Mavchen-1: A Conformational Ensemble Platform for Protein–Ligand Pose Prediction That Substantially Outperforms Static Structure Prediction in a Category-Stratified Benchmark

Benchmarking AI Protein Structure Predictors Reveals a Persistent Bias in Multi-State Proteins

Protein structure predictors achieve high single-state accuracy, but it remains unclear whether they can recover functionally relevant conformational ensembles or account for the presence of ligands and/or binding partners. Here, we benchmark AlphaFold3, Boltz-2, Chai-1, and BioEmu on four canonical multi-state proteins (Pf-MATE, LAO, SecA, and β2AR), quantifying state bias and sampling breadth against experimental reference structures. Models frequently default to a dominant state represented in the PDB; small-molecule ligands have weak or inconsistent effects, while large protein partners drive clear conformational switching between states. Multiple sequence alignment (MSA)-based approaches (AF-Cluster and random subsampling) recapitulate similar biases, indicating that this behavior is not unique to newer architectures. These results underscore current limitations for multi-state protein structure prediction and structure-guided ligand discovery. TOC Graphic

MultiGeo: Predicting Drug-Target Affinity via Adaptive Multi-Conformation Ensemble Learning

Accurate structural modeling of chemically diverse molecular interfaces with Vilya-2

Structure-prediction networks built on co-evolutionary statistics have transformed protein-based drug discovery, yet their accuracy does not extend to peptide therapeutics--an increasingly important modality defined by non-canonical residues, macrocyclization, and complex topologies. We introduce Vilya-2, a diffusion transformer that extends the all-atom representation of Vilya-1 from modeling individual molecules to modeling their interactions with protein targets. This all-atom representation enables transfer learning between different molecular types, and delivers highly accurate structural modeling of peptides across sizes, classes, and compositions bound to therapeutically relevant targets. By generating diverse structural ensembles and ranking them with calibrated confidence, Vilya-2 recovers 59.1% of peptide interfaces to sub-2 {\AA} backbone RMSD, far exceeding the performance of a representative co-folding model even when that model is given the bound receptor as a template. In addition, Vilya-2 is state-of-the-art at small-molecule docking, and generalizes to novel protein-small molecule complexes unlike those seen in training. It also generalizes to modeling molecular conformations of diverse macrocycles and disulfide-stapled miniproteins several-fold larger than any molecule seen in training. Finally, Vilya-2 can be used as a foundation model, and fine-tuned to enrich for active compounds in hit-to-lead campaigns. By unifying predictive accuracy with broad generalizability across chemical space, Vilya-2 is the structure-prediction oracle that de novo peptide design pipelines require--establishing the all-atom approach as a general foundation for the design and evaluation of de novo peptide therapeutics.

AlphaFold-based peptide structure prediction: Opportunities, limitations, and future directions.

Accurate prediction of peptide structures and peptide-receptor complexes is essential for rational peptide drug development. However, the inherent conformational flexibility of short and disordered peptides presents a fundamental challenge. The AlphaFold model series, which has progressed from AlphaFold2 through AlphaFold-Multimer to AlphaFold3, has substantially advanced computational peptide structure prediction through innovations in geometric reasoning (invariant point attention) and interface-focused confidence metrics (ipTM score), achieving high accuracy for both monomeric peptide structures and multi-chain complexes. However, these models output static conformations, whereas many bioactive peptides adopt their functional conformations only upon binding-often corresponding to low-probability states that static predictions may overlook, leading to failures in virtual screening. This review synthesizes recent advances in the AlphaFold series for peptide studies and applications, discusses their current strengths in structure prediction and receptor-binding analysis, and examines the limitations in capturing conformational dynamics, transient interactions, and chemical modifications. Recent studies have suggested that integrated computational strategies that combine AlphaFold predictions with molecular dynamics simulations, free energy calculations, and ensemble sampling to enhance predictive accuracy and better represent the dynamic nature of peptide-drug interactions. These complementary approaches position AlphaFold as a central computational platform in structure-guided peptide drug design, enabling more efficient lead identification and optimization while bridging the gap between static computational predictions and the complex biophysical reality of peptide therapeutics.

Evaluating BioEmu-Generated Kinase Ensembles Reveals Structure Selection as the Virtual Screening Bottleneck

Virtual screening (VS) is an essential tool in drug discovery to prioritize potential drug candidates from vast chemical space. One key challenge limiting its performance is accounting for protein conformational flexibility. While ensemble docking methods have been developed to address this challenge by incorporating multiple protein conformations, these methods often rely on computationally intensive physics-based simulations to sample the relevant conformational space. Generative machine learning models offer a highly promising, scalable, and high-throughput alternative to overcome the limitations of these traditional approaches. We therefore investigate whether conformational ensembles generated by BioEmu, a recently developed generative model, can improve VS performance for kinase targets. Using the DUD-E benchmark data set and a validated AutoDock-GPU protocol, we generated and analyzed nearly 1300 structures across 26 kinases (approximately 50 structures each). BioEmu produces structurally diverse ensembles with substantial performance variation among individual structures. However, ensemble methods employing consensus or best-score selection fail to improve upon, and often degrade, VS performance compared to crystal structure baselines. To investigate the source of this limitation, we quantified the relationship between KinCoRe-based conformational state classification and screening performance. By calculating the coefficient of determination (R2) across the kinase subset, we found that the structural features governing VS performance differ substantially from those defining standard conformational states, with KinCoRe classifications leaving over 84% of performance variance unexplained. This critical gap demonstrates that structural diversity alone is insufficient to guarantee screening success. We show that prospective structure selection, rather than structure generation, represents the primary bottleneck in ensemble-based VS, highlighting an urgent need for novel structural descriptors to identify high-performing conformations.

Conditional molecular dynamics refinement for protein-ligand affinity prediction.

Protein-ligand affinity prediction is fundamental to structure-based virtual screening and lead discovery. However, most existing methods rely on a single static conformation of the complex and exhibit high sensitivity to pose uncertainty and conformational noise. To address this limitation, CMD-PLA is proposed as a dynamics-aware framework for protein-ligand affinity prediction. Within this framework, pocket-conditioned molecular dynamics refinement is performed, and the conformational evolution of the ligand is explicitly modeled as a pocket-dependent dynamical process rather than an isolated static update. Furthermore, a dual-view atomic representation is adopted to separately capture the intra-molecular covalent structure and the inter-molecular interaction geometry. Global representations of the ligand and the pocket are also incorporated to complement the local geometric modeling. Experimental results demonstrate that CMD-PLA achieves robust performance across various settings. The provided case study further illustrates the interpretability of the model.