Enhancing molecular property prediction with auxiliary learning and task-specific adaptation

Abstract

AbstractPretrained Graph Neural Networks have been widely adopted for various molecular property prediction tasks. Despite their ability to encode structural and relational features of molecules, traditional fine-tuning of such pretrained GNNs on the target task can lead to poor generalization. To address this, we explore the adaptation of pretrained GNNs to the target task by jointly training them with multiple auxiliary tasks. This could enable the GNNs to learn both general and task-specific features, which may benefit the target task. However, a major challenge is to determine the relatedness of auxiliary tasks with the target task. To address this, we investigate multiple strategies to measure the relevance of auxiliary tasks and integrate such tasks by adaptively combining task gradients or by learning task weights via bi-level optimization. Additionally, we propose a novel gradient surgery-based approach, Rotation of Conflicting Gradients ($$\mathop {\texttt{RCGrad}}\limits$$ RCGrad ), that learns to align conflicting auxiliary task gradients through rotation. Our experiments with state-of-the-art pretrained GNNs demonstrate the efficacy of our proposed methods, with improvements of up to 7.7% over fine-tuning. This suggests that incorporating auxiliary tasks along with target task fine-tuning can be an effective way to improve the generalizability of pretrained GNNs for molecular property prediction.Scientific contributionWe introduce a novel framework for adapting pretrained GNNs to molecular tasks using auxiliary learning to address the critical issue of negative transfer. Leveraging novel gradient surgery techniques such as $$\mathop {\texttt{RCGrad}}\limits$$ RCGrad , the proposed adaptation framework represents a significant departure from the dominant pretraining fine-tuning approach for molecular GNNs. Our contributions are significant for drug discovery research, especially for tasks with limited data, filling a notable gap in the efficient adaptation of pretrained models for molecular GNNs.