Protein Structure and Dynamics

GeoFlow-V3 designs nanobodies de novo with high hit rates, reducing experimental screening needs and improving antibody discovery efficiency.

The paper introduces TF-MAPS, a platform to map functional and allosteric sites of DNA-binding proteins, showing distance-dependent allostery and druggable pockets in three human TFs.

The paper introduces a benchmark for evaluating protein ensemble generators across ordered and disordered proteins and presents a new model that improves performance on disordered regions.

This paper introduces ProteinZen, a method for all-atom protein design using SE(3) flow matching, achieving high sequence-structure consistency and competitive performance in design tasks.

This paper introduces a constrained diffusion framework for protein design that enforces strict structural constraints while maintaining structural diversity, achieving state-of-the-art results on motif scaffolding and pocket design tasks.

Odyssey generates and designs proteins using a multimodal language model with a consensus mechanism, achieving strong performance in protein generation and structure discretization.

This study shows that hydrophobic asymmetry in non-charged protein condensates creates electric potential gradients, enabling potential-dependent chemistry.

GhostFold generates synthetic multiple sequence alignments (MSAs) from single sequences to improve protein structure prediction, achieving accuracy comparable to existing methods without relying on homology searches.

This study investigates how isoelectric point (pI) influences evolutionary residue selection and structural stability in JAM paralogs through integrated analysis of conservation, dynamics, and pathogenicity.

This paper shows how AlphaFold3 can be guided by experimental data to predict accurate protein ensembles, including alternate conformations.

This study shows that circular permutants of azurin exhibit molten globule states not observed in wild-type azurin, suggesting CP can modulate protein energy landscapes.

The paper constructs a structure-guided kinase-transcription factor interactome atlas using AlphaFold-Multimer predictions, revealing docking motifs and allosteric hotspots that regulate kinase activity and lipid metabolism.

This paper introduces DiffTopo, a diffusion model for generating novel protein folds, and MirrorTopo, a pipeline to expand known fold space by mirroring native protein topologies.

Protein Hunter uses diffusion models to design proteins de novo by iteratively improving sequence and structure through hallucination and re-design.

The paper develops a framework to assess the quality of TCR-pMHC structural models without experimental references, demonstrating its effectiveness in expanding structural coverage and improving model evaluation.

The paper reveals how thermodynamic principles shape the biophysical properties of proteins through allosteric regulation and evolutionary tradeoffs.

FlexRibbon is a pretrained protein model that jointly learns from sequences and structures, achieving state-of-the-art performance in mutation-rich settings.

This paper introduces ProteusFold, a compact and interpretable protein folding model that achieves state-of-the-art accuracy with high efficiency and provides residue-level and atomic-level attribution analyses.

The paper introduces a model to study how proteins transmit information over long distances without structural changes, finding that discrete motion enables this despite thermal noise.

This work introduces an iterative sampling framework using coevolutionary information and AlphaFold to improve the diversity of sampled protein conformations.

This study develops area-based methods to select near-native interfaces for protein-protein complex structure prediction, showing high success rates in interface selection.

The paper introduces a compositional model to predict protein interactions, showing that functional group composition can reliably predict survivin binding without relying on amino acid sequences.

This study designs miniprotein binders targeting viral entry proteins of Chikungunya and Dengue viruses using computational methods and evaluates their stability through molecular dynamics simulations.

SLAE learns all-atom protein representations from local atomic neighborhoods, achieving state-of-the-art performance in downstream tasks through transfer learning.

The paper identifies a conserved motif linked to β-sheet oligomerization in EAA1 isoforms through evolutionary analysis of helix-to-sheet transitions across species.

This paper uses protein language models to discover new-to-nature protein domains with structures outside known CATH superfamilies, suggesting natural protein structure-space is incomplete.

This paper introduces GCP-VQVAE, a geometry-complete tokenizer for protein structures, achieving state-of-the-art performance on benchmark datasets and demonstrating robust generalization to unseen proteins.

MSACLR uses contrastive learning to map multiple sequence alignments to protein conformations, achieving better fold separation and capturing conformational variation.

The paper introduces Kanzi, a flow-based tokenizer for protein structures that simplifies training and outperforms existing tokenizers on reconstruction metrics.

This paper shows that protein language models prioritize sequence novelty over structural diversity, leading to a tradeoff in generating realistic protein structures.