Nvidia Corp. is making significant advancements in the field of artificial intelligence (AI) for drug discovery and development. The company has recently announced expanded partnerships with Amgen Inc. and Recursion Pharmaceuticals Inc.
Amgen's deCODE Genetics Building Nvidia Supercomputer
Amgen's subsidiary, deCODE Genetics, is partnering with Nvidia to develop a state-of-the-art supercomputer. This powerful system will be used to create genomics "foundation models" that are trained on massive datasets. These models will enable scientists to tackle a wide range of drug discovery challenges.
Nvidia Introduces Beta Testing for BioNeMo Platform
Nvidia has also introduced its own generative-AI platform for drug discovery, called BioNeMo. This platform is currently in beta testing and has gained significant interest from players in the computer-aided drug discovery field. Recursion and Insilico Medicine are among the companies adopting BioNeMo. By leveraging computational methods, BioNeMo allows scientists to rely on generative AI to optimize experiments and potentially even replace them.
Recursion's Foundation Model to be Available in BioNeMo
As part of Nvidia's collaboration with Recursion, the first third-party model available in BioNeMo will be Recursion's foundation model for drug discovery. This integration further enhances the capabilities of Nvidia's platform.
Revolutionizing Drug Discovery with Generative AI
Kimberly Powell, Vice President of Healthcare at Nvidia, emphasized the transformative nature of these advancements. She stated that the pharmaceutical industry is witnessing a groundbreaking moment as digital biology and generative AI reinvent the drug-discovery process. With an industry valued at $250 billion, the adoption of AI technologies has the potential to revolutionize the field.
Addressing the Costly Failure Rate of Traditional Drug Development
The momentum behind AI-powered drug discovery is driven by the industry's need to address the high failure rate and costs associated with traditional drug development methods. Currently, approximately 90% of drug candidates fail during clinical trials, and the journey to bring a successful drug to market can take up to 15 years and cost around $2.5 billion.
These developments by Nvidia have the potential to significantly accelerate the drug discovery process, reduce costs, and ultimately lead to more effective treatments for patients worldwide.
Major Drugmakers Embrace AI in Drug Discovery
In recent months, major pharmaceutical companies have been forging partnerships and collaborations related to artificial intelligence (AI) in drug discovery. Novo Nordisk and Valo Health Inc. joined forces in September to focus on new cardiometabolic programs, while Roche Holding AG's Genentech unit teamed up with Nvidia in November for AI-based drug discovery.
However, there are lingering questions about the potential of AI to enhance the speed and efficiency of drug discovery. The same way the chatbot ChatGPT sometimes generates fake responses, AI tools in drug discovery could suggest substances that are impossible to create, as pointed out by an editorial in the journal Nature. To overcome this issue, it may be necessary to manually code knowledge of molecular structures and employ additional AI tools.
Another challenge lies in collecting sufficient data for effective AI-powered drug discovery. Pharmaceutical companies need to find a way to share information without compromising their competitive advantage gained through exclusive datasets, as Amgen researchers highlighted in a paper published in Nature. One potential solution proposed by the researchers is the implementation of "federated learning," where each company updates a shared model using datasets without sharing the underlying data.
The Recursion foundation model, offered through BioNeMo, focuses on transforming images of human cells into mathematical representations of biology. Recursion CEO Chris Gibson expressed his belief that this model has the potential to be as impactful as genomics, which involves gene mapping and editing.
"We want to advance the field, and sharing this foundation model will accelerate the sharing of other models," Gibson explained, signaling a commitment to collective progress. Nevertheless, Recursion still keeps numerous aspects proprietary, including more than 50 petabytes of biological data used for internal programs and partnerships.