protein design

The paper identifies novel leucine-associated reduction in protein abundance linked to 2A-like peptides from eukaryotes using a high-throughput fluorescent reporter tool.

This paper introduces a generative AI framework for designing transmembrane β-barrel nanopores with customizable properties for ion sensing, DNA translocation, and transport.

MDMI is a structure-guided pipeline that designs functional peptide variants from minimal data, achieving high accuracy and discovering divergent sequences with measurable activity.

This paper introduces a physics-based generative method for designing ligand-binding proteins and demonstrates its application in creating biosensors for small molecules and metal ions.

This paper engineers the bacterial cellulose-forming surface to recruit proteins, creating functionalized biocomposites with improved material properties.

A prediction-guided FGF21 construct with a longer C-terminus gave less protein than the literature design yet concentrated to 1.4 mg/mL versus 0.7 mg/mL (n = 1).

The paper designs structurally diverse antibody repertoires with drug-like properties using a computational method called CADAbRe, demonstrating their utility in binding to challenging targets.

This paper introduces azatryptophans to engineer enzymes with enhanced activity and preserved stability, demonstrating their effectiveness in PETase optimization.

This study shows that larger protein language models can overestimate sequence likelihood, leading to poor fitness prediction, and recommends moderate model sizes for better performance.

ProtBLIP2-SST generates open-ended protein function captions by integrating sequence, structure, and text using a two-stage framework.

The paper describes a computational workflow for designing NAZ-mAb variants targeting p95HER2, showing improved expression yields but weaker binding affinities compared to the parent antibody.

This paper evaluates pLM-guided data augmentation for protein sequence prediction, showing that it improves performance across tasks with variation levels between 2-30%.

The paper designs protein sequences with varying single-site biases and pairwise covariances, finding that single-site biases determine stability while pairwise covariances affect catalytic activity.

This paper shows that substituting ribonucleotides with deoxyribonucleotides in mRNA improves translation efficiency and therapeutic potential in vivo.

MAERM predicts enzyme-reaction matching relationships using a mixed-attention model, outperforming baselines and enabling enzyme design.

InstaNovo-P improves phosphorylated peptide detection and site localization in phosphoproteomics by using a transformer-based model fine-tuned on phosphoproteomics data.

IgGM2 is an all-atom foundation model for immune receptor design that jointly predicts CDR sequences and full-atom structures, outperforming existing methods in structure prediction and sequence design benchmarks.

The paper introduces a hybrid method using a protein language model and Monte Carlo sampling to generate enzyme mutants that preserve structural features while varying overall sequence, providing sequences for experimental testing.

This paper designs and tests hyper-thermostable CO2 reductases with cobalt porphyrin cofactors, achieving high activity and suppressing competing reactions.

The paper describes a cost-effective E. coli-based method for producing endotoxin-free cytokines, demonstrating bioactivity comparable to commercial products at 5%-10% of the cost.

This paper designs blue melanin using peptide motifs and computational methods, achieving stable blue color with specific absorption wavelengths and thermal/photostability.

The paper identifies RPAP3 and PIH1D1 as R2TP subunits and shows their role in assembling INO80, SRCAP, and TIP60 complexes through physical interactions and functional validation.

PINPOINT predicts protease-inhibitory activities of fungal SSPs lacking annotated inhibitor domains using protein language models and structural modeling.

The paper identifies naturally occurring ACE2 stalk variants that differentially affect soluble ACE2 release and surface expression of B0AT1.

This paper reprograms a protein ligase into a functional aminoacyl-tRNA synthetase, enabling the incorporation of non-canonical amino acids and post-translational modifications into proteins.

The paper designs a multi-epitope vaccine against Influenza A H10N5 and H3N2 using immunoinformatics and validates it computationally, showing activation of immune responses.

The paper uses machine learning to guide cell-free expression of carbonic anhydrases, identifying variants with improved stability under high temperatures.

The paper designs peptides with custom secondary structures using a generative AI model trained on protein data, achieving success in creating novel proteins with desired motifs and reduced amino acid alphabets.

This study designs high-fidelity CRISPR-Cas variants using a physics-driven approach combining molecular dynamics and protein language models, achieving significant improvements in fidelity with minimal experimental validation.

The paper evaluates zero-shot models' ability to predict epistatic effects on protein properties and finds they fail to predict strongly epistatic mutation combinations.