Protein Structure and Dynamics

Across 8,127 cellular and viral proteomes, ordered SCOP domains originated first in a phylogenomic chronology; highly disordered domains emerged late, and viruses dominate highly disordered proteomes.

ESM3, a 98-billion-parameter multimodal protein language model tokenized across sequence, structure, and function, designs a bright green fluorescent protein at 58% sequence identity to known fluorescent proteins.

An open AlphaFold3 reproduction, with weights, inference and training code, that matches AF3 on ligand and RNA benchmarks, beats AlphaFold-Multimer 2.3 on protein interfaces and flags possible memorisation.

Open-source Boltz-1 predicts biomolecular complex structures comparably to AlphaFold3 and Chai-1 on CASP15 and recent PDB targets; Boltz-steering lets 97% of test-set poses pass physical checks.

Hallucinating binders by backpropagating through AlphaFold2 weights gave, across 12 targets, binders in 10-100% of the designs tested per target, mostly at nanomolar affinity.

ProstT5, a ProtT5 fine-tuned to translate between amino acids and Foldseek 3Di tokens, predicts structure strings for remote homology search and performs inverse folding.

Likelihoods from Progen2 and ESM2 protein language models are systematically higher for sequences from some species, tracking evolution-weighted database sequence counts, and pull likelihood-guided designs away from extremophile properties.

A Local Interaction Score (LIS), computed from AlphaFold-Multimer predicted aligned error, separates curated fly and human protein interactions from controls better than ipTM, recovering small or flexible interfaces.

ProteinGym unifies over 250 deep mutational scans and ~65k clinical variants to benchmark over 70 models in zero-shot and supervised settings; TranceptEVE ranks best overall.

A Foldseek-based linear-time structural clustering groups 52 million AlphaFold DB representatives into 2.27 million non-singleton clusters; 31% lack known structural or domain annotation.

With active-state templates and shallow ortholog alignments, AlphaFold2 models all 437 human catalytic kinase domains as active; 80% of 130 benchmarked kinases are under 1 Å activation-loop RMSD.

Fine-tuning RoseTTAFold as a diffusion denoiser gives RFdiffusion, a general protein backbone generator; a designed influenza haemagglutinin binder matched its model by cryo-EM.

Foldseek's 3Di tertiary-interaction alphabet enables structure search four to five orders of magnitude faster than Dali and TM-align at 86% and 88% of their respective sensitivities.

A sequence-only protein language model, sampled under a structure-compatibility energy, produced designs of which 152 of 228 tested behaved as soluble monomers, some with no detectable UniRef90 match.

PyTorch reimplementation of AlphaFold and AlphaFold-Multimer with training code, retrained from scratch to equal or better accuracy, about 2.2 times faster training and roughly 2% higher multimer TM-Score.

From a single sequence, OmegaFold reaches mean LDDT 0.82 on 146 CAMEO targets, between MSA-based RoseTTAFold (0.75) and AlphaFold2 (0.86), and runs about ten times faster.

ColabFold couples 40–60-fold faster MMseqs2 homology search to AlphaFold2, matching it on CASP14 and AlphaFold-multimer on ClusPro complexes, yielding close to 1,000 structures daily per GPU.

Adding 12 million AlphaFold2-predicted structures to training lifts native sequence recovery on held-out backbones to 51.6%, versus 42.2% for the best experimental-only model; likelihoods score mutation effects zero-shot.

AlphaFold2, fed concatenated chains with optimized organism-paired alignments and pDockQ-ranked seeds, gives acceptable models for 63% of heterodimers; the same predicted DockQ score flags interacting pairs.

AlphaFold trained on complexes, with cross-chain paired alignments and symmetry-aware losses, gets 70% of recent heteromeric interfaces acceptable or better (DockQ ≥ 0.23), 27 points above linker-joined AlphaFold.

Splitting AlphaFold2 complex models, re-docking with ClusPro, AlphaFold2 refinement and predicted-aligned-error ranking raised successful docking-benchmark targets from 88 (five AlphaFold2 models) to 123 (top 10) of 204.

AlphaFold, an attention network over sequence alignments and residue pairs, rivaled experimental accuracy in most CASP14 cases: median backbone r.m.s.d.95 0.96 Å versus 2.8 Å for the runner-up.

Attention maps of the 650M-parameter protein language model ESM-1b, read out by a logistic regression fitted on 20 proteins, predict contacts better on average than Gremlin's alignment-based Potts models.

Graph convolutions on atom graphs with distance, orientation and bond edge features predict per-residue and global lDDT of protein models, outperforming VoroMQA and retrained Ornate while using about 100k parameters.

Cascaded CNNs segmenting secondary structure, backbone and Cα voxels, plus tabu-search tracing, place 88.5% of Cα atoms within 3Å across 50 cryo-EM maps versus Phenix's 66.8%.

A 1.2B-parameter autoregressive model trained on 280M tagged sequences completes held-out proteins with near-native Rosetta energies and ranks high-fitness GB1 variants without any GB1 training.

NetWheels, an R/shiny web application, draws customizable peptide helical wheel and net projections, accommodates non-α helix geometries, and exports publication-quality images plus reloadable parameter sets.