Two barriers to benchmarking structure-based virtual screening against the AcrB efflux pump: a sample-size ceiling and an identifiability failure
- 1Faculty of Health Sciences
- 2Bahçeşehir Cyprus University
Abstract
Background. The AcrAB-TolC system is the principal constitutive efflux pump of Escherichia coli and an attractive adjuvant target. Virtual screening campaigns against its distal binding pocket have been published without the retrospective validation standard elsewhere in the field. This study asked whether such a benchmark can be built from the published literature. Methods. Compounds reported as AcrB inhibitors were filtered against an independent reproducibility screen and assigned to a binding site by the strongest structural or mutational evidence available. Property-matched decoys, a library-representative background and a class of reported-but-non-reproducing inhibitors served as negatives. A pilot docking screen against the hydrophobic trap of 5ENQ was re-analysed under the curated labels, with areas under the receiver-operating-characteristic curve reported with analytic and bootstrap confidence intervals. Two controls were pre-specified: a wrong-site control, and a ligand-only ablation deleting the receptor entirely. Results. Of 49 compounds reported as AcrB inhibitors, 38 have been independently retested and 17 reproduced; 13 maps to the distal pocket and 5 have a co-crystal there. Two barriers follow. With 5 actives the 95% interval cannot be narrower than ±0.26; reaching a conventional ±0.06 would require 94 actives, while raising the decoy ratio from 40:1 to 500:1 narrows the interval by 0.8%. More seriously, the ligand-only ablation reached 1.000 against a docking area of 0.707: deleting the receptor improved performance. 4 of the 5 co-crystal actives are analogues spanning 2 scaffolds, with a median pairwise ECFP4 similarity of 0.574 against 0.077 to the negatives; every active–active pair lies above the decoy-exclusion threshold of 0.35 and every active–negative pair below it. Conclusions. A retrospective docking benchmark for the AcrB distal pocket cannot currently be built. The first barrier is quantitative and yields to more confirmed inhibitors of any kind; the second is qualitative and yields only to inhibitors that are chemically unlike one another. Screening against this pocket therefore proceeds unvalidated and cannot be validated until structurally diverse confirmed inhibitors exist, which is a medicinal-chemistry result rather than a computational one. The curated compound table, the metrics implementation with its unit tests and the analysis notebook are released.
Related
- LASSBio-classFLOW: A Semiautomated KNIME Workflow for Classification Model Benchmarking and Virtual Screening
- Structure-based virtual screening identifies naturally derived histidine kinase VbrK inhibitors as potential countermeasures against β-lactam antimicrobial resistance
- TMK-VS-Pred: Impact of Decoy Choice on Target-Specific Scoring for Structure-Based Virtual Screening of Thymidylate Kinase, a Potential Anti-Mycobacterium Tuberculosis Target
- An Experimentally Validated Structure-Based Virtual Screening Approach to Identify Nucleotide-Binding Protein Inhibitors as a New Source of Kinase Inhibitors
- V-SYNTHES2—the next generation tool for structure-based virtual screening of giga-scale chemical spaces
BibTeX
@misc{shuaibu2026barriers,
title = {{Two barriers to benchmarking structure-based virtual screening against the AcrB efflux pump: a sample-size ceiling and an identifiability failure}},
author = {Ibrahim Ibrahim Shuaibu},
howpublished = {ChemRxiv},
year = {2026},
month = oct,
doi = {10.26434/chemrxiv.15009731/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009731/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Two barriers to benchmarking structure-based virtual screening against the AcrB efflux pump: a sample-size ceiling and an identifiability failure
AU - Ibrahim Ibrahim Shuaibu
PY - 2026
DA - 2026/10/01
PB - ChemRxiv
DO - 10.26434/chemrxiv.15009731/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009731/v1
AB - Background. The AcrAB-TolC system is the principal constitutive efflux pump of Escherichia coli and an attractive adjuvant target. Virtual screening campaigns against its distal binding pocket have been published without the retrospective validation standard elsewhere in the field. This study asked whether such a benchmark can be built from the published literature. Methods. Compounds reported as AcrB inhibitors were filtered against an independent reproducibility screen and assigned to a binding site by the strongest structural or mutational evidence available. Property-matched decoys, a library-representative background and a class of reported-but-non-reproducing inhibitors served as negatives. A pilot docking screen against the hydrophobic trap of 5ENQ was re-analysed under the curated labels, with areas under the receiver-operating-characteristic curve reported with analytic and bootstrap confidence intervals. Two controls were pre-specified: a wrong-site control, and a ligand-only ablation deleting the receptor entirely. Results. Of 49 compounds reported as AcrB inhibitors, 38 have been independently retested and 17 reproduced; 13 maps to the distal pocket and 5 have a co-crystal there. Two barriers follow. With 5 actives the 95% interval cannot be narrower than ±0.26; reaching a conventional ±0.06 would require 94 actives, while raising the decoy ratio from 40:1 to 500:1 narrows the interval by 0.8%. More seriously, the ligand-only ablation reached 1.000 against a docking area of 0.707: deleting the receptor improved performance. 4 of the 5 co-crystal actives are analogues spanning 2 scaffolds, with a median pairwise ECFP4 similarity of 0.574 against 0.077 to the negatives; every active–active pair lies above the decoy-exclusion threshold of 0.35 and every active–negative pair below it. Conclusions. A retrospective docking benchmark for the AcrB distal pocket cannot currently be built. The first barrier is quantitative and yields to more confirmed inhibitors of any kind; the second is qualitative and yields only to inhibitors that are chemically unlike one another. Screening against this pocket therefore proceeds unvalidated and cannot be validated until structurally diverse confirmed inhibitors exist, which is a medicinal-chemistry result rather than a computational one. The curated compound table, the metrics implementation with its unit tests and the analysis notebook are released.
ER -