Tag

Protein Design

All articles tagged with #protein design

technology6 days ago•54 min saved

Google's SynthID Bio Embeds Invisible Watermarks in AI-Designed Proteins to Boost Biosecurity

Google DeepMind has introduced SynthID Bio, a new technology that embeds imperceptible watermarks into AI-generated protein sequences and 3D structures. Published in Nature on September 30, 2026, the system aims to verify the provenance of synthetic biological designs without compromising their biological function. By integrating with tools like ProteinMPNN and AlphaFold 3, SynthID Bio allows DNA synthesis providers to distinguish trusted AI designs from potentially hazardous unknowns, enhancing biosecurity and scientific integrity.

via Nature|

#ai-watermarking #biosecurity #google-deepmind

technology-and-science9 days ago•6 min saved

Google DeepMind Introduces SynthID Bio to Watermark AI-Designed Proteins for Biosecurity

Google DeepMind has released SynthID Bio, a new system that embeds imperceptible watermarks into AI-generated protein sequences and 3D structures. Published in Nature on September 30, 2026, the technology aims to enhance biosecurity by allowing DNA synthesis providers to verify the provenance of novel biological designs. By integrating with popular tools like ProteinMPNN and AlphaFold 3, the system ensures that watermarked proteins retain their biological function, enabling trusted researchers to distinguish their work from potentially hazardous unverified sequences.

via Ars Technica|

#ai-safety #biosecurity #deepmind

biotech2 months ago•32 min saved

ProteinGuide enables on-the-fly conditioning of protein design models with experimental data

ProteinGuide provides an inference-time conditioning framework that lets a broad class of protein sequence generators (including masked language models like ESM3, autoregressive models like ProteinMPNN, and discrete-space diffusion/flow models) be guided by auxiliary experimental data without retraining. Built on a unifying statistical approach, it enables designs targeting properties such as stability and activity, supports multi-property Pareto optimization, and, when combined with wet-lab data, can enhance in vivo base-editor activity beyond multi-round directed evolution. The work provides code and data (Zenodo, GitHub) and a ProteinGen package to facilitate design workflows.

via Nature|

#bioengineering #biotech #experimental-data

technology2 months ago•42 min saved

Raygun: AI-driven, length-agnostic protein design that preserves function

Raygun is an autoencoder-based AI framework that encodes proteins as fixed-dimensional Gaussian distributions derived from protein language model embeddings, enabling rapid, single-shot generation of variants at any length with controlled substitutions and indels while preserving predicted structure and functional sites. It demonstrates miniaturization of fluorescent proteins and TurboID, enlargement of EGF with improved EGFR binding, and broad applicability to template-guided protein redesign, offering a faster alternative to diffusion-based de novo design.

via Nature|

#bioengineering #language-models #protein-design

science2 months ago•85 min saved

AI-designed starting points accelerate enzyme evolution for enhanced stability and precision

A Nature study shows that AI-designed, stabilized BoNT proteases (via ProteinMPNN) used as starting points for phage-assisted continuous evolution (PACE/PANCE) consistently yield enzymes with higher stability, better soluble expression, and greater catalytic efficiency across substrates than wild-type starts. Redesigned starting points expand the mutationally accessible sequence space, enabling faster adaptation and access to high-function genotypes that WT enzymes cannot reach. Notably, redesigned BoNT/E variants evolved to specifically cleave disease-relevant ataxin-2 with substantially higher specificity and minimal native-substrate activity, illustrating a practical AI-plus-evolution workflow to rapidly reprogram enzymes for new targets with improved safety profiles.

via Nature|

#ai #enzyme-evolution #protein-design

science2 months ago•4 min saved

Doudna’s AI-designed tiny nucleases promise on-demand gene editing

Nobel laureate Jennifer Doudna and colleagues report an AI-driven platform for designing non-natural nucleases, producing a tiny TnpB enzyme capable of binding and cutting DNA in human, plant, and bacterial cells. By using an inverse protein-model to tailor sequences to predefined backbones, the team envisions on-demand enzyme design with potential medical and agricultural applications and notes possible IP implications as the field advances.

via Fierce Biotech|

#artificial-intelligence #gene-editing #jennifer-doudna

science4 months ago•25 min saved

Programmable quasisymmetric protein cages from two complementary building blocks

Nature reports a computational design strategy using geometric frustration to create two-component, quasisymmetric protein cages that assemble into sphere-like structures by embedding curvature-inducing pentagonal defects. By pairing complementary trimeric and dimeric blocks, the authors programmably control cage size from ~40 nm to >200 nm and mass from 2 to >50 MDa, comparable to viral capsids. The cages are functionalized for ribonucleoprotein cargo loading and cellular uptake, enabling studies of cargo delivery and size-dependent diffusion in cells. Data and code are publicly available (Zenodo, GitHub), underscoring a new route for biologics delivery and cell biology tools.

via Nature|

#cryo-em #de-novo-design #protein-cages

science4 months ago•5 min saved

Bacteria Survive on 19 Amino Acids in Ribosomes for 450 Generations

Columbia University researchers redesigned 21 ribosomal proteins in E. coli to remove isoleucine, using AI-guided protein design, and created a viable strain that survived and reproduced for over 450 generations. The genome still largely relies on isoleucine, so it's not a full 19-amino-acid organism, but the work shows life can function with a reduced amino acid alphabet and provides a framework for studying early protein synthesis.

via Indian Defence Review|

#amino-acids #biology #genetics

science5 months ago•4 min saved

AI-Designed NovoTags Expand Multicolor Live-Cell Imaging

Researchers at HHMI’s AI@HHMI, led by David Baker and Luke Lavis, are using RFdiffusion-based AI to design NovoTags—small protein binders that pair with Janelia Fluor dyes—to create a new class of fluorescent probes. This approach could let scientists label many colors and proteins simultaneously, without traditional chemical linkers, enabling longer, multi-color imaging and accelerating discovery. The team plans to rollout NovoTags for about a dozen dye colors and expand to dyes that blink or respond to physiological signals, with the tools becoming broadly available to the scientific community.

via HHMI|

#ai #fluorescent-imaging #novotag

technology5 months ago•5 min saved

No‑Code AI Turns Biologists into Protein Designers

OpenProtein.AI offers a no‑code platform that lets biologists upload data and use AI-powered tools (including PoET and PoET‑2) to design, predict, and optimize proteins, with access for academia and collaborations with industry partners like Boehringer Ingelheim, advancing a shared language for describing protein behavior and accelerating therapeutics.

via MIT News|

#no-code-platform #openproteinai #poet

science9 months ago•17 min saved

Synthetic Heteropolymers Mimicking Enzymes

The article discusses the design and analysis of random heteropolymers as enzyme mimics, including data and code availability, experimental methods, and computational studies to understand their structure and function, supported by extensive references and acknowledgments.

via Nature|

#enzyme-mimics #heteropolymers #molecular-dynamics

science11 months ago•4 min saved

AI-Designed Antibodies Achieve Breakthroughs in Drug Development and Protein Design

Researchers from Nobel Laureate David Baker's lab and the University of Washington have used AI to design antibodies from scratch, successfully creating molecules that bind to their targets, potentially revolutionizing the antibody drug industry and enabling faster, more efficient drug development.

via GeekWire|

#ai-designed-antibodies #antibody-engineering #biotechnology

biology1 year ago•98 min saved

AI-Driven Innovations in Cytokine Signaling and Drug Control

The article discusses the design of protein systems that utilize facilitated dissociation to control the timing of cytokine signaling, enabling rapid and tunable regulation of protein interactions through allosteric mechanisms and structural strain, with applications in cellular signaling and biosensing.

via Nature|

#biology #cytokine-signaling #facilitated-dissociation

science1 year ago•75 min saved

BindCraft Enables One-Shot Design of Functional Protein Binders

BindCraft is a novel computational pipeline that leverages deep learning, specifically AlphaFold2, to de novo design functional protein binders targeting diverse proteins, including challenging membrane receptors, allergens, and nucleases, with high success rates and potential therapeutic applications.

via Nature|

#bindcraft #deep-learning #protein-design

science1 year ago•5 min saved

BindCraft AI Excels in One-Shot Protein Design

BindCraft, an AI-powered pipeline for de novo protein binder design developed by EPFL researchers, achieves high success rates in creating functional protein binders against diverse targets, including challenging proteins like CRISPR-Cas9, with potential to accelerate drug discovery and therapeutic development. Its open-source availability has garnered widespread industry and academic adoption, marking a significant advancement in computational protein engineering.

via Genetic Engineering and Biotechnology News|

#ai #bindcraft #drug-development