Amazon’s drug‑discovery tool

Drug discovery still starts with a long guessing game: researchers test huge numbers of molecules to find a few that might work in the body. On April 14, Amazon Web Services said its new Amazon Bio Discovery tool is meant to shrink that early search by letting scientists run those screening workflows without writing code. (reuters.com) Amazon Web Services launched the product on Tuesday, April 14, and described it as a no-code application for early-stage drug research. The service gives scientists access to specialized biological foundation models, which are artificial intelligence systems trained on large biology datasets to generate and rank possible drug candidates. (reuters.com, aws.amazon.com) The company said researchers can use an artificial intelligence agent to choose models, set experiment parameters, and interpret results, then send shortlisted candidates to lab partners for synthesis and testing. Test results are routed back into the software so the next round of designs can be updated with real lab data. (reuters.com, aws.amazon.com) That setup targets a practical problem in pharmaceutical research: the people who can translate a biologist’s goal into machine-learning pipelines are scarce, and handoffs between software teams and wet labs slow projects down. Amazon Web Services said the application is designed to let bench scientists and computational biologists work from the same system instead of passing files across separate tools. (reuters.com, aws.amazon.com) Amazon Web Services is also moving beyond selling raw cloud capacity to drugmakers and into packaged research software built for a specific industry. Reuters reported that 19 of the top 20 global pharmaceutical companies already use Amazon Web Services cloud services, giving Amazon a large installed base to market the new product to. (reuters.com) Amazon Web Services said the application includes more than 40 biology models, plus tools for customers to upload their own models or licensed third-party models. The company said computational biologists can turn those multi-step workflows into reusable templates for other teams inside the same organization. (aws.amazon.com) Rajiv Chopra, vice president of healthcare artificial intelligence and life sciences at Amazon Web Services, told Reuters that a process that once took 18 months to produce 300 potential drug candidates can now produce 300 candidates in a couple of weeks. Amazon Web Services also said that, in work with Memorial Sloan Kettering Cancer Center, the platform generated nearly 300,000 novel antibody molecules and narrowed them to 100,000 candidates for lab testing by Twist Bioscience. (reuters.com, aboutamazon.com) Early users named by Amazon Web Services include Bayer, the Broad Institute, Memorial Sloan Kettering Cancer Center, and Voyager Therapeutics. The company said it will start with a free trial that includes five experimental units before moving to subscription pricing. (reuters.com, aws.amazon.com) Amazon Web Services said the tool is meant to help scientists decide which molecules deserve real lab time, not replace the labs themselves. The pitch is simple: cut the number of manual steps at the front of drug discovery, and move more quickly from a computer-generated idea to a tested candidate. (reuters.com, aws.amazon.com)