It is demonstrated that BioEmu can generate plausible conformational ensembles for relatively large, six-and seven-pass membrane proteins, sampling rare states at a fraction of the computational cost of conventional MD simulations, suggesting that AI-based ensemble generation could provide an accessible approach for exploring membrane protein dynamics and complement conventional molecular modelling approaches.
Abstract
Proteins dynamically switch between a continuum of interconverting conformational states, and understanding these structural dynamics is important for understanding protein function and for developing therapeutics. Molecular dynamics (MD) simulations can provide insight into protein conformational ensembles, but sampling rare conformational states can require substantial computational resources. The recent development of AI-based approaches for generating protein conformational ensembles, such as the Biomolecular Emulator (BioEmu), offers a potential alternative, although it remains unclear whether these approaches can accurately capture the conformational landscapes, especially for special cases such as membrane proteins. Here, we assess the ability of BioEmu to model the conformational dynamics of a model membrane protein, the bacterial rhomboid intramembrane proteases GlpG. We find that BioEmu generates a range of conformations corresponding to both open and closed states of the rhomboid lateral gate, including states associated with different stages of the catalytic cycle. These conformations broadly correspond to states sampled during microsecond-timescale MD simulations, although BioEmu does not reproduce the full conformational landscape observed using MD. BioEmu also samples substantial conformational heterogeneity within the soluble domains of rhomboids, which are highly flexible and poorly represented in experimental structures. Overall, our findings demonstrate that BioEmu can generate plausible conformational ensembles for relatively large, six-and seven-pass membrane proteins, sampling rare states at a fraction of the computational cost of conventional MD simulations. These results suggest that AI-based ensemble generation could provide an accessible approach for exploring membrane protein dynamics and complement conventional molecular modelling approaches.
Pi-Ensemble (Predicting Interpolated Ensemble), a sequence-guided framework for generating protein conformational ensembles interpolating between two structural anchor states, provides an extensible framework for studying protein flexibility, guiding adaptive sampling, and accelerating mechanistic investigations of protein function.
Hassan Nadeem, D. Kleiman, Yuming Zhou et al.· bioRxiv· 0 citations
Proteins in solution often populate conformational ensembles that differ from the static states captured by crystallography or AI-based structure prediction. Conventional molecular dynamics (MD) simulations often fail to cross the energy barriers separating these states on accessible timescales, and statistical reweighting cannot recover conformations never sampled. Here we present Carbonara, a framework that uses experimental small-angle X-ray scattering (SAXS) data to predict alternative physically plausible protein conformations. Carbonara builds on Wiggle, a standalone Cα-based SAXS forward model validated against explicit-solvent calculations and experimental benchmarks. Using two case studies, an AI-predicted multi-domain helicase (SMAR-CAL1) and a crystallographic antibody fragment (ChiLob7/4 IgG2), we show how seeding MD simulations from Carbonara conformations enables efficient exploration of solution-state conformational landscapes. In both cases, MD ensembles initiated from available models either fail to match the SAXS data or do so only after discarding nearly all sampled conformations, whereas Carbonara-seeded ensembles reach agreement while retaining the majority of conformations. Our modelling framework provides a route from static structural models of flexible multi-domain proteins and multimeric assemblies to solution-state ensembles.
Together, these insights position conformational dynamics at the center of understanding and engineering the evolutionary logic of protein function, opening the door to study how proteins are tuned to operate under the nonequilibrium conditions of living cells.
Sixto M. Herrera, Elías Manríquez-Benítez, Exequiel Medina· Current Opinion in Structura...· 0 citations
Cyclic peptides are a promising therapeutic modality, offering the potential to target challenging intracellular protein-protein interactions involved in cancer and other diseases. However, their clinical utility is frequently restricted by poor membrane permeability. While deep learning offers new methodologies to predict permeability, current models are limited by a reliance on 2D molecular representations that fail to capture the conformational flexibility inherent to macrocycles. Existing 3D resources also lack physics-based sampling of conformational dynamics across solvent environments that are critical for membrane permeability. To bridge this gap, we present CycPeptMPDB-4D, a comprehensive dataset comprising atomistic molecular dynamics trajectories for 5,160 structurally diverse cyclic peptides, including unnatural, N-methylated, and D-residues in circle and lariat topologies. Each peptide was simulated using the AMBER14SB force field in both explicit water and hexane environments for 50 nanoseconds to generate conformational ensembles in aqueous and membrane-mimicking phases. The trajectories capture the “chameleon-like” property, evidenced by markedly reduced conformational flexibility and polar surface area in the hydrophobic phase. Technical validation demonstrates that the simulated ensembles are in high agreement with experimental NMR data, covering NMR conformers within an RMSD of 1.6 Å. The dataset provides clustered ensembles, representative structures, and specialized descriptors such as desolvation free energy. This resource is designed to facilitate the development of deep learning models that incorporate 3D or 4D (trajectory- or ensemble-based) information to improve the prediction of cyclic peptide membrane permeability.
Wei Liu, Nguyen Hung Pham, Chandra S. Verma et al.· Scientific Data· 0 citations
It is argued that, since physics-based simulations and machine learning provide complementary approximations to the underlying probability distribution associated with biomolecular recognition events, and they excel respectively in consistency with free-energy landscapes and state populations and in predictive accuracy, the central challenge for the coming decade will be integrating them into hybrid frameworks that are scalable and transferable.
R. Khalil, Elena Frasnetti, Han Kurt et al.· Journal of Physical Chemistr...· 0 citations
This work assesses ML potentials for exploring RNA conformations using the adenine–adenine dinucleoside monophosphate (ApA) dimer, a fundamental RNA building block, and parametrized ML potentials based on the equivariant MACE architecture and informed by both ab initio and semiempirical property data.
Leonardo Medrano Sandonas, Macarena Tolmos Nehme, L. F. Cofas-Vargas et al.· Journal of Chemical Theory a...· 0 citations