Enterprise AI Profile: The Roche Group
Futuriom Enterprise AI Profile
Organization: The Roche Group and its subsidiary Genentech
Vertical industry: Healthcare
Description: Roche and Genentech discover, develop, and manufacture pharmaceutical drugs, and Roche also specializes in diagnostic products, such as cancer markers and other pathology tools.
The companies work in a sector in which nearly all proposed drugs fail to make it to market, while those that do can take a decade to develop. For several years, both Roche and Genentech have engaged AI and machine learning to streamline the development of pharmaceuticals.
Core to their strategy is what they term a "lab in a loop" process, which starts with training AI models with data from the lab and clinical trials to locate potential drugs or molecules where drug therapies might be applied. The results are then retested in the lab, and those results are used to train the models further. This streamlines the traditional trial-and-error approach to pharmaceutical development, speeding up drug discovery and effectiveness.
“The ‘lab in a loop’ is a mechanism by which you bring generative AI to drug discovery and development,” says Aviv Regev, Head and EVP of Genentech Research and Early Development. The technique is used for a variety of AI-infused applications, including neoantigen selection of proteins for personalized cancer vacccines; molecular simulation, in which AI produces simulated modules to test therapies; antibiotic discovery; and antibody optimization.
Notably, to institute lab in a loop, Roche had to organize its data across multiple systems in over 80 countries. Reportedly, this involved a five-year plan that included reorganizing engineering teams, prioritizing DevOps, and enforcing CI/CD pipelines across the organization.
In drug manufacturing, Roche uses AI to refine processes, reducing waste and improving output. Using AI-powered predictive modeling, Roche can optimize process settings, predict any potential problems, and address any deviations from critical quality attributes (CQAs).
Roche also uses AI in its diagnostic products, including imaging of lung cancer legions with AI models, and applying AI algorithms to the identification and evaluation of tumors for various forms of cancer. Roche also uses AI to analyze imaging of the eye to help ophthalmologists identify diabetic macular edema.
AI Platforms and Models Used: Roche was one of the early adopters of NVIDIA’s BioNeMo Framework, a platform for developing AI models for pharmaceutical and biotech companies. This suite of tools and services helps streamline molecular design, which is essential to pharmaceutical R&D, as well as the creation and customization of models for drug discovery (finding the molecules for which drug therapies could be applied). Along with the BioNeMo Framework, Roche and Genentech also use NVIDIA DGX Cloud, which provides managed AI resources, including specialized hardware, software, and services, on hyperscaler cloud environments, including AWS, Azure, and Oracle Cloud. (Roche has reportedly used AWS to host DGX Cloud.)
Other Vendor Partnerships: The Roche Group has partnered with numerous companies in its AI journey. It has also acquired technology and formed partnerships for its subsidiaries. For example, Roche last year announced that is collaborating with eight companies (Deep Bio, DiaDeep, Lunit, Mindspeak, Owkin, Qritive, Sonrai Analytics, and Stratipath) to integrate over 20 AI algorithms into its Digital Pathology Open Environment, a platform that integrates Roche’s software for pathology workflow coordination with third-party AI-driven image analysis algorithms to speed diagnoses. This year, Chugai Pharmaceutical, a Roche subsidiary based in Japan, joined up with Gero, an AI-based biotech firm, to target therapies for age-related illnesses.
Key Success Factors: Roche estimates that AI reduces drug discovery timelines and costs by 25% to 50%. AI reduces clinical trial costs by up to 70%. In drug manufacturing, AI increases manufacturing yields by 5% to 10% and reduces write-offs for defective products by 50%.
Note: According to Futuriom's Enterprise AI Index sample of over 130 major companies worldwide, the healthcare vertical comprises over 20% of AI implementations:
Futuriom Take: The Roche Group, including Genentech, offers an example of how AI is being used to reduce failure rates and improve delivery times in the pharmaceutical industry. The company also has demonstrated the power of partnering with AI companies and of instituting a methodology for organizing data for use in AI.
