Fundamental Determinants of Misranking in Affinity-Based Ligand Ranking
- 1Department of Chemistry, York University, 4700 Keele St, Toronto, Canada, M3J 1P3
Abstract
Affinity-based prioritization depends on preserving the relative order of ligands, which is distinct from determining each equilibrium dissociation constant, K d , accurately. Errors that substantially bias individual K d values do not necessarily alter ranking if they affect ligands similarly, whereas smaller differential errors can reverse the order of ligands whose true affinities are close. Here, we establish a quantitative framework for understanding this distinction in full-titration binding assays. First-principles analysis of equilibrium 1:1 binding shows that the target-to-affinity ratio, T 0 / K d , governs the amplification of concentration errors into fitted-affinity errors, while the true affinity ratio between two ligands defines the margin that differential errors must overcome to produce rank inversion. Controlled end-to-end simulations extend this analysis to additive and nonadditive response models and map the regimes in which affinity ranking becomes vulnerable. The results show that ranking progressively destabilizes as local affinity separation decreases and T 0 / K d increases. They further demonstrate that ligand-specific signal nonadditivity can introduce differential distortions that impair ranking even when concentration-dependent error amplification is weak, whereas response transformations shared across ligands have substantially less effect on their relative ranking. Library-scale simulations additionally show that strong overall rank correlation can coexist with substantial errors in recovering the highest-affinity candidates, indicating that global rank statistics alone may not adequately assess prioritization performance. We translate these findings into a practical workflow emphasizing appropriate concentration design, validation of response additivity, control of ligand-to-ligand variability, and selection-focused performance metrics for reliable affinity ranking.
Related
- Incorporating Polarization into ML/MM Endpoint Calculations for Protein–Ligand Affinity Ranking
- Comparative Analysis of Basic Machine Learning and Rigorous Free-Energy Calculation on Benchmarking Datasets of Protein-Ligand Binding Affinity
- Machine-learning force-field scoring rivals free-energy perturbation for congeneric ligand ranking across public benchmarks
- Ligand Substitution in Three Organic-Inorganic Hybrid Zincophosphite Analogs: Modulating Lanthanide Affinity and White-Light Emission via Hard-Soft Acid-Base Interplay.
- Quantitative Protein–Ligand Binding Affinity Prediction from Quantum-Mechanical Scoring
BibTeX
@misc{wang2026fundamental,
title = {{Fundamental Determinants of Misranking in Affinity-Based Ligand Ranking}},
author = {Tong Ye Wang and Parmeetpal Dhillon and Sergey N. Krylov},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15009539/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009539/v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Fundamental Determinants of Misranking in Affinity-Based Ligand Ranking
AU - Tong Ye Wang
AU - Parmeetpal Dhillon
AU - Sergey N. Krylov
PY - 2026
DA - 2026/09/28
PB - ChemRxiv
DO - 10.26434/chemrxiv.15009539/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009539/v1
AB - Affinity-based prioritization depends on preserving the relative order of ligands, which is distinct from determining each equilibrium dissociation constant, K d , accurately. Errors that substantially bias individual K d values do not necessarily alter ranking if they affect ligands similarly, whereas smaller differential errors can reverse the order of ligands whose true affinities are close. Here, we establish a quantitative framework for understanding this distinction in full-titration binding assays. First-principles analysis of equilibrium 1:1 binding shows that the target-to-affinity ratio, T 0 / K d , governs the amplification of concentration errors into fitted-affinity errors, while the true affinity ratio between two ligands defines the margin that differential errors must overcome to produce rank inversion. Controlled end-to-end simulations extend this analysis to additive and nonadditive response models and map the regimes in which affinity ranking becomes vulnerable. The results show that ranking progressively destabilizes as local affinity separation decreases and T 0 / K d increases. They further demonstrate that ligand-specific signal nonadditivity can introduce differential distortions that impair ranking even when concentration-dependent error amplification is weak, whereas response transformations shared across ligands have substantially less effect on their relative ranking. Library-scale simulations additionally show that strong overall rank correlation can coexist with substantial errors in recovering the highest-affinity candidates, indicating that global rank statistics alone may not adequately assess prioritization performance. We translate these findings into a practical workflow emphasizing appropriate concentration design, validation of response additivity, control of ligand-to-ligand variability, and selection-focused performance metrics for reliable affinity ranking.
ER -