AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

Drugging the Undruggable: Benchmarking and Modeling Fragment-Based Screening

Haichuan Tan, Bowen Gao, Jiaxin Li, Yinjun Jia, Wenyu Zhu, Wenxuan Xie, Yihong Liu, Yanwen Huang, Jianhui Wang, Yuanhuan Mo, Ya-Qin Zhang, Wei-Ying Ma and 1 more

ICLR 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Protein interface and docking prediction

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
add fusion module (ablation)FragCLIP · FragBench (fragment retrieval, 90% homology filter)EF@0.5%: 6.428 -> 6.853Improved
add fusion module (ablation)FragCLIP · FragBench (fragment retrieval, 90% homology filter)BEDROC: 0.105 -> 0.115Improved
add molecule modality (molecule encoder) (ablation)FragCLIP · FragBench (fragment retrieval, 90% homology filter)EF@0.5%: 6.170 -> 6.853Improved
add molecule modality (molecule encoder) (ablation)FragCLIP · FragBench (fragment retrieval, 90% homology filter)EF@1%: 5.296 -> 5.797Improved
add fusion module at inference (ablation)FragCLIP · DUD-E (molecule-level)EF@1%: 33.56 -> 37.23Improved
add fragment-level contrastive supervision during training (molecule-only retrieval)DrugCLIP-style molecule retrieval · DUD-E (molecule-level)EF@1%: 31.87 -> 33.56Improved
multi-granularity ensemble of molecule, fragment and fusion scoresFragCLIP · DUD-E (molecule-level)EF@1%: 37.23Improved
replace DrugCLIP with FragCLIPfragment retrieval · FragBench (fragment retrieval, 90% homology filter)EF@0.5%: 4.110 -> 6.853Improved
replace DrugCLIP with FragCLIPfragment retrieval · FragBench (fragment retrieval, 90% homology filter)AUROC: 0.597 -> 0.593No clear effect
filter training proteins at 90% instead of 100% sequence identity (ablation)FragCLIP · FragBench (fragment retrieval, 90% homology filter)EF@0.5%: 10.741 -> 6.853Hurt
filter training proteins at 30% instead of 90% sequence identity (ablation)FragCLIP · FragBench (fragment retrieval, 90% homology filter)EF@0.5%: 6.853 -> 4.337Hurt
require 3 docking replicates instead of 1 (k=2) (ablation)fragment labeling · PDBbind 500 complexesprecision: 0.784 -> 0.894Improved

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