At AstraZeneca, data and AI are more than game changers – they are life changers

As one of the world’s largest biopharmaceutical companies, AstraZeneca pushes the boundaries of science to deliver life-changing medicines that create enduring value for patients and society. To accelerate growth through innovation, the company is expanding its use of data science and artificial intelligence (AI) across the business to improve patient outcomes. 

AstraZeneca has been on a multiyear journey to transform its scientific capabilities to enhance its understanding of disease, design next-generation therapeutics, pioneer new clinical approaches, and better predict clinical success. For example, as part of its efforts to unlock different human genomes, AstraZeneca is working toward the analysis of up to 2 million individual genomes by 2026. This initiative alone has generated an explosion in the quantity and complexity of data the company collects, stores, and analyzes for insights. 

“We needed a new approach to manage and analyze that data to accelerate the delivery of life-changing medicines for patients,” said Gurinder Kaur, Vice President of Operations IT at AstraZeneca. 

The new approach involved federating its vast and globally dispersed data repositories in the cloud with Amazon Web Services (AWS).  Unifying its data within a centralized architecture allows AstraZeneca’s researchers to easily tag, search, share, transform, analyze, and govern petabytes of information at a scale unthinkable a decade ago. 

What began as an initiative focused on R&D now has extended to the company’s three other major business units: Commercial, Operations, and Clinical, according to Kaur. The goal, she explained, is to knock down data silos between those groups, using multiple data lakes supported by strong security and governance, to drive positive impact across the supply chain, manufacturing, and the clinical trials of new drugs. 

“Our ambition is finding a way to take these amazing capabilities we’ve built in different areas and connect them, using AI and machine learning, to drive huge scale across the ecosystem,” Kaur said. “Beyond R&D, we see value in extracting insights from data sources to improve patient outcomes and deliver personalized medicines.”

The cloud-based platform allows AstraZeneca scientists to move from ideas to insights faster, accelerating both drug discovery and clinical trials, to improve patient outcomes.

Moving from ideas to insights faster

AWS’s expertise with scaling cloud services was invaluable in helping AstraZeneca build an end-to-end machine learning platform, called AI Bench, to make it easier to apply machine learning across the enterprise. “AI Bench is a set of automated tools and guardrails that help us spin up the right environments in an automated fashion, so our data scientists can quickly begin working in a safe, secure, environment while ensuring regulatory compliance,” said Brian Dummann, AstraZeneca’s Vice President of Insights & Technology Excellence. “Before AI Bench, every data science project was like a separate IT project. We would spend weeks getting the right environment in place.”

Built on Amazon SageMaker, a service to build, train, and deploy ML models, AI Bench has accelerated the pace of innovation and reduced the barrier of entry for machine learning across AstraZeneca.  

“We have reduced the lead time to start a machine learning project from months to hours,” Kaur said. “This allows for engineers and data scientists to go from idea to insight quickly, delivering meaningful impact. Modern technology solutions provide our data science teams with fingertip access to synchronized information and data sets, allowing rapid re-use of models to ultimately accelerate outcomes and delivery for our patients.”

Accelerating drug discovery and clinical trials

More quickly moving from ideas to insights has aided new drug development and the clinical trials used for testing new products. AstraZeneca’s ability to quickly spin up new analytics capabilities using AI Bench was put to the ultimate test in early 2020 as the global pandemic took hold. 

“When Covid first appeared, we knew we had to step up quickly with our pandemic response,” Dummann said. “We were able to establish validated environments within 24 hours to begin working on evaluating Covid. This would have taken weeks or even months without the work we had already done to build out AI Bench.”

AstraZeneca’s increased investment in the cloud and AI capabilities offers the potential for a similar impact on clinical trials.  “Clinical trials currently account for 60% of the cost and 70% of the time it takes to bring a potential new drug to market^[1],” said Kaur. “AI and machine learning are helping us optimize that process and reduce the time it takes. The quicker we can complete clinical trials, the quicker we can get new medicines to patients.”

Four ways to improve data-driven business transformation 

Kaur and Dummann offered four pieces of advice to other IT leaders looking to get more value from their data transformation activities: 

Start small, think big, and scale fast. “You always need to have the big picture and vision in mind, but you don’t have to develop that picture right out of the gate,” Kaur said. Instead, focus on getting solutions out quickly, testing and improving them, and then scale them out across the company. The ability to scale also means promoting the re-use of data products where possible. “We want to maximize our investment in AI,” said Kaur. “We don’t want to keep reinventing the wheel, and we want our data scientists to be able to re-use AI assets across the enterprise.” AstraZeneca’s data scientists have launched more than 100 AI projects, and the number continues to grow.

Build internal expertise and understanding. Data-driven transformation is as much about people and process as it is about data and technology. To succeed, you need to get people to believe in the value of the transformation and show them a clear path to get there. “Attracting and retaining some of the best data scientists in the world has been critical to unlocking the value of data,” said Dummann. “So a big part for us is focusing on improving the experience of the data scientists. We’re keen to democratize data projects so that data scientists can get on with their daily tasks without reliance on IT. We don’t want to make them wait for weeks or days to get their work done.”

Modernize your approach to data and technology. “Data is an asset, and it needs to be treated as such,” said Dummann. It’s critical to ensure the integrity of the data for AI and machine learning models to work effectively. For the broader technology architecture, Dummann suggests moving away from best-of-breed point solutions. Instead, “invest in a few big, critical capabilities to really get the scale and speed you need.”

Don’t be afraid to fail. “There are multiple ways to solve a problem,” said Kaur. “Adjust as you go.”

Through its commitment to the cloud, data, AI, and machine learning, AstraZeneca is seeing its pace of innovation increase – and is eager to see where the journey leads. 

“Our data science community is moving faster than ever before, harnessing the power of data and AI to help discover new drugs, accelerate clinical studies and regulatory approvals, and maximize impact on patient lives,” says Dummann. “It’s an exciting time to be at AstraZeneca!” 

Learn more about ways to put your data to work on the most scalable, trusted, and secure cloud. 

---

^[1] Clinical Development Success Rates, 2006-2015. BIO, BioMed tracker, Amplion, 2016

Artificial Intelligence