Path Complex Message Passing for Molecular Property Prediction
Keywords: Molecular Property Prediction; Path Complex; Geometric Deep Learning; High-order Interaction; Low-order Interaction
Abstract: Geometric deep learning (GDL) has demonstrated enormous power in molecular data analysis. However, GDL faces challenges in achieving high efficiency and expressivity in molecular representations when high-order terms of the atomic force fields are not sufficiently learned. In this work, we introduce message passing on path complexes, called the Path Complex Message Passing, for molecular prediction. Path complexes represent the geometry of paths and can model the chemical and non-chemical interactions of atoms in a molecule across various dimensions. Our model defines messages on path complexes and employs neural message passing to learn simplex features, enabling feature communication within and between different dimensions. Since messages on high-order and low-order path complexes reflect different aspects of molecular energy, they are updated sequentially according to their order. The higher the order of the path complex, the richer the information it contains, and the higher its priority during inference. It can thus characterize various types of molecular interactions specified in molecular dynamics (MD) force fields. Our model has been extensively validated on benchmark datasets and achieves state-of-the-art results.
The code is available at \url{https://anonymous.4open.science/r/Path-Complex-Neural-Network-32D6}
Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 1420
Loading