Latest news with #RFdiffusion
Geek Wire
24-07-2025
- Health
- Geek Wire
UW researchers discover AI-powered breakthrough that could boost precision cancer treatment
The co-lead authors of a study publishing today in Science describing AI-designed proteins for target health theraphies, from left: Julia Bonzanini, Nathan Greenwood and Bingxu Liu. All are researchers with UW Medicine's Institute for Protein Design. (IPD Photo) A breakthrough in the science used to customize the treatment of cancer and other diseases is so promising that researchers at the University of Washington are planning to launch a company to commercialize the technology. The strategy uses artificial intelligence to create proteins that recognize and bind to specific markers on diseased cells, creating what are essentially biological neon signs that attract immune cells to destroy the targets. The research comes from the lab of Nobel laureate David Baker and UW Medicine's Institute for Protein Design, which Baker leads, and is being published today in the journal Science. Additional study authors come from multiple UW departments, the Garcia Lab at Stanford University, and the Scheinberg Lab at Memorial Sloan Kettering Cancer Center. 'Detecting unhealthy cells is one of the main jobs of the immune system, but it doesn't always notice the subtle signs of cancer or viral infection,' Baker said in a statement. 'In this study, we show that computer-designed proteins can help human immune cells spot the right targets and function more effectively.' The work could lead to significant advancements in precision immunotherapy, which delivers disease-fighting drugs that are crafted for individual patients. The key to this personalized approach lies in understanding how cells display their identity on their surfaces. The outside membrane of a cell is studded with molecules that include small proteins called peptides that reveal a cell's inner workings, including if it's cancerous or infected by a virus such as HIV. The scientists used RFdiffusion and ProteinMPNN — AI-tools built by the Institute for Protein Design — to efficiently and cheaply engineer proteins that recognize the unique peptides. University of Washington biochemist and Nobel Prize laureate David Baker at his office in Seattle. (GeekWire Photo / Lisa Stiffler) The custom-made proteins can then be integrated into chimeric antigen receptors (CARs), which are engineered molecules that attract and activate immune-system warriors called T cells to attack specific targets. In their study, the researchers designed proteins for 11 peptide targets and eight successfully triggered a T-cell response. Of those eight, two produced such a strong immune response that the T cells killed the targeted cells. The peptide targets included HIV fragments and tumor-related protein mutations. 'We've shown how advances in protein design could make personalized cancer therapy possible, and we intend to start a company to turn these results into real therapies that benefit patients,' said Bingxu Liu, co-lead author of the study and postdoctoral scholar in the Baker Lab. Since 2014, the Institute for Protein Design has spun off 10 startups, and Baker has co-founded 21 tech companies. When Baker won the Nobel Prize in October, the Royal Swedish Academy of Sciences noted that 'his research group has produced one imaginative protein creation after another, including proteins that can be used as pharmaceuticals, vaccines, nanomaterials and tiny sensors.' The researchers involved in the latest study explained that the strategy can be easily adapted to new scenarios. Within days, the team was able to take one of its successful proteins and use AI to modify it to target new cancer and virus-associated peptides. The process could make it much cheaper to create personalized treatments. 'I'm hopeful that this will lead to new therapies that are more accessible to patients around the world who do not benefit from current state-of-the-art cancer treatments,' said Julia Bonzanini, a co-lead author and Baker Lab graduate student. Researchers worldwide can access online the open-source software used in the research. Other authors of the Science paper, titled 'Design of high-specificity binders for peptide–MHC-I complexes,' are Nathan Greenwood, Amir Motmaen, Jeremy Meyerberg, Tao Dao, Xinyu Xiang, Russell Ault, Jazmin Sharp, Chunyu Wang, Gian Marco Visani, Dionne Vafeados, Nicole Roullier, Armita Nourmohammad, David Scheinberg and Christopher Garcia. RELATED:
Business Wire
11-06-2025
- Science
- Business Wire
Diffuse Bio Launches DSG2-mini AI Model for Protein Binder Design
SAN CARLOS, Calif.--(BUSINESS WIRE)-- Diffuse Bio, a leading provider of AI systems for designing protein therapeutics, today announced the public launch of DSG2-mini, its latest AI model, now accessible to researchers worldwide through DiffuseSandbox, a new web-based platform for designing proteins that can bind to target proteins of interest. DSG2-mini is a streamlined version of Diffuse Bio's upcoming flagship foundation model DSG2. The model is now available through DiffuseSandbox, marking the first time researchers can directly access Diffuse Bio's cutting-edge protein design capabilities. Powerful Capabilities for Protein Engineering DSG2-mini enables researchers to design custom protein binders with unprecedented precision, and DiffuseSandbox makes these capabilities accessible through an intuitive web interface: Proven performance: DSG2-mini outperforms leading models including DSG1, RFdiffusion, and RFantibody in computational benchmarks for nanobody design No coding required: Upload target structures and generate designs through point-and-click interface Real-time design generation: Get novel protein binder candidates in minutes, not days Flexible input control: Design binders against any target protein using simple specifications Currently supports nanobody design, with upcoming support for scFvs and minibinders Downloadable results: Export designed structures for laboratory testing and validation "DiffuseSandbox puts the power of DSG2-mini into researchers' hands, whether they're at major institutions or working independently," said Namrata Anand, CEO at Diffuse Bio. "This launch advances our mission to enable researchers to engineer proteins with unprecedented control and accuracy." DiffuseSandbox launches with free starting credits for all researchers, enabling immediate experimentation with protein binder design. About Diffuse Bio Diffuse Bio is pioneering the future of protein engineering through AI models and large-scale biological data collection. The company focuses on enabling rapid, accurate protein design for applications spanning therapeutics, diagnostics, and novel enzyme development. For more information about DiffuseSandbox and to start designing protein binders, sign up here and follow Diffuse Bio on LinkedIn and X or visit
