protein sequence

emb2dis predicts protein disorder using a deep learning model combining ResNets and dilated convolutions, achieving top performance on benchmark datasets.

Baktfold improves protein functional annotation for microbial genomes using structural information, achieving higher annotation rates than existing tools for bacteria and archaea.

This study shows that residue burial information can predict protein backbone conformation more efficiently than other representations, suggesting core identity is a better encoding for fold prediction.

IDBSpred predicts IDP-binding sites on structured proteins using ESM-2 embeddings and a multilayer perceptron, achieving an ROC AUC of 0.87.

The paper designs lanthanide-binding mini-proteins with enhanced luminescence using non-canonical amino acids and computational methods.

This paper introduces BLAST, a light-controlled protein secretion system that enables reversible and precise protein release from the endoplasmic reticulum.

eSIG-Net predicts how single mutations affect protein interactions using a sequence-based interaction language model, outperforming existing methods.

The paper develops a transformer model to predict codon sequences from amino acids, improving rare codon prediction and linking sequence variation to protein function.

EnzyGen2 co-designs protein sequences and structures with ligand-guided functionality, outperforming existing methods in in silico benchmarks and generating active enzymes with high novelty.

This paper introduces a symmetric self-play framework for protein inverse folding that improves sequence design self-consistency by decoupling structural objectives.

DualLoc improves protein subcellular localization prediction by using full-parameter fine-tuning of dual transformers, achieving better accuracy for multi-compartment localization.

This paper introduces StructureDCA, a method that improves mutational landscape predictions by integrating structural information into direct coupling analysis.

GYDE provides an open-source platform for bench scientists to access AI tools for protein design and drug discovery through a collaborative interface.

Phyling is a tool for phylogenetic inference from protein sequences, offering faster performance than OrthoFinder with comparable accuracy.

The paper benchmarks four AI tools for designing BMPR1A-binding peptides, identifies top candidates, and highlights differences in binding site targeting.

ProteinMCP accelerates and democratizes protein engineering by automating workflows and integrating 38 tools, achieving a comprehensive fitness modeling task in 11 minutes.

The paper introduces SelectZyme, a platform using protein language models and dimensionality reduction to visualize and explore protein sequence space for enzyme discovery, enabling candidate selection without relying on sequence identity or annotations.

CombinGym benchmarks machine learning for combinatorial protein variant design using curated datasets and evaluates model performance across multiple tasks and data splits.

The paper analyzes SNARE transmembrane domains to show how their composition adapts to different membrane environments, with early pathway SNAREs having more Phe and late pathway SNAREs having more Ile.

The paper identifies and analyzes human intrinsically disordered regions (IDRs) in transcription factors, revealing their unique functional, phenotypic, and evolutionary characteristics compared to non-TF IDRs.

The paper introduces DIAMOND DeepClust, a fast clustering method for protein sequences, identifying 1.7 billion clusters across 19 billion sequences.

This study identifies and classifies over 400 metallothioneins (MTs) in chelicerates, revealing three types with distinct structural and evolutionary features.

VPF-Class 2.0 improves viral classification by using a supervised model on marker-composition features, achieving high accuracy and interpretability.

A ProtT5-embedding, ConvNext-plus-attention ensemble predicts human pathogenic capacity from whole bacterial genomes, outperforming prior predictors on species absent from training and flagging influential proteins.

PepBERT is a lightweight peptide language model that outperforms ESM-2 on most peptide-related tasks, enabling better representation of bioactive peptides for downstream applications.

ST-PARM improves multi-objective protein design by aligning a frozen language model with user-specified trade-offs using a reward model that handles uncertainty and achieves broader Pareto coverage.

evedesign provides a unified framework for conditional biosequence design, enabling multiobjective workflows and experimental integration across various applications.

This study uses a conditional variational autoencoder to design new-to-nature PHA synthases, which are validated for activity in yeast and show potential for expanding PHA material properties.

This study develops a baculovirus-mediated method to express PIEZO1 in muscle satellite cells, restoring their proliferative capacity and enabling functional studies of large proteins.