Building—and Tracking—a Data Strategy at the Novartis Institute for Biomedical Research
By Allison Proffitt
November 10, 2022 | At the Bio-IT World Conference and Expo, Europe, last month, Philippe Marc, Executive Director and Global Head, Integrated Data Sciences, Novartis Institutes for BioMedical Research, outlined NIBR’s updated data and data management strategy—going into transparent detail about how NIBR has arranged their data strategy vision and how they intend to assess it moving forward.
As part of the larger enterprise digital journey, and as part of the Novartis Research Master Plan, the Novartis Institute for Biomedical Research (NIBR) defined an updated data and data management strategy. This data strategy falls into a broader digital strategy which has many additional priority areas: Information technology, Artificial Intelligence, external science, decision support for drug discovery and early development, and more. At Bio-IT World Europe, Marc focused only on data strategy at NIBR.
Data is our core business, he explained, and for that reason, NIBR knew it needed to get those data under control. The NIBR portion of Novartis is, on its own, very large. More than 5,600 scientists, physicians, and business professionals work at NIBR, running approximately 300 research projects in eight disease areas. The organization has a budget of $2.6 billion.
The strategy, in a nutshell, Marc explained was to set up a “north star” for data topics, create a home for existing projects, identify gaps in the current data strategy and a path to improve, and better engage all partners.
Fundamentally, Marc said, the strategy is built on the assumption that data are pre-competitive—they should be used and shared. The resulting NIBR data management strategy grew around four pillars:
Data Culture: Treat data as a corporate asset;
Data Management: Structure and link data;
Data Science: Develop products and insights based on data; and
Data Enterprise: Lead the enterprise on data.
But, Marc warned, “Strategy is nothing but nothing but nothing if you don’t act on it.”
For each of the pillars, NIBR breaks out the core tenants. Creating a data culture at NIBR hinges on communicating broadly and frequently on data strategy topics to build a data citizen culture, setting data-related goals in every NIBR department and tracking them centrally, offering continued training on compliance and regulations, and making data governance clear, central, and accessible.
Culture is difficult, Marc acknowledged. While he applauded NIBR teams for their general openness, he said the view of “my” data is still present. Continued trainings, though, are making a difference, and soon each department will have data-related goals with associated key performance indicators.
The data management pillar is built on achieving “FAIR-enough” datasets, referring to data that are findable, actionable, interoperable, and reusable. No one really wants their data to be FAIR, Marc said—the bar is too high. Instead, he said, “We use FAIR as a proxy for data that are usable for what we want to do.” NIBR aims for what it calls FAIRPlay. “You want your data to exist with the right metadata, to be able to find, to be able to integrate,” Marc said.
Key to this type of data management is ensuring that data related processes are efficient and non-obstructive to decision making. NIBR has a centralized and actively maintained list of the most important scientific data assets, and regularly engages with and adopts Novartis’s global data management initiatives.
Legacy data can be a particular data management challenge, and NIBR’s approach is to only FAIRify or curate legacy data that offer a current competitive advantage to the company.
“With the explosion of AI, we see many people going crazy with curation of data,” Marc said. “We’ve seen also so many things where FAIRification was lost and did not bring any benefit.” He warned: “Don’t do it if you don’t know why!”
The third pillar of the NIBR data strategy intends to act on the data management goals. “We have so many tools it’s hard to find them,” Marc said. “We need to organize.”
Borrowing the F from FAIR, goals include enabling scientists to find existing datasets, tools, and subject matter experts; identifying and filling unmet data analysis needs; creating a comprehensive index of all datasets in their “home systems”; and listing priority data science activities to ensure data readiness.
“The hope is that that pillar will help people to elevate data topics to the level of the management a little bit easier than in the past,” Marc said.
Finally, the last pillar is one Marc said he “didn’t see coming,” though once the other data strategy components were spelled out, the need was clear. NIBR needed a strategic way to make data decisions moving forward, and the data enterprise pillar sets out these best practices.
The strategy ensures the long-term commitment to NIBR and the enterprise data management efforts and encourages researchers to prioritize use of global Novartis systems when possible, and if more local solutions are used, contribute resulting data back to central systems.
One key point in the Data Enterprise pillar is the valuation of data.
“Our data should have a price tag,” Marc said, a recent idea gaining ground among data scientists. “If you are a data company, your data should be integrated into your stock price,” he added.
Calculating the cost of your data is fairly straightforward, Marc said. If you acquired datasets, you know the price. If you produced them internally, you can calculate the cost of instruments and labor. But, he said, that is not the same as knowing the value of the data.
“You don’t really care how much was the cost of producing the data. You care what it will be used for. If it’s useful to you, that has a value,” he said.
NIBR has created two frameworks for data valuation so far, Marc said, but it still isn’t a solved problem. “If we can get that one right, that would be a very interesting thing for decision making. Then you can assess ROI.”
Operations and Implementation
With a strategy in place, NIBR is well-positioned to address data issues, but even good plans require regular attention. Marc explains that the NIBR Data Strategy calls for reporting on implementation every six months and strategy updates every two years. Each of the pillars currently has seven to nine large projects underway and four to seven coming up. May and June 2022 were the last reports to the NIBR Computational Sciences Council and Leadership Teams respectively.
At the top levels of leadership, Marc said, Novartis knows that achieving such an ambitious data strategy will not be quick or easy—but it will be worth it. “I would argue that data is actually what we produce at the end—that’s the only goal of research,” Marc said. “We need to have these data under control. That’s really our core business in research.”