Bayer accelerates drug research through data analysis in Microsoft Azure

The challenge: Evaluating data from clinical trials more efficiently

“Our goal is to provide patients with new treatment options as quickly as possible,” says Benedikt Egersdoerfer, Vice President at Bayer. He is also the company’s Head of Clinical Data Sciences and Analytics, overseeing just under 1,000 internal and external employees based all over the world. His team looks after the entire value chain for the clinical data used in Bayer’s drug research. But speed is not the only factor in play: “Efficiency and productivity are also absolutely crucial for us, because drug research has to be affordable to society,” Egersdoerfer says. “Anything else wouldn’t match Bayer’s identity. After all, our vision is ‘Health for all, hunger for none’.”

One obstacle on the way to getting drugs ready for the market is that research already generates a vast amount of data—and that amount is growing massively. In today’s clinical trials, traditional methods of collecting data have been joined by many new data sources such as sensors, devices, and third-party providers. A great way to process this abundance of data is to use artificial intelligence (AI) and machine learning (ML). Five years ago, Bayer had already recognized the importance of developing better ways to process clinical data. “We didn’t know where to store the growing amount of heterogeneous data because we didn’t know how big the datasets would get. At the same time, we were asking ourselves where to source the scalable computing power required to handle increasingly complex data processing,” Egersdoerfer says. 

Bayer also had to consider GxP requirements—good x practice, where “x” stands for the specialized field such as clinical: “When we process patient or other data gathered during clinical trials with a view to getting the drug approved, we always have to ensure that the data is consistently traceable and that our systems meet regulatory requirements,” Egersdoerfer explains. “However, none of our ML tools offered an adequate level of structure and validation for use in a GxP environment capable of handling sensitive data on people participating in clinical trials.”

This is because, as is common in large companies, Bayer operates a heavily fragmented IT infrastructure. “Considering the new requirements for device data and computing power, it was important that we consolidate everything on a scalable platform where we can store and analyze our data,” explains Abi Velurethu, Vice President Clinical Data Science & Digital Solutions at Bayer, who is responsible for Technology & Application Management. 

So, toward the end of 2019, Bayer launched the ALYCE (Advanced Analytics Platform for the Clinical Data Environment) to find a high-performance, scalable solution architecture. With Velurethu acting as project sponsor and Holger Dach, Data Science & Analytics Technology and Application Manager at Bayer, as business lead, the project team identified three initial use cases:

After conducting an exhaustive review of existing in-house solutions, the team decided to issue a public invitation to tender. They invited ten providers to present their products; the four highest rated provided demo versions of their products for a three-month period. At the end of 2020, the project team settled on Microsoft.

The solution: Data analysis underpinned by machine learning and based on Microsoft Azure

To map Bayer’s exacting requirements for collecting, storing, and analyzing data, the project team selected a solution architecture, built using Microsoft's Intelligent Data Platform, with data storage in Azure Data Lake. Here, the Azure tools for SecDevOps—in other words, the dovetailing of security, development, and IT operations—make it possible to automate and industrialize processes. In addition to core infrastructure services and Data Lake, ALYCE’s core components are: Azure Synapse Analytics, Azure Databricks and Data Factory, Azure Active Directory, Purview, Azure Kubernetes Service, and Power BI for visualization. Bayer also uses solutions from third-party providers such as SAS and TIBCO. “Azure lets us import all different kinds of data,” Velurethu says. “The platform’s programming interfaces were also appealing. They make it possible for us to tie in external analysis applications, dashboards, and programming tools.”

“Azure offers highly integrated platform services, but we also always have the option of adding something on,” Dach says. “For us, one of Azure’s major benefits is that all its services follow Azure Security & Compliance Blueprint for HIPAA. Thanks to that, we can setup these services in a qualified manner fully automatically and easily configure them to support our use cases. Azure saves us a lot of time and money because we have accelerated our qualification process and the final validation activities. That gives us the chance to focus on the business challenges which is a key advantage.”

Bayer has been using pattern recognition in its clinical trials since November 2021, and remote data collection since March 2022 with the new Azure technology. The first MIRA data visualizations are finished since summer 2022: “We’ve developed interactive dashboards for our oncology unit. They make checking tumor reactions as part of our clinical trials extremely interactive, which didn’t use to be possible,” Egersdoerfer says. “ALYCE has sped up this process significantly.”

There are also other ways in which ALYCE helps ensure speedy results: “The platform flags unwanted events, missing data, and information that might have been incorrectly reported,” Egersdoerfer explains. Dach offers an example: “Up to now, it’s been very difficult to recognize time series anomalies for clinical trials. But now, if the analysis software reveals that a patient was recorded as having suddenly lost 30 kg and then gained the weight back again shortly thereafter, then it’s clear that something’s not right.” Egersdoerfer also points out that, in addition to data cleansing, the ML technology helps identify risks: “We can now uncover any instances of fraud and misconduct in our trials, such as a patient breaking the rules by participating in two trials.” Bayer can then commission specialists to investigate on-site.

The benefits: Accelerated insights for more efficient research

“ALYCE helps us speed up our processes, which ultimately makes us faster at creating innovations,” Egersdoerfer says. “We’ve achieved all of our original use case goals, and even surpassed some.” What he is referring to is that ingesting external data was not originally part of the project. “Bayer was using a different solution, but ingesting external data proved easier in Azure. Personally, I felt that was the icing on the cake,” he says.

All this makes the project team confident about the future: “Now that we have ALYCE, I can imagine that our fragmented infrastructure might organically merge into a cohesive technology platform,” Egersdoerfer says. “This means we can focus on complex data rather than on complex technology.” Velurethu adds: “The journey we’ve embarked on with ALYCE and our partnership with Microsoft is providing us with new skills with which to meet future requirements. That’s a really big payoff.”

At the end of the day, the platform benefits not only Bayer’s employees but also the people for whom the company does its research: “What our team has achieved,” Egersdoerfer says, “allows us to use ALYCE to deliver the data required for drug approvals as quickly and efficiently as is possible today.”