There is no shortage of discussion around the promise of artificial intelligence (AI) in general and machine learning (ML) in particular in drug discovery and development. But delivering on this potential requires much more than computation alone.
“At Roche and Genentech, advanced computational methods such as AI are permeating all aspects of R&D, creating a ‘lab-in-a-loop’ which allows a harmonious, iterative flow of data and information from experimental scientists back to computational teams and vice versa,” says Tommaso Biancalani, senior principal scientist and director, AI/ML Research Biology at Genentech. “This exchange not only accelerates our ability to run existing experiments, but also helps refine computational models that can then be used to inform new experiments.”
These computational approaches have been elevated as a core pillar of our R&D strategy. However, they are just one of four critical, interconnected “levers” for drug discovery. The other three include a deeper characterization of human biology, the ability to conduct experiments at exceptionally high resolution and massive scale, and the exploration and application of diverse therapeutic modalities. It is the interplay among these levers that has the potential to deliver multiplicative benefits for patients.
Levers in action
The time is right to integrate AI approaches, such as ML, with these other levers, as drug developers can now mine data for insights that are unreachable with traditional methods, at a scale and speed that were previously unattainable. In turn, this enhanced insight could uncover new therapeutic targets and improve the design and optimization of novel medicines.
For example, scientists in Genentech’s research and early development (gRED) group are using AI for small molecule drug discovery to examine the chemical structure of billions of potential novel antibiotics and determine which ones have the highest potential to be effective against antibiotic-resistant bacteria. With our acquisition of Prescient Design, we’re also applying AI to large molecule drug discovery. By bringing Prescient’s ML approaches and expertise in-house to work in a “lab-in-a-loop” alongside our antibody engineers, we work to discover and develop better antibodies and bring new medicines to patients faster.
Our colleagues at Roche are also leveraging AI in innovative ways. Scientists in Roche’s pharmaceutical research and early development (pRED) are using DELT-AI, which is a DNA-encoded library technology combined with AI, to identify and prioritize new therapeutic targets for small molecule discovery. Roche scientists are also using AI-driven pattern recognition algorithms to analyze imaging data from cancer and immune cells to identify rare antibody drug candidates that may help the body’s immune system to fight off cancer.
Collaboration across the healthcare ecosystem is also more important than ever as we look to realize the potential of AI. Through our partnership with Recursion Pharmaceuticals, for example, both Roche and Genentech’s R&D groups are exploring new territory in cell biology to drive the discovery and development of novel treatments in neuroscience and a specific area of oncology. The partnership will leverage Recursion’s ML-based and technology-enabled drug discovery platform in combination with Genentech and Roche’s extensive single-cell data generation and ML capabilities to cast a wide, comprehensive net for novel drug targets.
“Through collaborations, we get more than just the ability to create and build on large preclinical data sets,” says Barbara Lueckel, Head of Research Technologies, Roche Pharma Partnering. “Combining technologies and expertise allows for a much deeper understanding of human biology otherwise not possible. Partnerships also open the door to diversity of thought — critical for the continued advancement of this field.”
The drug discovery field is at an inflection point, with AI dramatically accelerating our experimental capabilities and helping to decipher unanswered questions related to every aspect of biology. “ML methods can be applied to all types of data, so every disease for which we can collect large datasets can benefit from adding ML to our strategies,” says Tommaso Biancalani. “So what is next? We continue to invest in researching and improving ML algorithms, and we continue to apply what we learn. When combined in our lab-in-a-loop with the other key R&D levers, the hope is that AI will allow us to deliver new medicines and targets that otherwise would have been very challenging to discover.”