protein sequence

This paper introduces a deep-learning model, PALM, that predicts peptide aggregation using protein language model embeddings, showing improved performance with larger datasets.

This study evaluates using ProtT5 embeddings with gradient boosting for Gene Ontology prediction, finding gradient boosting provides the best balance between performance and stability.

The paper shows that ESM-2 attention maps encode the directional hierarchy of allosteric communication in the adenosine A2A receptor through topological asymmetry.

MuSIC predicts cross-species RBP-RNA interactions using evolutionary conservation and label smoothing, outperforming existing methods and quantifying variant effects on binding.

The paper introduces a context-aware framework integrating text and protein language models to predict multiple antibody properties without retraining, achieving high correlation across developability properties.

The paper benchmarks a new substitution matrix derived from AlphaFold-predicted structures against existing matrices, finding similar performance and highlighting the diminishing role of substitution matrices with more sequence data.

PROBind predicts, analyzes, and visualizes protein-protein and protein-nucleic acid binding residues using sequence and structure data, integrating multiple predictors and providing interactive tools.

The paper shows that MULTI-evolve's multimutant predictions correlate closely with additive models, suggesting its performance is not due to epistasis learning.

This paper introduces a two-layer framework using protein language models to identify transcription factors and predict TFs with methylated DNA binding preferences, achieving better performance than existing methods.

PISCO improves codon optimization by integrating protein structure and host-specific codon usage, enhancing solubility and expression.

This study investigates the evolutionary diversification of plant actins, finding that actin copy number increases correlate with plant complexity and that functional distinctions are maintained through specific sequence changes.

Aiki-XP predicts within-species relative protein expression across bacteria using multimodal data, outperforming single-omics models in leakage-controlled settings.

The paper introduces yakRNA Design, a language model that generates functional synthetic RNAs without 3D structural data, achieving efficient frame-shifting capabilities in ribosomes.

This paper introduces a framework using protein language models to discover novel enzymes with specific functions by clustering sequence embeddings and recursively exploring functional reservoirs.

This paper uses a phylogenetic approach to analyze bradyzoite differentiation in Toxoplasma gondii and identifies conserved and novel proteins involved in this process.

PEACE improves effector protein prediction by using contrastive embeddings and prototype-aware training, outperforming EffectorP 3.0 on fungal and fungi+oomycete datasets.

This paper introduces a bi-level diversity-optimization framework for selecting representative protein panels, reducing redundancy and improving diversity compared to existing methods.

This study benchmarks agentic large language models for protein-set functional annotation, finding moderate reproducibility and common error modes, but improved evidence handling and reproducibility.

The paper uses FAMPNN to redesign SrBDH1 for increased thermostability while maintaining stereoselectivity.

This study compares how stabilizing versus overproducing target proteins affects PROTAC activity using β-catenin as a model.

This paper shows how ionic strength affects the structural disorder and oligomerization of the Phd transcription factor, revealing its ability to function as a conformational rheostat.

Hybrid Gated Fusion improves protein function annotation by integrating sequence, structure, text, and interaction data using bilinear gating and auxiliary supervision, achieving state-of-the-art results on CAFA3.

The paper presents a computational workflow for designing nanobody binders against a novel cancer target, achieving sub-nanomolar affinities without experimental structure or prior antibody data.

This paper introduces iClust, an interpretable clustering method for biological sequences that provides clear cluster explanations and competitive performance on simulated and real datasets.

This paper presents a generative framework for designing intrinsically disordered proteins using protein language models, showing that large datasets are essential for accurate conformational control.

The paper expands the classification of reflectin domains in cephalopods to accommodate sequence diversity and introduces a framework for identifying reflectins across species.

The paper identifies context-dependent sequence features that improve prediction of de novo binder success beyond structure and affinity.

The paper develops a toolkit of membrane-localising tags tested across ten animal species to identify effective signals for targeting proteins to the plasma membrane.

The paper benchmarks MSA pairing strategies for protein-protein complex structure prediction and finds that increased MSA depth, not specific pairing, improves accuracy.

BABAPPAlign improves multiple sequence alignment by using a learned residue-level scoring function, outperforming existing tools on standard benchmarks.