AWS Launches Bio Discovery

Drug discovery often starts as a search problem: researchers sift through huge numbers of possible molecules to find a few worth testing in a lab. On April 14, Amazon Web Services launched Amazon Bio Discovery, a new application that packages that search into a cloud service for drug researchers. (aws.amazon.com) Amazon said the service gives scientists access to more than 40 biological foundation models, or artificial intelligence systems trained on large biology datasets, plus tools to upload their own models and licensed third-party ones. The company said the software can guide model selection, configure workflows, and send promising candidates to contract research partners for lab validation. (aws.amazon.com) AWS described the product as a no-code system for running complex early-stage drug-discovery workflows, rather than a single model for one task. Reuters reported the launch as part of Amazon’s push to speed early-stage drug research by letting scientists design and test novel drug candidates more quickly. (usnews.com) The company’s example is antibody discovery, where researchers try to design proteins that bind to a disease target the way a key fits a lock. Amazon said Memorial Sloan Kettering Cancer Center used the system to design nearly 300,000 novel antibody candidates, narrow them to 100,000, and move them into wet-lab testing in weeks instead of up to a year with older methods. (aws.amazon.com) That workflow is what biotech companies call “lab in the loop”: software proposes candidates, a physical lab tests them, and the results feed back into the next round. AWS said Amazon Bio Discovery ties those steps together in one application so teams do not have to shuttle models, data, and results across separate systems. (aws.amazon.com) Amazon launched the service with named customers and partners including Bayer, the Broad Institute, Memorial Sloan Kettering, Revolution Medicines, Apheris, and NVIDIA. Bio-IT World reported that AWS is also emphasizing data isolation and customer ownership of proprietary data and intellectual property for regulated pharma and biotech users. (aboutamazon.com) (bio-itworld.com) The release extends a broader AWS strategy in life sciences: sell not just raw computing and storage, but packaged industry software that sits closer to research work. AWS has long marketed cloud infrastructure for genomics, imaging, and high-performance computing in pharmaceutical research and development. (aws.amazon.com) Amazon is entering a field crowded with specialized biotech software groups and model makers, but it is betting that cloud scale and existing enterprise relationships will matter. AWS said commercial and open-source biology models from partners including Apheris and Boltz are available now, with Biohub and Profluent listed as coming soon. (aws.amazon.com) (bio-itworld.com) The pitch is straightforward: keep the model catalog, orchestration layer, and lab handoff in one place, then shorten the cycle between idea and experiment. Amazon’s public case study does not claim an approved drug, but it does claim that one of the slowest parts of antibody research can now run on a cloud timeline measured in weeks. (aws.amazon.com)