Redefining Drug Discovery with AI

AI is evolving how new medicines are being discovered.

Seemingly overnight, artificial intelligence (AI) has become an integral part of our everyday lives. Streaming services recommend the next show to watch, chatbots help us write our emails, driverless taxis navigate for us – and the list goes on.

While mainstream applications dominate the daily headlines, the AI revolution is also well underway in the healthcare industry. At Genentech, we have been pioneering the use of AI in research and development to transform and accelerate how we discover, design, and test new medicines.

Overcoming the Data Challenge

Making new medicines is complex. It is a time- and resource-intensive process that often results in limited success. Industry-wide, about 90% of drug candidates fail, and it can take more than 10 years to find out if they will even work. Not only do we need new approaches that allow for more shots on goal – we need to improve the quality of our shots.

But the staggering scale and complexity of the science – and associated data – underlying drug discovery and development remain major barriers to progress. For example, the number of possible small molecules – the type of medicines in pills – and therapeutic antibodies that could be considered to treat the approximately 20,000 different conditions far exceeds the number of stars in the universe. While computational approaches have reshaped the collection and analysis of data, prior computational approaches alone have not dramatically improved the speed of bringing new medicines to patients or the likelihood that research programs will succeed.

After years of investment and continuous scientific innovation, Genentech is at an inflection point – where an accelerated drug discovery process is driven by the convergence of data and technology. Computational approaches in biotech have evolved far beyond crunching numbers and running statistics. AI and machine learning (ML), a subset of AI, are helping us surmount the data challenge by generating better insights, hypotheses, and predictions that streamline the classic process of trial and error in the design and testing of novel therapies.

Lab in a Loop

Recognizing we are at this inflection point, at Genentech, we pivoted our approach to make AI a core and integral part of our entire drug discovery and development approach, leading the way in utilizing AI to address major challenges in the process. The foundation of our strategy centers on creating a “lab in a loop,” where data from the lab and clinic feed AI models and algorithms – that are designed by our researchers – to identify trends, make predictions, and generate new molecules. After our scientists conduct experiments based on these predictions and designs, they input the results back into the model, thereby improving its performance across all programs. It's an iterative cycle of data, computation, experimentation and discovery.

The “lab in a loop” strategy involves training AI models with massive quantities of data generated from lab experiments and clinical studies. These models generate predictions about disease targets and designs of potential medicines that are experimentally tested by our scientists in the lab. This leads to new data that can train and further improve the models. This iterative process of data - AI model - experiment is repeated with the goal of accelerating the discovery of new medicines.

And its impact is already being felt. For example, AI is enhancing our efforts to develop a new class of therapies called personalized cancer vaccines. Cancer vaccines work by training the immune system to recognize proteins generated by tumor-specific mutations, called neoantigens. However, not all neoantigens are equal, with some acting as better cancer vaccine targets than others. AI approaches are being used to select which neoantigens have the highest chance for success, and these algorithms are actively helping design individualized cancer vaccines for patients.

AI and ML also enable us to rapidly generate the virtual structure of thousands of new molecules and simulate their interactions with therapeutic targets, accelerating the discovery and testing of new potential treatments. We are using this and other AI strategies to optimize the design of antibodies, predict the activity of small molecules, identify new antibiotic compounds that can address the hardest-to-treat bacterial infections, find new disease indications for investigational therapies in our pipeline, and much more.

Despite these advances, the use of these technologies in drug discovery presents its own challenge: we need increasingly powerful computing capabilities to train the algorithms and process the ever-growing amount of data. That’s why we’re bolstering our ongoing work in AI and ML through collaborations with leading technology companies.

Building a Next-Generation Drug Discovery Platform

Across Genentech and the Roche Group, we’ve built massive, robust datasets sourced from decades of laboratory and clinical research and used them to train powerful AI models that help us solve formerly intractable problems every day. To enhance the capacity and speed of our ongoing and planned efforts, in November 2023, we announced a collaboration with NVIDIA. Using our unique resources and expertise, Genentech is creating an industry-leading, next-generation drug discovery platform that harnesses the power of generative AI.

Through the collaboration with NVIDIA, Genentech is enhancing the power of our proprietary ML algorithms and models using NVIDIA accelerated computing and software, including platforms like NVIDIA BioNeMo that scale generative AI applications in drug discovery.

Similar to what has occurred in areas like gaming, search engines, and text-to-image generation, generative AI is acting as a catalyst in healthcare, and our collaboration with NVIDIA is helping us to amplify our “lab in a loop” framework. This boost in computing power could ultimately speed up our drug development process and improve the success rate of R&D.

Looking to the Future

Ten years ago, few people would have guessed that AI would allow us to create paintings in seconds using a few lines of text. Or that it would become a seamless, indispensable part of how we travel, shop, listen to music, consume news, and so much more. As with those now familiar applications, it will take time to realize the full potential of AI in drug discovery and development.

But we're well on the way. The proof-of-concept phase is nearly over, and we're beginning to see real progress throughout many aspects of our work.

In the next decade, the impact of AI on human health will likewise be unimaginable. By helping us untangle disease biology, predict which approaches could fix the underlying defects, and design better therapies faster, AI will enable us to treat or cure many illnesses that have eluded us with the goal of extending and improving the lives of millions of patients.