Evidential Deep Learning for Guided Molecular Property Prediction and Discovery

TLDR

While neural networks achieve state‑of‑the‑art performance for many molecular modeling and structure‑property prediction tasks, they often struggle with out‑of‑domain generalization, poor sample efficiency, and uncalibrated predictions. The paper demonstrates a new approach to uncertainty quantification for neural network‑based molecular structure‑property prediction using evidential deep learning, without additional computational cost. The authors develop evidential 2D message‑passing and 3D atomistic neural networks and apply them to a range of molecular property prediction tasks. The study shows that evidential uncertainties yield calibrated predictions, enable sample‑efficient active learning, and improve virtual screening validation rates, indicating that evidential deep learning is an efficient uncertainty‑quantification method for molecular property prediction, discovery, and design.

Abstract

While neural networks achieve state-of-the-art performance for many molecular modeling and structure-property prediction tasks, these models can struggle with generalization to out-of-domain examples, exhibit poor sample efficiency, and produce uncalibrated predictions. In this paper, we leverage advances in evidential deep learning to demonstrate a new approach to uncertainty quantification for neural network-based molecular structure-property prediction at no additional computational cost. We develop both evidential 2D message passing neural networks and evidential 3D atomistic neural networks and apply these networks across a range of different tasks. We demonstrate that evidential uncertainties enable (1) calibrated predictions where uncertainty correlates with error, (2) sample-efficient training through uncertainty-guided active learning, and (3) improved experimental validation rates in a retrospective virtual screening campaign. Our results suggest that evidential deep learning can provide an efficient means of uncertainty quantification useful for molecular property prediction, discovery, and design tasks in the chemical and physical sciences.