Genentech leverages generative AI to get lifesaving medicines to patients faster

Developing new medicines is an enormously complex challenge due to the sheer scale of biological possibilities that must be explored. The human body contains thousands of different cell types and states, interacting with roughly 20,000 genes in the genome. Scientists must search through an astronomical number of potential treatments, including 10⁶⁰ different small molecules with medicine-like properties and 20³² possible therapeutic antibodies. This scale makes it impossible to test every possibility through traditional lab experiments alone.

To navigate this complexity, Genentech's scientists must analyze massive amounts of scattered data, including 35 million biomedical journals in PubMed, public repositories like the Human Protein Atlas, and their internal repository of hundreds of millions of cells across hundreds of diseases. Scientists must not only identify promising drug targets but also validate biomarkers—biological indicators that help track disease progression and treatment effectiveness. Currently, this process is largely manual, with scientists spending weeks searching through data sources, totaling nearly five years in biomarker validation across therapeutic areas. 

Genentech partnered with Amazon Web Services (AWS) to develop a solution that automates the time-consuming manual search process, enabling their scientists to focus on high-impact research and accelerate drug discovery innovation. It designed a generative AI system called gRED Research Agent, built using Anthropic Claude Sonnet 3.5 in Amazon Bedrock Agents, which transforms how scientists interact with vast scientific datasets. The system can process complex scientific queries like "What cell surface receptors are enriched in specific cells in inflammatory bowel disease?”, simultaneously search through multiple data sources from PubMed journals to internal repositories, and synthesize findings with cited summaries—ensuring the scientific rigor and traceability critical for drug development.

What makes this solution powerful is its use of autonomous agents that can break down complicated research tasks into dynamic, multi-step workflows. Unlike traditional automation systems that follow predetermined paths, these agents adapt their approach based on information gathered at each step, access and analyze multiple knowledge bases using Retrieval Augmented Generation (RAG), and execute complex queries by interfacing with Genentech's internal APIs and databases.

The impact has been transformative for the company’s scientists. John Marioni, Senior Vice President and Head of Computational Research, at Genentech observed, “One of the things that I'm especially excited about, through the development of tools such as autonomous agents, is the ability to democratize access to data sets and computational tools to scientists who maybe have less computational background. Through interacting with an agent in a really iterative and coherent way, we’re able to take advantage of these tools and use that to directly accelerate the research. The agent really gives us an unbelievable boost in the work that we're trying to achieve.

The gRED Research Agent, powered by Amazon Bedrock Agents, transforms processes that once took scientists weeks to accomplish into tasks that take just minutes. This is expected to help Genentech automate over 43,000 hours – nearly 5 years — of manual effort in biomarker validation across therapeutic areas, reducing time-to-target identification. This will ultimately enable Genentech to bring new medicines to patients much faster.

Aviv Regev, Executive Vice President and Head of Genentech Research and Early Development observed “Agents can't replace our scientists, but they actually boost our scientists. The really cool part happens when you let the agent and our top performing scientists work together. They then perform together far better than our best scientist, and that means that our best scientist is now a better version of themselves.”

Genentech is building for the future by developing a data infrastructure that allows agents to seamlessly access and integrate data across their organization. The strategy involves expanding beyond a single comprehensive agent to a network of specialized sub-agents through Amazon Bedrock's multi-agent collaboration capabilities. This approach creates dedicated agents for specific domains—for instance, one agent focusing on public data sources like PubMed, while another specializes in querying Genentech's internal Single Cell Hub database. This specialized agent approach generally results in greater speed and precision, as multiple focused agents can work in parallel while being easier to maintain, test, and scale.

Finally, Regev concludes that “There are many things that only humans can do, with their brilliance, creativity, and ability to push the frontier. Genentech's ultimate benefit of using our gRED Research Agent is that it's able to identify drug targets, and to design studies that will get lifesaving medicines to patients faster.”

Learn how Genentech is Automating Drug Research with Amazon Bedrock Agents | Amazon Web Services