Yuechuan Lin | Xu Lab

An electron-density point-cloud framework for robust protein-ligand interaction prediction

Thu, 11 Jun 2026 00:00:00 +0000

Research question

Can an electron-density representation improve binding-affinity prediction when atomic coordinates are uncertain?

How the method works

E-CloudBind combines covalent molecular graphs with point clouds representing ligand electron density and a protein-pocket density proxy. Ligand densities are approximated with GFN2-xTB; pocket clouds use van der Waals-guided Gaussian functions. A heterogeneous graph network combines these features to predict binding affinity.

E-CloudBind: from electron-density representations to binding-affinity prediction.

Figure 2 · Yujian Liu et al., Nature Communications 17, 7424 (2026). © The Author(s) 2026. Full figure reproduced; display size and image format only are adjusted. · Published figure · CC BY 4.0

View full-size figure

What the study shows

The reported tests span crystallographic resolution, experimental and AlphaFold2 structures, and ligand-scaffold, protein-sequence and joint holdouts. They support improved robustness to structural uncertainty. An ablation also finds similar prediction performance using a cheaper Gaussian approximation for ligand clouds.

Scope of the evidence

The model uses prepared three-dimensional structures and predefined pockets. Explicit solvent and polarization effects are omitted, and substantial geometric corruption reduces performance. These results support affinity prediction; mechanistic explanations require independent testing.

Read the publisher record for the abstract and access options.

MPFusion-MIL: Morphology-Guided Fusion with Precise Cross-Scale Interaction for Whole Slide Image Analysis

Thu, 01 Jan 2026 00:00:00 +0000

Accepted for ACM International Conference on Multimedia (ACM MM) 2026. Proceedings details will be added when available.

PyraE2E: Enhancing End-to-End WSI Analysis via Cross-Scale Super-Resolution

Thu, 01 Jan 2026 00:00:00 +0000

Accepted for European Conference on Computer Vision (ECCV) 2026. Proceedings details will be added when available.

View the conference’s accepted-paper list.

Minimal High-Resolution Patches Are Sufficient for Whole Slide Image Representation via Cascaded Dual-Scale Reconstruction

Sun, 03 Aug 2025 00:00:00 +0000

A conference version has been accepted at PRCV 2026. The accepted title is “Learning Whole Slide Image Representations from Few High-Resolution Patches via Cascaded Dual-Scale Reconstruction”. The author list and downloadable citation on this page describe the 2025 preprint; the conference citation will be added after its final metadata is confirmed.

Read the preprint and its version history.