protein sequence

DELPHI is an open platform for predicting antibody developability from sequence, achieving high AUC scores and enabling residue-level engineering insights.

The paper shows that human signal peptides are inefficiently cleaved by yeast SPCs due to differences in SPC-lipid interactions and membrane thinning profiles.

NoroScope uses context-aware machine learning to prioritize amino-acid substitutions in the norovirus capsid protein, achieving high precision in predicting mutation effects with limited data.

This paper introduces Leff, a spectral measure to quantify effective information in multiple sequence alignments, showing evolutionary constraints reduce sequence diversity in RNA and proteins.

The paper evaluates multi-view gene representations under a matched contract, showing condition-dependent fusion operator rankings across different view combinations and tasks.

The paper builds an InterVir Atlas to map temporal interactome changes across RNA viruses and identifies convergent host vulnerabilities for antiviral drug repurposing.

The paper identifies artifacts in bacterial proteomes by analyzing cluster participation and protein length, finding that short proteins are often errors due to frame-shifts or termination codons.

The paper identifies that peptide hydrophobicity affects their availability in yeast cells, impacting synthetic interaction strength beyond in vitro measurements.

This paper introduces RiPLE, a method for discovering engineered protein binders, and shows it produces high-affinity, selective binders for ROR1 and GS.

The paper identifies gene-specific evolutionary-rate associations with quantitative mammalian brain phenotypes using integrated phenotypic and genomic data.

StackHPpred identifies peptide hormones from amino acid sequences using a stacking-based ensemble framework, achieving high accuracy on independent datasets.

ANIMA predicts protein-protein interactions across species using an AI model trained on diverse animal species, achieving high accuracy in under-researched species.

This study shows ACAN promotes aortic dissection by disrupting mitochondrial homeostasis in VSMCs via the YAP1/TAZ pathway, and its reduction improves disease outcomes.

This study tests using a variational autoencoder to reconstruct ancestral sequences by interpolating in its latent space, finding it less accurate than standard methods despite containing evolutionary signal.

VES predicts viral mutation fitness and evolutionary trajectories using a diffusion model trained on protein sequences, outperforming existing models in mutation fitting and forecasting future mutations.

Pep-PU-GAN improves peptide function prediction using PU learning and GANs with graph neural networks, achieving high F1 and AUROC scores on neuropeptide classification.

Reward-guided inference-time search over the Proteína-Complexa sequence-structure codesign model yields validated binders across 127 targets, kinases and a free carbohydrate, with codesigned sequences outperforming ProteinMPNN redesign.

This paper introduces Protea-Tac, a Ub-independent protein degrader that targets proteins to the proteasome, demonstrating in vivo tumor suppression.

The paper introduces a generative model framework that maps protein sequences to microscopy images and vice versa, enabling virtual mutagenesis and functional motif analysis.

This study identifies uterocalin as a specialized protein supporting embryo nutrition and capsule formation in equids, rhinos, and tapirs, showing unique amino acid and structural features across perissodactyls.

This paper fine-tunes a protein language model on nanobody thermostability data and uses sparse autoencoders to uncover biophysical rules governing stability, identifying both known and novel stabilizing features.

5,776,417 ColabFold-AlphaFold2 structures cover 98.9% of viral UniProt 2025_03 sequences, 16.4-fold more than earlier versions, with 75.3% high-confidence and 72.6% of viral reference proteomes fully covered.

DepoCat is a database of experimentally verified phage depolymerases that enables sequence and structure-based classification and specificity prediction.

The paper models protein sequence evolution using an Ornstein-Uhlenbeck process in a latent space to improve contact prediction by integrating directed evolution data with natural homologs.

RealKcat predicts enzyme kinetics from sequence using a curated dataset and embeddings, showing sensitivity to catalytic mutations and robustness across evaluations.

This study evaluates mRNA vaccines encoding PfMSP2, showing they induce strong IgG responses and functional antibodies against malaria merozoites, comparable to or better than a protein-based vaccine.

Cyanochelin C, a non-acylated peptide siderophore from Atacama Desert Myxacorys chilensis, was fully stereochemically resolved; its gene cluster carries CcsQ, a putative deacylase from a cyanobacteria-specific, siderophore-linked acylase lineage.

This paper compares GANs and VAEs for generating viable antibody sequences, finding GANs outperform VAEs when sampling strategies are optimized.

This study shows that mTOR inhibition restores RNA-regulatory proteins in AD models, suggesting a reversible mTOR-dependent RNA regulatory program in cerebrovascular dysfunction.

The paper develops a foundation model trained on fully human heavy-chain-only antibodies, showing it captures sequence and functional constraints distinct from conventional antibody models.