Nvidia, the Silicon Valley-based microprocessing giant, has expanded its growing footprint in artificial intelligence (AI)-based life sciences through the new AI drug discovery collaboration it has launched with Roche’s Genentech subsidiary.

The companies have committed to speeding up drug discovery and development by combining their mutual AI capabilities, Nvidia’s accelerated computing technologies and Genentech’s biological and molecular datasets and expertise in research.

John Marioni, PhD, senior VP and head of computational sciences at Genentech Research and Early Development (gRED)

“We’re ready to take that next leap in our use of AI, and that requires truly outstanding scientific computing. And it’s here that we’re so excited to partner with Nvidia,” said John Marioni, PhD, senior vice president and head of computational sciences at Genentech Research and Early Development (gRED), addressing reporters at a media briefing on the collaboration.

“They’ve pioneered the use of accelerated computing, and AI is positively changing so many industries and already beginning to have transformative impacts in the field of healthcare,” Marioni said. “We’re certain that working together, we’ll be able to similarly revolutionize how quickly we can train and use our foundational models in a variety of different contexts. Most importantly, this is going to change how we discover and develop medicines that impact lives right across the world.”

Like its parent Roche, Genentech aims to transform its generative AI models and algorithms into a next-generation platform. That would expand an AI presence that has yet to produce a clinical candidate, but has generated several collaborations with partner companies focused on developing such treatments.

Genentech and Roche are looking to advance therapies in 40 programs that include “key areas” of neuroscience and an undisclosed oncology indication through an up-to-$12 billion partnership with Recursion launched in 2021. That year Genentech acquired Prescient Design, developer of a deep-learning protein design platform designed to help identify and design antibodies.

A year earlier, Genentech teamed up with Genesis Therapeutics to use its deep learning and molecular simulation platform to discover small molecules for challenging targets that would elude other methods, and launched an alliance of undisclosed value with Reverie Labs to discover and develop next-generation kinase inhibitors using AI.

Also in 2020, Genentech named Aviv Regev, PhD, to head gRED. Regev is among leaders in AI, machine learning, and computational biology recently appointed to the scientific advisory board of the Kempner Institute for the Study of Natural and Artificial Intelligence at Harvard University. She co-authored a paper published in Nature Medicine earlier this year detailing an algorithm designed to predict the risk of pancreatic cancer by applying deep learning to real-world longitudinal datasets of disease trajectories.

In its new partnership with Nvidia, Genentech plans to accelerate and optimize its machine learning (ML) algorithms and models through NVIDIA DGX™ Cloud, designed to facilitate generative AI applications in drug discovery through a training-as-a-service platform built dedicated Nvidia AI supercomputing and software. That includes BioNeMo™, a generative AI cloud-based service designed to enable faster discovery and design of drugs.

Genentech said its AI/ML teams are developing and leveraging foundational models across numerous research areas including diverse therapeutic modalities, with the aim of gleaning new insights for target and drug discovery and answering fundamental questions about human biology and disease.

Therapeutic areas undisclosed

Addressing a GEN Edge question during the briefing, Marioni said Genentech is not disclosing the therapeutic areas in which it will pursue targets in its collaboration with Nvidia. “We would anticipate the types of work that we’re doing here having an impact right across all of the areas within our organization,” Marioni said. “The impact should be broad because of the potential of generative AI and the ability to use accelerated computing to really drive that.”

Genentech’s pipeline focuses on immunology, infectious disease, neuroscience, oncology, and ophthalmology. Roche’s focus areas in pharma includes all those plus hematology, rare disease, respiratory, and women’s health.

Marioni said the collaboration will also help accelerate Genentech’s “lab-in-a-loop”, through which it feeds extensive amounts of experimental data into computational models designed to uncover patterns and formulate new, experimentally testable predictions. The results from lab tests are fed back into the models to improve the underlying computational model, an approach that according to Genentech will improve the development of therapies.

“This lab-in-a-loop requires multiple things to make it work,” Marioni explained. “In house, we have access to literally hundreds of petabytes of proprietary data of numerous types. We could complement that with public data. It also requires deep expertise in the biological and the clinical sciences, something that is at the very heart of Genentech as well as in AI and computation.”

Genentech said it will control the sharing of its proprietary data, with Nvidia not having direct access to that data unless granted by Genentech for use in a particular project during the term of that project.

Marioni leads gRED Computational Sciences (gCS), a central organization within gRED in which more than 400 employees focus on integrating data, technology and computational approaches in order to revolutionize how targets and therapeutics are discovered and developed.

The companies say their collaboration could expand from its current emphasis on early target discovery and development of molecules, gCS’ sweet spot, toward applications that include manufacturing. “It could be that there are interesting opportunities to explore that as we move forward, but that’s not our initial focus,” Marioni added.

Kimberly Powell, vice president of healthcare at Nvidia

Genentech and Nvidia did not disclose the value of their partnership or its duration, other than to call it a multi-year arrangement. “Our teams will be continuously exchanging expertise on the advancement of science and the state-of-the-art methods emerging in accelerated computing, AI and simulation across this entire drug discovery process,” said Kimberly Powell, vice president of healthcare at Nvidia.

Powell noted the vastness of the potential universe of molecular structures waiting to be discovered: The number of chemical compounds has been estimated at 1060, while the number of atoms in the observable universe is estimated at between 1078 and 1082.

The number of potential proteins is twice that range, she said, at an estimated 10160.

“This is a near infinite search space. And what other than generative AI would allow us to search it?” Powell said. “That’s why we built BioNeMo, to use generative AI to intelligently search this enormous complex space. We can search it for targets, we can search it for leads and we can even intelligently search it for optimized molecules.”

Biopharma partners

Genentech is the latest of several drug developers that are partnering with Nvidia in AI-based drug development:

  • Amgen, an early user of BioNeMo, has shrunk from three months to four weeks the time needed to pre-train large language models for molecular biology on its proprietary data, according to Nvidia.
  • AstraZeneca partnered with Nvidia in 2021 to develop MegaMolBART, a transformer-based generative AI model for chemical structures used in drug discovery, envisioned to help researchers conceptualize molecules that could be potential drug candidates but do not exist in databases. MegaMolBART was among the first projects to run on NVIDIA’s Cambridge-1 supercomputer, launched that year.
  • GlaxoSmithKline (GSK) joined Nvidia to launch an AI drug discovery partnership in 2020 that was also tied to use of the superco

Also partnering with Nvidia has been Insilico Medicine, a Hong Kong-based designer/developer of drugs using generative AI. Insilico uses NVIDIA Tensor Core GPUs in its generative AI drug design engine, Chemistry42, designed to generate novel molecular structures—and was among first adopters of an early precursor to NVIDIA DGX systems in 2015. Insilico is also a premier member of NVIDIA Inception, a free program that provides cutting-edge startups with technical training, go-to-market support and AI platform guidance.

Nvidia’s growing focus on AI in drug development has enabled the company to emerge as one of the biggest beneficiaries as investors have hopped on the AI bandwagon over the past year. During 2023, Nvidia’s stock has more than tripled, rocketing 252% from $143.15 to $504.09 as of November 20, before dipping to $482.42 on Monday.

A report issued in June by Boston Consulting Group (BCG) and commissioned by Wellcome Trust found 72 AI-based drug candidates in clinical trials, of which the largest share—nearly half (32)—were developed for indications in oncology. A limited range of indications favoring those with commercial potential, such as oncology, was one of five barriers to AI drug development identified by BCG and Wellcome. The other four are:

  • Concerns about what rapid advances in AI could mean for science and wider society.
  • Lack of high-quality data sets, access to mature tools, and relevant AI and drug discovery capabilities.
  • Access to inter-disciplinary capabilities such as computational chemistry and bioinformatics
  • The dearth, absence, or lower quality of longitudinal population datasets that can be mined to understand diseases and identify new targets, particularly in low- and middle-income countries.

BCG and Wellcome called for increasing trust in and understanding of AI in drug discovery; further developing of high-quality datasets and AI tools; as well as expanding AI capabilities.

“A number of mature use cases are delivering value today and provide immediate opportunities to help researchers discover new medicines to improve human health. At the same time, barriers risk concentrating the benefits of AI to already data-rich and commercially attractive TAs [therapeutic areas] with limited opportunity for researchers in other areas to engage,” BCG and Wellcome observed.

“Concerted action is needed today to shape this emerging field and set the ‘rules of the road’ that will allow equitable benefit from the transformational opportunities of AI in drug discovery,” BCG and Wellcome added.

Alex Philippidis is Senior Business Editor of GEN.

Previous articleLung Cancer May Be Halted by Tumor Suppressor Protein
Next articleNucleus RadioPharma: Building CDMO Sites to Ride the Radioligand Therapy Wave