Reuters Events | Pharma

Thought leadership and innovation for the Pharmaceutical Industry

AstraZeneca’s elite computer scientists are turning their skills to drug development.

By Adam Hill on Jul 6, 2015

With a PhD in high energy particle physics, and postdoctoral research into differences between matter and anti-matter, James Weatherall may not perhaps sound like a typical pharma man. But he has worked for AstraZeneca since 2007 and is now Executive Director and Head of the manufacturer’s Advanced Analytics Centre.

He sees analytics as the use of data within a company to aid business decisions and thereby add value to the business, and the centre is a hotbed of activity, with 30 people spread across the US, UK and Sweden within AstraZeneca. His team is split into four main groups:

• Statistical innovation
• Scientific computing solutions
• Biomedical Informatics
• Health Informatics

“There are overlaps between all those skills groups,” he says. In his academic work, Weatherall was “trying to make sense of big, messy, dirty data”, writing his own software to process what he found – and it is exactly this sort of work which pharma companies need to make sense of the tsunami of data which everyone now faces.

Big Data challenge

With data mining we can establish, for example, subsets of patients in whom a drug works best.We can apply statistical rigor to the design of clinical trials.”

Analyzing millions of electronic health records is one Big Data challenge, but Weatherall sees others in a world which is technologically-enabled to an unparalleled degree. Devices and sensors are everywhere – in supermarket tills, under car bonnets, in smart watches. “What do you do with all that?” he asks rhetorically.

The answer in AstraZeneca’s case is to examine things forensically, trying to identify patterns and trends which will have some commercial advantage. “It’s quite broad,” says Weatherall. “We help develop new medicines by trying to solve research problems based on data, modelling and simulation.”

There is a mass of clinical data on medicines out there already. “With data mining we can establish, for example, subsets of patients in whom a drug works best,” he continues. “We can apply statistical rigor to the design of clinical trials.” The team also analyzes care pathways, using its blend of scientific and technical skills - statistics, informatics, software programing and tool development – to develop and deliver cutting-edge solutions to critical scientific and business issues.

Reacting to questions

Weatherall’s team reacts to questions from within the business, and also strikes out on its own to examine the boundaries of what is possible – whether anyone at AstraZeneca has asked them to or not. “There is a balance between answering questions and blue sky thinking,” he says. But while there is a need to steer clear of being too prescriptive, or conservative, in their approach Weatherall knows the team’s work cannot be done in an ivory tower. He is careful to emphasize that the Advanced Analytics Centre’s focus is squarely on solving business problems. “We can’t become the ‘University of AstraZeneca’!” he laughs. “I’m very keen to get people communicating effectively about the value we bring.”

Recurring questions tend to be answered at the Advanced Analytics Centre through the development of new best practice, or the creation of a new software tool or type of training. “We might proactively bring new technologies to bear,” Weatherall suggests. “We have to maintain business relevance. We’re here to solve offbeat clinical data problems.”

Maintaining business relevance

Interestingly, there are statisticians within AstraZeneca – but outside the Advanced Analytics Centre – who can solve “80-90% of problems”, he suggests. “We’re there for the more challenging ones,” Weatherall goes on. “We do run into the issue that people don’t fully understand what we do, and in some ways a group like ours can be perceived as a luxury if we’re not careful.”

This is where the benefit of C-suite support comes in. Weatherall is extremely well connected to the very top of AstraZeneca through a management chain which starts with his boss – the Vice President of Biometrics & Information Sciences – who in turn reports to AstraZeneca’s Chief Medical Officer. From there, it is a direct line to the group’s Chief Executive Officer, which is a pretty short chain in an organization of 50,000 people. “We have top-level support,” Weatherall says. “Senior leaders recognize what we do: we have good visibility and I count myself lucky.”

Weatherall began with the Anglo-Swedish group working as an applied computer scientist on biomedical scientific problems in support of clinical research. But where on earth does his own academic specialism – particle physics and anti-matter - come into all that? “They are perhaps not normal in pharma but they are increasingly in data science,” he laughs. “Data scientists are not necessarily coming primarily from maths or statistics backgrounds, but from disciplines like engineering and physics. They bring a lot of experience of practical problem solving and working with experimental data.”

Understanding the domain

“A lot of people in pharma R&D may have more training in chemistry, biology or medicine,” he explains. “And it is true that data experts have to have an understanding of the domain in which they are working. But if they are smart, analytical problem solvers they will be good at learning new scientific disciplines. For example, I was able to go on that learning curve quite quickly – it didn’t matter that I hadn’t studied biology since I was 14 at school.” This approach represents a 180-degree turn from the thinking of 15 years ago, in which familiarity with the scientific subject matter was seen as the key, and analytics knowledge as secondary.

It all goes to show that data science is a compound discipline, which means Weatherall’s team members must have some facility with maths and statistics, as well as scientific domain knowledge – and practical computer skills. “You need enough knowledge of statistics to explore the data in a scientific and sensible way,” he points out. But computer scientists’ real skill today lies in using computers to manipulate data, dipping into a knowledge bank of software tools and programing languages to make sense of what they find: this is known as ‘data carpentry’. And there is always something else to learn: Weatherall maintains his academic links with a post as honorary reader at the School of Computer Science, University of Manchester and is looking forward to AstraZeneca’s move to Cambridge, UK, for the links it offers to the city’s university and other top research organisations which are based there.

Practical problem solving

We came up with a mathematical approach.We simulated various scenarios and distilled our learnings into a tool which colleagues in drug distribution can use.”

He highlights a couple of practical problems which the Advanced Analytics Centre has got its teeth into. The first concerned a problem with the distribution of an experimental drug in an international clinical trial: the logistical issue of getting the physical doses to the various sites, ensuring there is no wasteful overstocking – and indeed that there is enough of it in the locations where it is needed at any given time. “We came up with a mathematical approach,” he explains. “We simulated various scenarios and distilled our learnings into a tool which colleagues in drug distribution can use.”

It is hard to think of a less rarefied pharma problem than actually ensuring your products are where they need to be, suggesting that Weatherall’s team is as far from any ivory tower as he suggests. “It was a practical, logistics issue rather than a scientific one,” he concedes. “But it was inefficient and costing money before.”

The second example of the centre’s work is, in a different way, about making a process work more efficiently. In this case, the team used visual analytics to transform the way people conducting trials looked at data when dealing with patients as they made daily decisions over dosing and safety. “They were printing out figures on paper and trying to match them up,” Weatherall says. “We converted that data to a visual analytics framework.” He sees it as relatively simple, but the effect was significant, allowing trends to immediately be seen by looking at timelines, giving insight into how vital signs stood, for instance, and allowing decisions on dosage – or on a patient’s suitability - to be made much more quickly than before.

Looking to the future

When it comes to the future, Weatherall doesn’t believe the way forward is just about his team’s headcount. “We don’t need to expand per se,” he says. “We have a good critical mass and I suspect we will forever be in recruitment mode.” But he does think that the Advanced Analytics Centre’s work could move into other areas of AstraZeneca’s business. “There are opportunities elsewhere in the company,” he suggests. “We are focused on clinical trials and also looking at more on-market data. But what about the discovery part of the business, where drugs are being designed and tested?” He sees this as a rich domain for the use of Big Data in the genomic sense. “This would involve bringing together the molecular and clinical worlds, which we don’t do currently: we look purely at clinical data so there’s an opportunity there.”

He also believes that there is scope for more activity in the group’s commercial activities, using analytics from a non-scientific perspective to help AstraZeneca to better understand sales and marketing. “We collect a lot of data on interactions,” Weatherall says. “For example, how the sales force talks to the people who prescribe our medicines. Are we having the right conversations? Can we work towards greater segmentation?”

He has been working with researchers at his alma mater in Manchester on the latter issue, looking at clustering, which involves grouping patients in ways that the computer thinks is logical. “You can get different insights once you let the computer go forward,” he smiles. “So long as you are combining that with human experience, seeing if the combination can be greater than the sum of its parts. A.I. [artificial intelligence] doesn’t mean letting the computer take over, where you can run down all sorts of blind alleys: it means using the computer. The best programs marry human experience and common sense with the power of computing.”

It is a philosophy that AstraZeneca is clearly living by at present. And Weatherall is leading the way.