NVIDIA, which has dominated the global semiconductor market, is expanding its presence in the healthcare sector by promoting its own artificial intelligence (AI) drug development platform. NVIDIA has added a service that designs therapeutic proteins, moving beyond just discovering drug candidates using AI.
NVIDIA noted that its ultimate goal in the healthcare sector is to achieve a 'digital twin.' A digital twin refers to accurately replicating physical entities in a virtual world and predicting and solving potential problems through simulation.
Kimberly Powell, vice president of NVIDIA's healthcare division, said at the 43rd JP Morgan Healthcare Conference, the world's largest pharmaceutical and biobased investment event, held on the 13th in San Francisco, that "we will collaborate with four corporations and institutions including IQVIA, a market research organization and clinical research services firm, Illumina, a genomics analysis company, Mayo Clinic, a medical institution, and Arc Institute, a research organization on data and technology utilization."
According to Powell, Mayo Clinic plans to use NVIDIA's Blackwell chip and the open-source medical AI platform MONAI to process its vast pathology data. IQVIA will develop an AI platform to accelerate clinical research using NVIDIA's AI Foundry, while Illumina plans to enhance genomic analysis using NVIDIA's accelerated computing technology and AI platform.
She explained, "The partnership with Illumina aims to leverage next-generation genomics to enhance drug development and improve human health," adding, "NVIDIA will build genomic data reflecting regional population characteristics." The regions for data construction disclosed by NVIDIA included only China and Japan in Asia. This is because NVIDIA learns based on open-source data from sources like the National Center for Biotechnology Information (NCBI) and the Ensembl Genome Browser, which do not include Korean data.
Powell said, "Our ultimate goal is to create digital twins of humans incorporating medical imaging, pathology, health records, and wearables," noting that "this collaboration will serve as a cornerstone for new applications in drug development and diagnostic medicine."
NVIDIA has added a protein design tool to its healthcare-focused generative AI platform, BioNeMO, unveiled last year. Generative AI creates three-dimensional (3D) models of proteins, while AI specialized in inference and reasoning structures optimal binding between proteins. While ChatGPT learned the rules of human language through massive training, BioNeMO has learned the language of 'amino acid sequences and protein structures.' This enables the prediction of protein structures and the generation of protein sequences, thus allowing for the design of therapeutic proteins.
Powell emphasized that "AI, armed with accelerated computing and biomedical data, is transforming healthcare into the largest technology industry," adding that "protein-based therapeutics have revolutionized medicine, from insulin to antibodies, as safe treatment methods, but they were complex and time-consuming, a paradigm that has now shifted with AI."
The traditional drug development process undergoes rigorous steps, including target discovery and screening, substance optimization, and toxicology testing, before entering clinical phases 1 to 3. Typically, it takes 4.5 to 10 years from target discovery to toxicology testing, and 6 to 8 years from clinical trials to approval. The entire process of discovering candidate substances and optimizing materials has required substantial manpower, but AI can replace this and accelerate the pace.