protein design

The paper introduces SPURS, a deep learning framework for protein stability prediction that integrates generative models and achieves accurate, scalable, and generalizable results.

The paper develops a cross-ancestry information transfer framework to improve protein abundance prediction and protein-trait association identification in underrepresented populations.

The paper introduces a framework for engineering tunable oscillatory protein expression using Tet-On and dTAG degron systems, demonstrating synchronous oscillations with the COD system.

The paper maps the photodynamic diversity of LOV domains, revealing kinetic diversity and functional classes, and demonstrates computational design of new LOV photosensors.

The paper designs RNA switches using Restricted Boltzmann machines and validates their structural switching properties through experimental testing.

This study shows that electrostatic interactions in the Nuc domain of Cas12a orthologues regulate trans cleavage activity, with specific mutations enhancing or abolishing cleavage efficiency.

This study designs a protein that precisely controls Gremlin-1 antagonism by releasing BMPs, revealing the role of interface rigidity in antagonist function.

This study identifies RNA-binding activity of the SACSIN HEPN domain as linked to ARSACS, providing insights into its molecular function and potential therapeutic targets.

The paper designs protein competitors for small molecule detection using deep learning, demonstrating effective binders for digoxin detection.

This paper computationally reconstructs T. rex collagen using paleo-proteomics and protein models, enabling tunable biomaterial stiffness through inducible crosslinking.

This paper introduces a streamlined platform for luciferase engineering using continuous hypermutation, resulting in improved light output variants without manual library generation.

This paper introduces PS and IIOS, AI tools for analyzing and engineering molecular interactions by decomposing binding interfaces with stage-resolved force-wise insights.

The paper investigates the role of cyclic nucleotide binding sites in the Drosophila protein SWS, showing that mutations in these sites affect its catalytic function and phospholipase activity.

This paper introduces CUT&ID, a method that simultaneously profiles protein-DNA and protein-protein interactions without requiring transgenesis.

The paper presents evidence that eukaryotic circular RNAs can be translated in the 3' to 5' direction, producing novel proteins with distinct features, and identifies millions of potential BT proteins across species.

The paper engineers a covalent linkage into a de novo enzyme, revealing that the modified protein rescues cells through upregulating siderophore biosynthesis rather than catalyzing iron release.

The paper uses sequence-to-function machine learning models to predict and design improved red fluorescent protein variants, identifying variants with enhanced photophysical properties.

The paper designs and engineers minimal protein-based molecular inverters and switches that respond to ligands, demonstrating tunable and modular control of post-translational signaling.

This paper presents a computational pipeline for designing VHH nanobodies against LGR5 without structural templates, resulting in high-affinity, stable binders validated by cryo-EM.

This paper designs small single domain proteins that present three distinct viral epitopes using deep learning, showing improved immunogenicity compared to single-epitope designs.

This paper designs novel solenoid proteins using in silico evolution, achieving successful folding for alpha-solenoids and improved results for beta-solenoids with refinements.

This paper introduces Riff-Diff, a machine learning and atomistic modeling method for designing enzymes with high catalytic activity and stereoselectivity, demonstrating its effectiveness for two chemical reactions.

RFdiffusion2 scaffolds unindexed atom-level catalytic active sites via flow matching without inverse rotamer enumeration, solving all 41 in silico benchmark cases and yielding active de novo enzymes across four reactions.

BAGEL enables flexible protein design by exploring an energy landscape through a modular framework, supporting diverse design goals and integrating deep learning models.

This paper designs highly active metallohydrolases using a new generative AI method, achieving kcat/KM values up to 53,000 M-1 s-1 without experimental optimization.

The paper designs a multi-epitope immunogen for malaria using machine learning, demonstrating its ability to bind and orient three copies of the L9 antibody.

This paper introduces a discrete diffusion model for protein sequence design using all-atom representations, showing improved novelty and diversity compared to amino acid-based models.

This paper identifies a novel type III-E-like CRISPR-Cas effector composed of Cas7-like and Cas1-like domains, demonstrating its RNA-targeting capability and modularity for engineering synthetic Cas effectors.

This paper introduces LAffAb, a method to co-optimize antibody affinity and stability through combinatorial mutation design, achieving significant affinity gains while maintaining developability.

This study shows pyridostatin stabilizes telomeric G-quadruplexes, inhibiting lagging strand replication, which is overcome by FEN1/DNA2 cleavage and MMEJ repair, suggesting potential for anti-cancer therapies.