AgenticPosesRanker: An Agentic AI Framework for Physically Grounded Ranking of Protein-Ligand Docking Poses

TL;DR

AgenticPosesRanker is an AI framework that combines physical analysis tools and GPT-5 reasoning to evaluate and rank protein-ligand docking poses, aiming to improve accuracy and interpretability.

Contribution

This work introduces a novel agentic AI framework integrating physical analysis and large-language models for pose ranking in molecular docking.

Findings

Achieved 50% best-pose accuracy on a benchmark, matching the baseline.

Significantly outperformed random chance with p < 0.001.

High alignment between tool weights and metric separations (median a +0.83).

Abstract

Scoring functions remain the principal bottleneck in molecular docking: they routinely fail to rank near-native poses above decoys, and their composite single-score design obscures the physicochemical basis of each ranking error. We present AgenticPosesRanker, an agentic AI framework that combines six deterministic, physically grounded analysis tools (interaction fingerprinting, solvent-accessible burial, conformational strain, steric-clash detection, unsatisfied-polar-atom penalty, and chemical-identity extraction) with large-language-model (GPT-5) chain-of-thought reasoning to evaluate and rank docking poses. On a curated benchmark of ten protein-ligand systems (162 poses) balanced by construction between Smina scoring-function successes and failures, the agent achieved 50.0% best-pose accuracy, matching the design-fixed Smina baseline of 50.0% and significantly exceeding a 7.7% uniformly random baseline (p < 0.001, one-sided exact binomial test). The balanced-benchmark accuracy decomposes symmetrically: the agent retained 80% (4/5) of the Smina-success systems and recovered 20% (1/5) of the Smina-failure systems, so the aggregate 50% reflects one regression offset by one recovery rather than any net improvement over the Smina reference. Decision-attribution analysis showed high alignment between the agent's self-reported tool weights and objective metric separations of the selected pose (median \r{ho} = +0.83), consistent across correct and incorrect outcomes, localising the performance ceiling to tool-suite coverage rather than reasoning inconsistency. These results establish a methodological template for evaluating agentic AI against objective ground truth in the natural sciences and position the framework as an interpretable curation layer for late-stage pose refinement in structure-based drug design.