protein sequence

The paper improves the GFmix model to better handle site-and-branch compositional heterogeneity in phylogenetic inference, showing improved accuracy in tree estimation.

PRISM enables sequence-independent, one-shot fine-tuning for simultaneous protein domain segmentation and classification, improving accuracy for complex and novel architectures.

This study investigates FimH conformational dynamics and catch bond formation using a single-domain variant, revealing ultra-slow dynamics and a dynamic allosteric mechanism.

This paper predicts antibody non-specificity using protein language models and biophysical parameters, showing that heavy variable domains and isoelectric point are key predictors.

GraphTox predicts peptide toxicity using a geometric graph transformer and LORA-based finetuning, outperforming existing methods by incorporating structural information.

SynFit improves protein fitness prediction by learning shared and property-specific sequence representations, outperforming existing methods in multi-objective optimization.

ARACoFusion predicts protein-protein interactions in Arabidopsis thaliana using a deep learning model that improves accuracy and reliability through uncertainty-aware techniques and benchmarking against existing predictors.

The paper develops an orthogonal initiation system for efficient N-terminal noncanonical amino acid incorporation in cell-free translation without disrupting the native methionine pathway.

This paper predicts influenza virus host tropism and zoonotic spillover risk using protein sequence embeddings, achieving high precision and recall scores across multiple host classifications.

CLIPepPI predicts domain-peptide specificity using contrastive learning, achieving competitive performance across benchmarks and enabling proteome-wide analysis.

The paper develops a stickiness scale for disordered proteins based on biophysical experiments and compares it to existing hydropathy scales.

The paper uses neural networks to predict protein evolutionary rates by integrating molecular and cellular features across multiple scales, showing that protein-level and site-level properties contribute additively to evolutionary rate variance.

This study identifies nanobody CDRs using deep learning and structural analysis, finding that their disorder and conformational diversity are adaptive features.

This paper uses an ensemble of protein language models to identify beneficial mutations in Aspergillus oryzae β-galactosidase, improving catalytic efficiency and thermostability without supervised training.

The paper introduces BioRAG-DRAG, a local-first multimodal retrieval system for biomedical agents, combining neural retrieval, BLAST verification, and graph-based evidence packaging.

The paper shows that high accuracy in predicting enzyme kcat fold-changes from sequence data often results from data biases, not true mechanistic understanding, and introduces a bias-aware framework to improve predictions.

ProMoNet improves drug discovery by predicting protein-ligand binding affinity and kinetics using sequence-based pre-training without 3D structural data.

CASPE is a lightweight platform for site-directed protein engineering that improves the identification of beneficial mutations for thermostability and pH tolerance.

TREAD improves repeat annotation by using protein embeddings instead of sequence profiles, outperforming HMMER in low-data and high-divergence scenarios.

This paper introduces a platform for efficient protein delivery using extracellular vesicles engineered with biogenesis-inspired features, achieving functional gene editing in vitro and in vivo.

ProteomeLM predicts protein-protein interactions and gene essentiality across species using a proteome-scale language model, outperforming existing methods in speed and accuracy.

The paper introduces a new chemical method, SAPLabTech, to create self-assembling artificial proteins with functional diversity by using supramolecular complexes for site-specific conjugation.

H3BERTa is a transformer model trained on CDR-H3 sequences to analyze antibody repertoires, capturing immunological signatures and supporting classifier development for broadly neutralizing antibodies.

ProtmRNA transfers knowledge from protein language models to mRNA sequences, achieving competitive performance with less computational resources.

This study identifies a second mobile region in mHORMADs that controls their conformation and function, showing its conservation across species.

The paper maps the functional landscape of the human intrinsically disordered proteome using evolutionary signatures to classify and predict IDR functions.

The paper introduces two models, PiCAP and CAPSIF2, to predict protein-carbohydrate interactions at the protein and residue levels, achieving high accuracy and demonstrating substantial interplay across proteomes.

This paper develops a hybrid feature approach combining deep learning and amino acid frequency features to improve protein-protein interaction prediction using a random forest classifier.

DiscrimAlign learns substitution matrices and gap penalties for sequence alignment to predict microRNA-target interactions, achieving performance comparable to neural networks while being more interpretable.

The paper identifies and classifies ion channels across species, revealing functional insights into understudied CALHM channels through sequence, structure, and functional data integration.