Protein Structure and Dynamics

This study evaluates how input format and charge affect protein-ligand structure predictions using four co-folding methods, finding format has a stronger impact than charge.

AlphaFast accelerates AlphaFold 3 by using GPU-accelerated MSA construction, achieving significant speedups and cost reductions while maintaining structural accuracy.

This paper develops a scalable NMR pipeline to characterize protein structure and dynamics, finding that designed proteins adopt intended folds and exhibit unexpected dynamics not captured by current models.

ProtFlow generates high-quality protein sequences by learning the full semantic distribution of protein space, outperforming existing methods in generating antimicrobial peptides with broad activity.

The paper improves protein complex structure prediction by integrating ion mobility mass spectrometry data with AlphaFold and Rosetta, achieving better RMSD values for many complexes.

This paper redesigns dsRBD domains to inhibit DHX9 helicase activity by targeting its autoregulatory interface, offering a computationally efficient strategy for therapeutic development.

The paper investigates how solvent isotopes affect protein stability using double-chain monellin, finding that wild-type protein is more stabilized in D2O due to hydrophobic interactions.

IntelliFold-2 improves protein structure prediction accuracy and robustness over AlphaFold 3 using architectural refinements and structural consistency, with strong gains in antibody-antigen and protein-ligand interactions.

This paper introduces NbForge, a nanobody folding model that improves prediction of HCDR3 blueprint and disulphide bonds, achieving accuracy comparable to large predictors with faster speed.

The paper maps a bidirectional allosteric pathway in Arf GTPases that controls access to a functional molten globule state through the GDP ligand, rather than directly affecting switching rates.

The paper uses FTMap to analyze 50 antibody-antigen complexes, finding that paratopes are hot spot-rich and that Trp/Tyr residues drive hot spot formation.

This paper introduces a method for fast de novo protein design using a structure-inspired alphabet and Metropolis sampling to generate sequences without known homologues.

Metadiffusion introduces a meta-energy biasing layer to diffusion models to generate diverse biomolecular conformations without retraining, matching molecular dynamics simulations.

The paper shows that fluctuating selection pressures improve coevolution-based inference of protein structure and function from sequence data.

The paper introduces a pattern-matching strategy to predict protein contact maps from sequences, achieving high correlation with experimental data using minimal computational resources.

This study uses molecular dynamics simulations to show that VISTA's conformational switching between Up and Down states regulates its immune checkpoint function through CC' loop accessibility.

This study introduces trainable subnetworks to analyze how protein language models factorize structural information and shows that structure prediction is sensitive to these factorized representations.

MemConverter reprograms protein localization between membrane and aqueous environments using MCP-guided sequence design and AlphaFold2 refinement, achieving stable integration or solubility.

This paper investigates the temporal dynamics of discrete protein diffusion models, revealing a decoupling between structural alignment and property optimization, and identifies a structural collapse regime due to Coulombic repulsion.

The paper uses molecular dynamics simulations to show how minor sequence changes in NarU enable symport instead of antiport compared to NarK.

This study investigates the conformational flexibility of staphylokinase variants to explain their reduced immunogenicity and altered properties.

This study evaluates seven single-point mutants of β-glucosidase B using computational design and experimental analysis, finding weak correlations between predicted thermodynamic stability and enzyme properties.

This paper introduces bAIes, a Bayesian framework for generating accurate atomic-resolution ensembles of intrinsically disordered proteins using AlphaFold2 and molecular mechanics, showing improved accuracy over existing methods.

The paper introduces FELS, a method to screen free energy landscapes for disordered proteins, and applies it to tropoelastin, revealing its conformational ensemble and resolving conflicting models.

The paper evaluates if AlphaFold distograms can predict binding-induced hinge motions in proteins, showing they capture flexibility not seen in predicted structures.

The paper introduces AtomPaint, a full-atom diffusion model for designing protein/non-protein binding interactions, showing it outperforms random guessing in sequence recovery.

The paper introduces MARS, a simulation framework that resolves conformational heterogeneity in IDPs by integrating experimental data from SAXS and PRE, revealing two major states in ER-NTD.

MViewEMA improves protein complex model accuracy estimation with multi-view representation learning, achieving higher efficiency and performance than previous methods.

The paper shows that peptides made from non-standard amino acids can form protein-like secondary structures, expanding the known sequence-structure space beyond Earth's 20 amino acids.