A systematic NMR-characterized dataset of mutants of the GA/GB model fold-switching system is presented and it is found that this benchmark revealed variable and position-dependent performance across methods, with certain AlphaFold2-based algorithms able to predict mutant effects at individual sites, indicating some understanding of physical effects of residue substitutions.
Abstract
Many proteins are known to adopt multiple distinct folded states which are often associated with key functional behavior. A predictive understanding of the properties of such fold-switching or metamorphic proteins can provide insights into protein dynamics and energetics, and enable the design of complex protein functions and molecular machines. Recently developed deep learning modeling tools, including AlphaFold, have led to dramatic increases in accuracy for prediction of protein structures from sequence, but their performance for prediction of point mutant effects or fold switching is unclear. Here we present a systematic NMR-characterized dataset of mutants of the GA/GB model fold-switching system and use it to evaluate whether current structure prediction and design methods can predict mutation-induced changes in fold state. We measured fold-state populations for variants at three key positions that differentially stabilize the 3α, 4β+α, mixed, or unfolded states, generating a quantitative experimental benchmark for mutation-level fold switching. Using this benchmark to assess and compare a panel of deep learning and physics-based modeling and design algorithms, we found that this benchmark revealed variable and position-dependent performance across methods, with certain AlphaFold2-based algorithms were able to predict mutant effects at individual sites, indicating some understanding of physical effects of residue substitutions. Additional comparisons of predictions with experimentally measured stability changes further highlighted position-dependent success and general challenges for predictive algorithms. Together, this study provides a new benchmark for mutation-induced fold switching and reveals the current capabilities and limitations of deep learning models for predicting mutation-dependent protein conformational states.
UniStab is introduced, an end-to-end framework for predicting stability changes across all mutation types by leveraging the implicit geometric reasoning of a pre-trained folding model and demonstrates state-of-the-art performance, particularly in the challenging scenarios of multi-point mutations and indels.
Hong Tan, Shenggeng Lin, Yi Xiong· Chemical Science· 0 citations
Abstract Motivation Proteins rely on conformational flexibility for biological function, yet predicting alternative states remains a major challenge in structural biology. Although deep learning models like AlphaFold2, AlphaFold3, and RoseTTAFold2 excel at static structure prediction, what these networks actually learn about the underlying conformational landscapes remains largely opaque. Results Here we introduce synthetic multiple sequence alignments (MSAs), designed by inverse folding to encode predefined structural constraints, as a programmable intervention for interrogating the internal logic of structure prediction systems. Synthetic MSAs systematically bias AlphaFold2, AlphaFold3, and RoseTTAFold2 toward distinct conformational states of fold-switching proteins, including alternative conformations inaccessible through natural sequence information alone. Adversarial experiments pairing query sequences with MSAs encoding competing folds reveal sequence-dependent responses, exposing how alignment-derived and sequence-derived signals are weighted within each system. Probing predictions initialized from molecular dynamics trajectories reveals a systematic bias toward compact, training-distribution-favored conformations. Hybrid alignments combining synthetic and natural MSA segments enable targeted steering toward specific conformational states. These results suggest synthetic MSAs as a generalizable framework for dissecting the conformational landscapes encoded by deep learning structure predictors, with direct implications for understanding model behavior and accessing biologically relevant hidden states. Availability and implementation Newly generated data can be found at https://zenodo.org/records/20916910. The code underlying this article is available on GitHub at https://github.com/ibmm-unibe-ch/msa-tests.
J. Gut, Noah Kleinschmidt, Thomas Lemmin· Bioinformatics· 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
Designing mutations that enhance protein stability is a central goal in protein engineering. However, experimentally screening large numbers of candidate mutations is costly and time-consuming, creating a strong need for computational methods that can identify potentially stabilizing mutations. Among these approaches, protein language models are particularly promising because they learn context-dependent amino acid preferences from large-scale sequence and structure datasets. Nevertheless, most existing stability prediction methods use these models primarily as feature extractors and do not fully exploit the amino acid probability distributions they encode. Here, we introduce MAXWELL (Matrix-wise Landscape Learning), a novel post-training method that calibrates the probabilistic outputs learned by protein language models during pretraining to generate mutational landscapes that quantify the effects of individual amino acid substitutions on protein stability. When applied to ProteinMPNN, MAXWELL yields a state-of-the-art predictor of the effects of protein mutations on stability, outperforming ThermoMPNN and other representative methods on a curated benchmark of experimentally measured stability changes. We next applied MAXWELL to the design of ten single-point mutations in the DhaA dehalogenase, seven of which (70%) increased thermal stability. Among them, G171W showed the largest improvement, with a measured ΔTm of 4.91 °C. These experimental results establish MAXWELL as a novel post-training strategy for protein language models and a practical framework for designing stabilizing mutations. Repository https://github.com/ai4protein/Venus-MAXWELL
Mingchen Li, Xiaoran Cheng, Fan Jiang et al.· bioRxiv· 0 citations
Three modeling frameworks are developed, including models based on handcrafted features, models using embedding representations extracted from ProteinMPNN, and ensemble models integrating a diverse set of state‐of‐the‐art predictors integrating a diverse set of state‐of‐the‐art predictors.
Yang Liu, Jian Zhang, Minghui Li· Protein Science· 0 citations
This framework provides a clearer understanding of how methodological shifts have shaped the capabilities, limitations, and practical roles of recent models.
Wengan He, Yongsheng Luo, Lihong Jiang et al.· 0 citations