Leakage-controlled benchmarking reveals generalization limits of deep learning for protein-ligand binding affinity prediction
bioRxiv 10.64898/2026.09.17.752433v1 (posted 2026-09-23); 30-page PDF · Preprint
What this source contributed
Records in this release that cite this source. A record can cite several sources.
Results
Values as the source printed them. Compare results only under compatible protocols, datasets and metrics.
Models, benchmarks and other records
Models (7)
Configurations (12)
- Boltz2 (CASP16 and ChEMBL35 evaluation, PLABench)
- DeepDTA (CASP16 and ChEMBL35 evaluation, PLABench)
- DeepDTA (retrained on Davis and KIBA, PLABench)
- FlowDock (CASP16 and ChEMBL35 evaluation, PLABench)
- FLOWR.ROOT (CASP16 and ChEMBL35 evaluation, PLABench)
- Graph_RG (CASP16 and ChEMBL35 evaluation, PLABench)
- LCDD-team (CASP16 and ChEMBL35 evaluation, PLABench)
- LLF (CASP16 and ChEMBL35 evaluation, PLABench)
- LLF (retrained on Davis and KIBA, PLABench)
- MFE (CASP16 and ChEMBL35 evaluation, PLABench)
- MixingDTA (CASP16 and ChEMBL35 evaluation, PLABench)
- MixingDTA (retrained on Davis and KIBA, PLABench)
Benchmarks (1)
Protocols (9)
Datasets (9)
- CASP16 ligand affinity series L1000 (chymase)
- CASP16 ligand affinity series L3000 (autotaxin)
- Davis cold-drug test set (PLABench)
- Davis cold-target test set (PLABench)
- Davis warm test set (PLABench)
- KIBA cold-drug test set (PLABench)
- KIBA cold-target test set (PLABench)
- KIBA warm test set (PLABench)
- PLABench ChEMBL35 leakage-controlled test set
Evaluations (45)
- Boltz2 on PLABench casp16-L1000
- Boltz2 on PLABench casp16-L3000
- Boltz2 on PLABench chembl35
- DeepDTA on PLABench casp16-L1000
- DeepDTA on PLABench casp16-L3000
- DeepDTA on PLABench chembl35
- DeepDTA on PLABench Davis-cold-drug
- DeepDTA on PLABench Davis-cold-target
- DeepDTA on PLABench Davis-warm
- DeepDTA on PLABench KIBA-cold-drug
- DeepDTA on PLABench KIBA-cold-target
- DeepDTA on PLABench KIBA-warm
- FlowDock on PLABench casp16-L1000
- FlowDock on PLABench casp16-L3000
- FlowDock on PLABench chembl35
- FLOWR.ROOT on PLABench casp16-L1000
- FLOWR.ROOT on PLABench casp16-L3000
- FLOWR.ROOT on PLABench chembl35
- Graph_RG on PLABench casp16-L1000
- Graph_RG on PLABench casp16-L3000
- Graph_RG on PLABench chembl35
- LCDD-team on PLABench casp16-L1000
- LCDD-team on PLABench casp16-L3000
- LCDD-team on PLABench chembl35
- LLF on PLABench casp16-L1000
Evidence
Source checking verifies the cited claim or transcription. It does not establish independent reproduction.
Evidence table
Inspect claims, sources and review details
Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.
One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.
24 evidence rows matching the loaded filters
| Property and statement | Original source and location | Review and provenance |
|---|---|---|
| Artifact SHA-256 5fba9691f7d0777d19dabc66144e5c23d40e430de4919f97b92e4d50fb397281 Source metadata | Leakage-controlled benchmarking reveals generalization limits of deep learning for protein-ligand binding affinity prediction No field-specific location recorded Version: bioRxiv 10.64898/2026.09.17.752433v1 (posted 2026-09-23); 30-page PDF | catalogued No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Artifact url https://www.biorxiv.org/content/10.64898/2026.09.17.752433v1.full.pdf Source metadata | Leakage-controlled benchmarking reveals generalization limits of deep learning for protein-ligand binding affinity prediction No field-specific location recorded Version: bioRxiv 10.64898/2026.09.17.752433v1 (posted 2026-09-23); 30-page PDF | catalogued No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Code url https://doi.org/10.5281/zenodo.22782934 Source metadata | Leakage-controlled benchmarking reveals generalization limits of deep learning for protein-ligand binding affinity prediction No field-specific location recorded Version: bioRxiv 10.64898/2026.09.17.752433v1 (posted 2026-09-23); 30-page PDF | catalogued No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Doi 10.64898/2026.09.17.752433 Source metadata | Leakage-controlled benchmarking reveals generalization limits of deep learning for protein-ligand binding affinity prediction No field-specific location recorded Version: bioRxiv 10.64898/2026.09.17.752433v1 (posted 2026-09-23); 30-page PDF | catalogued No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Licence CC-BY-NC-ND-4.0 Source metadata | Leakage-controlled benchmarking reveals generalization limits of deep learning for protein-ligand binding affinity prediction No field-specific location recorded Version: bioRxiv 10.64898/2026.09.17.752433v1 (posted 2026-09-23); 30-page PDF | catalogued No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Limitations3 values
| Leakage-controlled benchmarking reveals generalization limits of deep learning for protein-ligand binding affinity prediction No field-specific location recorded Version: bioRxiv 10.64898/2026.09.17.752433v1 (posted 2026-09-23); 30-page PDF | catalogued No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Media type application/pdf Source metadata | Leakage-controlled benchmarking reveals generalization limits of deep learning for protein-ligand binding affinity prediction No field-specific location recorded Version: bioRxiv 10.64898/2026.09.17.752433v1 (posted 2026-09-23); 30-page PDF | catalogued No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Publication status preprint Source metadata | Leakage-controlled benchmarking reveals generalization limits of deep learning for protein-ligand binding affinity prediction No field-specific location recorded Version: bioRxiv 10.64898/2026.09.17.752433v1 (posted 2026-09-23); 30-page PDF | catalogued No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Retrieved at 2026-10-11T21:28:04Z Source metadata | Leakage-controlled benchmarking reveals generalization limits of deep learning for protein-ligand binding affinity prediction No field-specific location recorded Version: bioRxiv 10.64898/2026.09.17.752433v1 (posted 2026-09-23); 30-page PDF | catalogued No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Review: date 2026-10-11 Source metadata | Leakage-controlled benchmarking reveals generalization limits of deep learning for protein-ligand binding affinity prediction No field-specific location recorded Version: bioRxiv 10.64898/2026.09.17.752433v1 (posted 2026-09-23); 30-page PDF | catalogued No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Sources and history
Release 2026-10-23-5347695500ed · Record review: source checked
Primary source retrieved and hashed for the 2026-10-11 microbial and interactions literature pass.
Technical metadata and extraction receipts
Stable ID: lit-interactions-20261011-leakctl-source-wang2026
- areas
- molecular-interactions
- contexts
- research
- url
- https://doi.org/10.64898/2026.09.17.752433
- artifact url
- https://www.biorxiv.org/content/10.64898/2026.09.17.752433v1.full.pdf
- version
- bioRxiv 10.64898/2026.09.17.752433v1 (posted 2026-09-23); 30-page PDF
- retrieved at
- 2026-10-11T21:28:04Z
- artifact sha256
- 5fba9691f7d0777d19dabc66144e5c23d40e430de4919f97b92e4d50fb397281
- doi
- 10.64898/2026.09.17.752433
- publication status
- preprint
- licence
- CC-BY-NC-ND-4.0
- media type
- application/pdf
- code url
- https://doi.org/10.5281/zenodo.22782934
- limitations
- bioRxiv adds a per-request banner, so the PDF hash describes this one retrieval.; Values were read from the pdftotext (poppler 26.08.0) layout text layer, SHA-256 acc4a3c5deb4068748abc233a5aa74e1633747737f438eebccf7270d49a16328. The licence (CC BY-NC-ND) is respected by not archiving the PDF or its text in the repository.; Bold marking of the best value per column is not carried by the text layer and is not recorded.
- review
- method: source-hash-verification; method note: PDF re-downloaded; the per-request bioRxiv banner changes the PDF hash, so the pdftotext -layout text layer was rehashed instead and matches the recorded text-layer hash.; reviewer: claude; reviewer note: Claude review agent reviewer-mi-2, independent of the extractor lit-microbial-interactions; no human review claimed; date: 2026-10-11