Shift-Robust Molecular Relational Learning with Causal Substructure

TL;DR

This paper introduces CMRL, a causal inference-based model for molecular relational learning that is robust to distributional shifts by focusing on causally relevant substructures, outperforming existing methods.

Contribution

We propose a novel causal substructure detection framework using SCM and conditional interventions to improve molecular interaction predictions under distributional shifts.

Findings

CMRL outperforms state-of-the-art models on real-world datasets.

The causal substructure approach reduces confounding effects.

Extensive experiments validate the robustness of CMRL.

Abstract

Recently, molecular relational learning, whose goal is to predict the interaction behavior between molecular pairs, got a surge of interest in molecular sciences due to its wide range of applications. In this work, we propose CMRL that is robust to the distributional shift in molecular relational learning by detecting the core substructure that is causally related to chemical reactions. To do so, we first assume a causal relationship based on the domain knowledge of molecular sciences and construct a structural causal model (SCM) that reveals the relationship between variables. Based on the SCM, we introduce a novel conditional intervention framework whose intervention is conditioned on the paired molecule. With the conditional intervention framework, our model successfully learns from the causal substructure and alleviates the confounding effect of shortcut substructures that are spuriously correlated to chemical reactions. Extensive experiments on various tasks with real-world and synthetic datasets demonstrate the superiority of CMRL over state-of-the-art baseline models. Our code is available at https://github.com/Namkyeong/CMRL.