Gen2-Allostery: A Replicate-Aware Framework for Evaluating Allosteric Binding-Site Hypotheses

Allosteric binding-site hypotheses are easy to propose computationally but much harder to evaluate rigorously after dynamical relaxation. Docking, pocket detection, and short simulations can each produce visually plausible non-orthosteric poses, yet such poses often fail to remain spatially coherent, mechanically stable, or reproducible under unbiased molecular dynamics. Here, we present Gen2-Allostery, a replicate-aware framework for evaluating whether a proposed allosteric binding-site interpretation remains supportable under predefined and falsifiable criteria. The framework is designed as a bounded pre-experimental decision layer rather than a discovery engine, affinity predictor, or replacement for experiment. Across structurally and mechanistically distinct test systems, Gen2-Allostery recovered reproducible control-like behavior in validated settings while also identifying cases in which apparent pocket engagement was replicate-sensitive, non-persistent, or more consistent with an alternative site interpretation. These results show that docking plausibility alone is not sufficient evidence for a defensible allosteric binding narrative. Instead, site-focused dynamical evaluation, replicate-aware interpretation, and explicit consideration of competing explanations are needed before downstream experimental investment. Within this scope, Gen2-Allostery provides a practical computational framework for distinguishing supported, rejected, and unresolved allosteric binding-site hypotheses under the tested conditions.