DEL-Ranking: Ranking-Correction Denoising Framework for Elucidating Molecular Affinities in DNA-Encoded Libraries
Keywords: DEL Denoising, Deep Learning, Bioinformatics
Abstract: DNA-encoded library (DEL) screening has revolutionized protein-ligand binding detection, enabling rapid exploration of vast chemical spaces through read count analysis. However, two critical challenges limit its effectiveness: distribution noise in low copy number regimes and systematic shifts between read counts and true binding affinities. We present DEL-Ranking, a comprehensive framework that simultaneously addresses both challenges through innovative ranking-based denoising and activity-referenced correction. Our approach introduces a dual-perspective ranking strategy combining Pair-wise Soft Rank (PSR) and List-wise Global Rank (LGR) constraints to preserve both local and global count relationships. Additionally, we develop an Activity-Referenced Correction (ARC)
module that bridges the gap between read counts and binding affinities through iterative refinement and biological consistency enforcement. Another key contribution of this work is the curation and release of three comprehensive DEL datasets that uniquely combine ligand 2D sequences, 3D conformational information, and experimentally validated activity labels. We validate our framework on
five diverse DEL datasets and introduce three new comprehensive datasets featuring 2D sequences, 3D structures, and activity labels. DEL-Ranking achieves state-of-the-art performance across multiple correlation metrics and demonstrates strong generalization ability across different protein targets. Importantly, our approach successfully identifies key functional groups associated with binding affinity, providing actionable insights for drug discovery. This work advances both the accuracy and interpretability of DEL screening, while contributing valuable datasets for future research.
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.
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: 1248
Loading