Graph-Enhanced Learning for Predicting Optimal Drug Combinations Using Contrastive Embedding
Keywords: Graph Learning, Contrastive Embedding, DDI
TL;DR: We introduce a novel DDI-aware optimal transport problem, formulated as a geodesic equation on an infinite-dimensional Finsler
Abstract: We present a groundbreaking unified theory for drug-drug interaction (DDI) aware domain adaptation (DA) in the context of drug synergy prediction. Our framework seamlessly integrates concepts from optimal transport, information geometry, and quantum information theory within the setting of abstract Banach spaces. We introduce a novel DDI-aware optimal transport problem, formulated as a geodesic equation on an infinite-dimensional Finsler manifold that encodes both DDI structure and optimal transport costs. This geometric formulation provides a unified perspective on DDI-aware domain adaptation, interpreting the process as the evolution of a transport map along a geodesic in a space that captures both domain discrepancy and drug interaction patterns. Our approach extends to a stochastic gradient flow on the space of probability measures, combining ideas from information geometry and stochastic analysis. We prove the existence of a unique invariant measure for this flow and establish its convergence properties using techniques from infinite-dimensional Markov processes and Γ-convergence. Our comprehensive mathematical framework not only unifies existing approaches to domain adaptation and DDI prediction but also opens new avenues for research at the intersection of these fields. By bridging the gap between abstract mathematical theories and practical drug synergy prediction, our work paves the way for more effective and theoretically grounded algorithms in drug discovery and personalized medicine. The proposed unified theory has far-reaching implications, potentially revolutionizing our understanding of cross-domain adaptation in complex biochemical systems and inspiring novel computational methods in pharmaceutical research.
Primary Area: transfer learning, meta learning, and lifelong learning
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: 13871
Loading