protein sequence

The paper introduces AtlasFold, a protein structure prediction method using metagenomic-scale language models, achieving state-of-the-art accuracy and efficient inference.

EMAP-SSN integrates sequence similarity network computation, visualization, and analysis to explore protein sequence space and identify functional motifs and conservation patterns.

The paper shows a sequence-only pipeline using TabPFN achieves high accuracy in multi-activity antimicrobial peptide profiling, outperforming complex deep models.

The paper introduces IMM-AGE, a metric of immune-aging, and shows it independently predicts cardiovascular disease risk across multiple cohorts.

The paper introduces a geometric framework to analyze protein sequence space coverage and connects it to heuristic discovery times using a stochastic exploration model.

The paper designs mutational paths for WW domains to switch specificity using computational methods and validates them experimentally, showing high success rates for structured paths over random ones.

The paper uses spatial multi-omics and single-cell transcriptomics to identify senescence-associated programs in colon adenoma to cancer progression, finding GDF15 mediates senescence-stemness coupling and immune exclusion.

CatESO optimizes enzyme sequences to improve catalytic turnover by integrating kcat prediction with differentiable design, achieving higher predicted activity and structural confidence than existing methods.

The paper introduces a sparse feature circuit framework to identify causal features in ESM-2 for zero-shot mutation effect prediction, showing that shared circuits explain mutation effects through 3D contact residues.

PocketScope retrieves binding sites across the AlphaFold human proteome using protein language model representations and achieves top performance on a benchmark.

The paper annotates neuropeptides and GPCRs in Gryllus bimaculatus and predicts ligand-receptor pairs using AlphaFold3 and Boltz-2, providing a resource for studying insect endocrinology.

This paper demonstrates that large language models can generate de novo protein sequences that bind specific ligands without copying existing sequences, with a 25% hit rate in experimental validation.

This study evaluates the impact of pre-trained protein language models on drug-target affinity prediction, showing that simple architectural changes can match large PLMs' performance.

This paper introduces SLIDE, a method to estimate landscape ruggedness in directed evolution using population-level phenotypic data and mutation rate, improving selection strategies for protein design.

This paper uses deep learning to design bacteriophage receptor binding proteins with multiobjective specificity, showing high tunability and success rates in targeting diverse strains.

The paper analyzes NLRP3 evolution and finds eutherian NLRP3 has unique regulatory features absent in non-eutherian NLRP3-like proteins.

This study shows that ENPP3 expression in HER2-positive breast cancer cells is linked to better prognosis by inhibiting epithelial-to-mesenchymal transition.

PathFold predicts protein folding pathways from sequence using a deep learning framework combining AlphaFold and diffusion models, showing agreement with experimental data.

The paper develops a high-quality dataset for protein sub-cellular localisation prediction, showing performance improvements and highlighting data leakage issues in existing models.

RegimeFormer identifies protein perturbation regimes using a large-scale model and atlas, improving prediction accuracy and enabling downstream applications.

Web resource offering predicted monomeric and oligomeric structures for 1,603 M. leprae proteins, with pockets, epitope scores, essentiality and a target priority ranking, plus a nanopore resistance-calling module.

FuncSeek improves enzyme function prediction by using contrastive learning with multiple PLMs, outperforming existing methods on benchmark datasets.

DyAb predicts antibody properties and designs new sequences with high binding affinity using a low-data approach.

This study shows disease-linked mutations in hnRNPA2B1 accelerate condensate maturation and promote amyloid aggregation by altering phase separation and fibril formation.

FigTreeKit provides a Python toolkit for programmatically styling and auditing phylogenetic trees, with scalability testing on large bacterial and archaeal datasets.

PMPNN-DDG is a Random Forest model predicting single-point mutation-induced protein stability changes using features from ProteinMPNN, achieving high accuracy on independent test sets.

A tiered catalogue of 5,280 Drosophila genes encoding putative secreted proteins (1,802 high-confidence) with a web portal; these genes are less conserved, more tissue-restricted and disproportionately down-regulated with age.

AntiSite predicts antibody paratopes using a single model that works with or without structural data through modality dropout.

This paper reconstructs a pan-genomic Tree of Life using protein domain combinations, revealing evolutionary relationships and processes across Bacteria, Archaea, and Eukaryota.

The paper shows that unobserved protein sequence space contains many functional proteins, challenging the assumption that existing proteins cover most functional space.