Boltz-Perturb is presented, a framework for addressing small molecule binding poses through perturbing model conditioning signals during model inference, and it is demonstrated that inference-time perturbations can unlock latent structural diversity in generative co-folding models and improve protein-ligand predictions without costly retraining.
Abstract
Protein-ligand co-folding models hold promise in structure-based drug discovery and small molecule interaction prediction, but often fail in predicting correct small molecule binding poses. We present Boltz-Perturb, a framework for addressing this through perturbing model conditioning signals during model inference, and show that such perturbations improve correct ligand binding mode predictions. We first show with true-coordinate injection experiments that the model’s learned energy landscape contains correct binding-mode basins, allowing us to reframe the problem as one of sampling deficiency. We then introduce two inference-time perturbation strategies, Token Bias Perturbation (TBP) and Token Conditioning Perturbation (TCP), which increase exploration of alternative binding poses. Across diverse protein–ligand systems, TCP improves top-20 oracle success rates by 2.6 to 7.8 fold. Boltz-Perturb attains higher oracle success rates compared to the Boltz-2 high diffusion temperature variant while requiring over 75% less compute. To our knowledge, this is the first systematic perturbation analysis of a co-folding architecture for small-molecule binding mode diversity. We demonstrate that inference-time perturbations can unlock latent structural diversity in generative co-folding models and improve protein-ligand predictions without costly retraining.
Boltz is benchmarked using a curated set of ligand-bound human G protein-coupled receptors from families unseen during training, showing that while Boltz generally predicts receptor backbones accurately, ligand poses can contain significant errors that lead to a limited ability to reproduce experimental affinity data when tested with FEP+.
Lichirui Zhang, R. Friesner, Edward B. Miller et al.· npj Drug Discovery· 0 citations
A novel method named Perturbed Flow Matching (PFM), which significantly reduces sampling steps by leveraging a Flow Matching framework and introduces a unique perturbed conditional probability path design that incorporates pocket binding site information and atom type-coordinate coupled information to enhance molecular generation performance.
A category-stratified, statistically powered benchmark comparing pose prediction from receptor conformational ensembles against AlphaFold2, used as a matched static-structure baseline, across 29 protein–ligand systems spanning cryptic-pocket, induced-fit, water-mediated, and autoimmune-indication target classes is presented.
Ryan Varghese, Pooja Tiwary, Krishil Oswal· bioRxiv· 0 citations
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
Muhui Ye, Yu-Hong Wang, M. Brogi et al.· bioRxiv· 0 citations
The resulting model, HydrAffinity, is an interaction-free, dynamic sparse model that uses pre-trained encoders and MoE for parameter-efficient learning and outperforms all interaction-free methods and matches state-of-the-art interaction-based methods on CASF-2016.
The kinetics of protein-ligand binding systems are increasingly recognized as a key determinant of drug efficacy, yet remain far harder to compute than binding affinities. Existing kinetics methods either bias the dynamics along a collective variable (CV), demanding careful system-specific CV design, or use path sampling, which keeps the dynamics unbiased but can struggle to converge rates out of deep free-energy wells and often relies on hand-engineered descriptors. By combining the `best of both worlds', we propose a method to compute accurate kinetics for general ligand-unbinding problems at modest computational expense and minimal fine tuning, building on the AI for Molecular Mechanism Discovery (AIMMD) path sampling framework. To avoid the need for feature engineering, we opt for modelling the committor with a single descriptor-free, equivariant graph neural network shared across all systems. We also partially flatten deep bound-state wells with a static, basin-restricted bias potential. This improves convergence by lifting the path sampling state boundary out of regions, where the committor is hard to learn, while leaving the reactive region strictly unbiased. Across host-guest and protein-ligand systems spanning roughly 17 orders of magnitude in residence time, the method robustly recovers rates in line with reference and experimental values. Simultaneously, and without further sampling, it also reconstructs the underlying unbinding mechanisms. We additionally find that accurate rates do not require globally accurate committor models, allowing for efficient kinetics estimation even in a low-data training regime. Requiring little system-specific setup, our approach offers an efficient and broadly generalizable route to binding kinetics, and its shared committor architecture lays crucial groundwork for probing structure-kinetics relationships across ligand series in drug discovery.