Nvidia is making significant advancements in the field of artificial intelligence (AI)-powered drug discovery and development. The tech company has announced expanded partnerships with Amgen Inc. and Recursion Pharmaceuticals Inc. to further accelerate these efforts.
Amgen’s subsidiary, deCODE Genetics, will be utilizing an Nvidia supercomputer to create genomics “foundation models” for drug discovery. These models, trained on massive datasets, will provide researchers with valuable insights to fuel drug discovery processes.
In addition, Nvidia has introduced its own generative-AI platform for drug discovery called BioNeMo, which is currently in beta testing. The platform’s computational methods enable scientists to leverage generative AI to reduce the need for extensive experiments and even replace them entirely. Several players in the computer-aided drug discovery sector, including Recursion and Insilico Medicine, are already adopting BioNeMo.
The first third-party model available in BioNeMo will be a Recursion foundation model. Kimberly Powell, Vice President of Healthcare at Nvidia, highlighted the transformative impact of these advancements, stating that the pharmaceutical industry is being reinvented with digital biology and generative AI.
The integration of AI in drug discovery aims to address the challenges faced by the industry, such as the high failure rate and lengthy development process. Currently, approximately 90% of drug candidates fail at the clinical trial stage, and successful drugs take around 10 to 15 years and cost $2.5 billion to bring to market.
Despite the promise of AI, concerns remain. Just as AI chatbots can fabricate answers, there is a risk that AI tools in drug discovery may suggest substances that are impossible to create. However, researchers suggest that by combining hand-coding of molecular structures with the assistance of other AI tools, these challenges can be overcome.
Another obstacle is the need for sufficient data to facilitate efficient AI-powered drug discovery. Drug companies must find ways to share information without sacrificing their competitive advantage. “Federated learning” is proposed as a solution, in which each company updates a shared model using datasets without revealing the underlying data.
The Recursion foundation model offered through BioNeMo focuses on translating images of human cells into mathematical representations of biology. Recursion CEO Chris Gibson believes that this model will be as groundbreaking as genomics, and hopes that sharing this foundation model will accelerate progress for all stakeholders.
While much is being shared, many companies, including Recursion, still keep proprietary data. With over 50 petabytes of biological data used for internal programs and partnerships, there is still more to uncover in this evolving field.
The source of the article is from the blog combopop.com.br