AI Paper Atlas

Research map / 3D vision / Face and shape

Learning Compressed Shape-Aware Molecular Representations for Virtual Screening

Robin Winter, Julian Cremer, Djork-Arné Clevert

ICML 2026 · 3D vision › Face and shape · approach: Optimization-based shape matching

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
train end-to-end with quantization-aware IVF-PQ instead of post-hoc FAISS k-means (ablation)SAND · Enamine 10M subset, compression sweepcompression at equal correlation: M=16 matches post-hoc M=64 (about 4x better compression)Improved
quantization-aware training changes embedding geometry (ablation)SAND · 1M test moleculesprincipal components for >90% variance: 317 -> 83Improved
retrain FAISS codebooks post-hoc on quantization-aware model embeddings (ablation)SAND · compression sweepquantization performanceNo clear effect
replace Spearman loss with triplet loss (ablation)SAND (GINE) · Enamine, Recall@100recall: 0.50 -> 0.42Hurt
replace Spearman loss with MSE loss (ablation)SAND (GINE) · Enamine, Recall@100recall: 0.50 -> 0.46Hurt
replace GINE (28M) with GPS (87M) encoder (ablation)SAND · Enamine and ChEMBLrecall: 0.50 -> 0.51 (Enamine, 100)No clear effect
replace GINE with GATv2 encoder (ablation)SAND · Enaminerecall: 0.50 -> 0.50No clear effect
replace IVF-PQ with residual vector quantization (RVQ) (ablation)SAND · Enamine / ChEMBLSpearman correlation: Enamine 0.78 -> 0.84; ChEMBL 0.73 -> 0.81Mixed
apply 64x compression (M=32 quantization) (ablation)SAND · Enamine / ChEMBLSpearman correlation: 0.86 -> 0.78 (Enamine); 0.81 -> 0.73 (ChEMBL)Hurt
train on shape overlay scores with rank-preserving contrastive loss vs ECFP/pretrained 3D baselinesSAND · Enamine, Recall@1000recall: 0.30 (ECFP) -> 0.89Improved
use pretrained 3D-aware representations (GraphMVP, DrugCLIP) instead of ECFPshape retrieval baselines · Enaminetop-100 recall vs correlation: R 0.20 -> 0.37 but recall 0.13 -> 0.05Mixed
retrieve analogs with SAND instead of ECFP or random for Flowr-generated ligandshit expansion pipeline · JNK3 and p38 MAPK dockingChemgauss4 score and post-optimization RMSDImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy