Pharma and data as gospel

Source: Het Financieele Dagblad (11/03/2017); Author: Arnoud Groot

Based on the case study ‘UCB – Data is the new drug’ written by Stijn Viaene, professor of DigitalTransformation at Vlerick Business School. 

Developing drugs is a time-consuming process. In order to speed it up, the Belgian biopharmaceuticals producer UCB has decided to switch to a data-driven organisation. It was necessary to ‘convert’ the entire company to achieve this, says CIO Herman De Prins.

In February 2011, Watson briefly made headlines all over the world: for the first time, the supercomputer had managed to beat human contestants in the American quiz show Jeopardy.

It was a special moment for owner IBM, but also for Herman De Prins. The power of the self-learning computer Watson encouraged the CIO of biopharmaceuticals producer UCB to take a critical look at both the use of data and the IT department within the company, a multinational which operates in forty countries.

‘UCB mainly develops drugs for neurological and immunological disorders such as epilepsy, Crohn's and Parkinson’s disease’, says De Prins (55). ‘In total, we invest over a quarter of our turnover – which is just under €4 billion – in R&D. This allows us to generate a huge quantity of data for internal purposes, both during the research into new drugs and when testing them. This data may relate to the latest insights into the development of the diseases we are fighting and the search for chemical substances which can affect the course of the disease, for example. Partly thanks to Watson, I realised that we could do much more with all this data.’

Cognitive computer systems such as Watson have a computing power of dozens of petaflops. One petaflop represents a quadrillion calculations per second, or a million to the power of four. This sheer processing power allows huge databases to be searched for relevant data or patterns in mere fractions of seconds. At a conference a few months after the Jeopardy stunt, De Prins met John Kelly, who is in charge of Watson at IBM. Animated discussions about the possibilities of data analytics in the pharmaceutical industry and healthcare sector led to the launch of a joint project.

‘The aim of this project was to develop a “clinical decision support system”’, says De Prins. ‘In principle, doctors who make a diagnosis can call upon their training, professional experience, colleagues and the professional literature which crosses their path’, he explains. ‘But new studies, best practices, reports and patient cases appear in this field almost every day. It's not humanly possible for a doctor to keep track of it all, but this information is still recorded in our databases. And supercomputers such as Watson now make it possible to search this data for the most relevant information in just a few seconds.’

Lookalikes

Among other things, the joint project with IBM should soon make it possible to find previous patients whose symptoms bear a strong resemblance to those of a patient who requires treatment.

For example, it will be possible to look at the medical and treatment history, frequency and intensity of epileptic seizures and other distinctive characteristics. This anonymised medical data will then allow doctors to find out how these ‘lookalikes’ responded to treatment with specific drugs. ‘This knowledge allows the doctors to come up with far more accurate diagnoses and treatment plans, and much more quickly too’, says De Prins.

In order to establish which information ties in best with a specific patient’s circumstances, their electronic patient records will need to be linked with the ‘clinical decision support’ system. UCB works in close collaboration with various parties including the renowned Georgia Tech university, which ensures that the required data standards are upheld and the patient privacy is protected.

The large amount of anonymised ‘real-world’ patient data received as a result is extremely valuable for UCB. Once analysed, for example, this data can provide very detailed insights into the course of the illness in thousands of patients and the impact of the drugs administered. In the future, algorithms may even make practical and strategic suggestions for the use of new drugs or promising areas of focus for new research. In order to make effective use of these opportunities, De Prins has been working extremely hard in recent years to develop a new company culture based on data analytics.

‘Six years ago, staff almost exclusively made decisions based on the often relatively limited amount of information which they could collect and process themselves’, explains De Prins. ‘It was a big step from there to the current data-driven decision-making process in which important choices are made by algorithms that are often fairly incomprehensible. That was certainly true in in the pharmaceutical sector, where people tend to want to see proof before they commit to something.’

The company's own IT department also needed to switch from a primarily supporting function to a much more strategic role. In order to tackle these barriers, De Prins launched his Future of IT programme at the end of 2011. At intensive workshops on topics such as cloud technology, social and artificial intelligence and analytics, UCB's 240 permanent IT staff learned that ‘digital’ would be playing an increasingly critical role in every section of the company. And that they would be expected to respond proactively with ideas and projects which did not revolve around project-based IT but focused on improving patient services: the ‘patient value’. As a result, one of his staff came up with the idea of a sensor that could be attached to an epilepsy patient’s body to predict seizures. In December 2016, a consortium initiated by UCB started production of these sensors.

In order to roll out the huge opportunities offered by data analytics across the entire pharmaceuticals company, with its staff of 8000, De Prins appointed Arnaud Lieutenant (formerly of IBM) as the Director of IT Advanced Analytics at the end of 2012. One of his first missions was to set up a series of ‘showcases’: projects which could achieve fast, visible results with the aid of analytics. De Prins explains: ‘We set clearly defined goals beforehand which we needed to be able to achieve in short sprints of a maximum of 50 days. Each time you succeed in getting fast results with this “agile” working method, your employees, departments and the company as a whole get to experience the value of analytics first hand.’

For example, the dozens of showcases include a project which gives UCB a better overview of ongoing and issued patent requests. ‘In our sector, many thousands of patents are requested every year’, says De Prins. ‘Performing a “text mining” analysis in this area allows you to improve your insights into developments and trends, and also shows you which talented researchers are working on research which is of above-average interest, for example. We sometimes approach these researchers for a partnership, and sometimes we just offer them a job.’

Evangelists

In order to guarantee maximum visibility, results, new insights and best practices are shared with the entire organisation by means of a new portal. ‘We used A/B tests, which compared the results of the new and old methods, to highlight the differences with the traditional working method’, says De Prins. ‘As a result, we were able to gradually convince our staff of the added value, and simultaneously built up a network of internal “evangelists” who started to actively promote the new working methods. This meant that analytics was not just the domain of our IT experts, but an important focal point for the organisation as a whole.’

De Prins, who was voted Belgian ‘CIO of the Year’ in November, finds it difficult to specify any concrete, clear-cut results of the working methods which he started to introduce six years ago. ‘It's impossible for me to say that we have reduced the development time for a specific drug by a certain amount’, he says. ‘We simply don't have a baseline measurement for that. In addition, these paths are always a unique combination of knowledge, the analysis of internal and external data and sometimes hundreds of thousands of different tests. However, as our efficiency increases, we are saving considerable amounts of time on each individual test and analysis. And we can also examine more and more options in each analysis, which in turn may lead to more new insights and better drugs.’

By way of an example, De Prins refers to the lightning-fast developments in mapping human DNA, a field in which UCB is also conducting an increasing amount of research. The first 1.5 terabyte (1500 gigabyte) data set was provided in 2003 by a few hundred scientists, who spent 13 years and $3 billion creating it. These days, a human being’s DNA can be sequenced within a day at a cost of less than a thousand dollars.

‘Thanks to new technology and methods, we are now also achieving similarly spectacular results at UCB’, says De Prins. ‘For example, we recently improved the speed at which specific sets of real-world patient data can be analysed within our company by a factor of 3600.’ In the future, this could lead to an entirely new Analytics as a Service (AaaS) revenue model, whereby UCB could share its knowledge of data and technology with organisations and consumers directly by means of apps and APIs. ‘The groundwork for this is currently being laid’, says De Prins, ‘but it is still a pipe dream for the time being’.