Novartis

Philippe MARC

what a year! Of course, 2020 will be remembered as the year where visio-conferencing and working from home became mainstream. But 2020 was also a glorious year for Data Science in drug discovery. Few examples: it took only two days for Moderna and the NIH to design their mRNA-1273 vaccine against SARS-CoV-2after they received the viral sequence. And deep learning methods did blow our mind again, enabling prediction of RNA expression and protein folding from raw sequences. Drug discovery has been using scientific computing for decades, but the availability of nearly unlimited computing power really transformed our capacity to both analyze data and get algorithms to the next level. It is really an exciting time where the previously undruggable targets are becoming druggable and no indication seems out of reach anymore.

That being said, the investment made by Pharmas, CROs, Biotechs and Startups companies in “Digital” and “AI” is so consequent that it is legitimate to ask if we get the full return of investment we expected. The Novartis Institute for Biomedical Research has been on a journey to expand use of Data Science for many years and we learned quite a lot along that path. We recently published what we think are the essential ten rules to create aData Science powered Research organization. Our set of rules is described in the figure below. If key expected elements of digital strategy like data and IT are represented in our 10 rules, most of them are about culture and organization. I will let the reader go to the full article for details of the rules, but I want to put emphasis and a single reality: more than ever drug discovery is a team sport. 

Data science enables business only when all elements aligns: computational scientist talents, paired to scientists knowing their domain, enabled by clean data, IT and management buy in, supporting the resolution of a clear problem. 

Subtracting any of these items will make the project a lot less effective. A popular misconception is to think that “AI” will identify gold in data in an unsupervised way. In our experience, biological data are still sparse and having a clear question in mind and knowledgeable domain scientists to scrutinize results are key to success. The crucial elements to impact the drug pipeline are still the following triad: clear question, enough clean relevant data, teamwork. Fundamentals of the game did not change, but Digital and Data sciences are both speed up activities and enabling new research avenues.

We have been quite successful at exploring these new avenues already, but new ones appears every day and a lot remains to be done. By definition research is about experimenting new things so we have a very exciting time ahead of us. I can’t wait to see how far that journey will bring us. I would really like to meet a time traveler visiting us from 2025 and ask him: Did we really got our data 100% FAIR? What are the computational methods that reached the plateau of productivity? Is quantum computing a reality yet? Is IT (finally!) non-obstructive? What is precompetitive and shared in drug discovery? Did data protection become a bottleneck for R&D? What are the “killer applications” of 2022? Overall: how quicker did we bring new drugs to patients thanks to Data Science?