providing insights into the underlying mechanisms affecting performance gains.

Abstract

Molecular Graph Neural Networks (GNNs) are increasingly common in drug discovery, particularly for Quantitative Structure-Activity Relationship (QSAR) studies; yet, their superiority compared to classical molecular featurisation approaches is disputed. We report a general strategy for improving GNNs for QSAR by pre-training to predict Extended-Connectivity Fingerprints (ECFP). We validate our approach with statistical tests and challenging out-of-distribution (OOD) splits. Across five out of six Biogen benchmarks, we observed a statistically significant improvement in standard performance metrics over all evaluated baselines when using ECFP pre-trained GNNs. However, for more heterogeneous datasets and more complex endpoints, such as binding affinity prediction, pre-trained GNNs underperformed in OOD settings. Importantly, we investigated the impact of substructure-level data leakage during pre-training on downstream performance. While we identified scenarios where pre-training on ECFPs was less effective, our findings show that ECFP-based pre-training can enhance downstream OOD performance on a diverse set of practically relevant QSAR tasks.

Problem

• Graph Neural Networks
• QSAR
• Out-of-Distribution
• Molecular Representation
• Pre-training

Innovation

• Graph Neural Networks
• Pre-training
• Extended-Connectivity Fingerprints
• QSAR
• Out-of-Distribution Generalization