A few weeks ago, I rode to Basel to find out how the pharmaceutical industry uses Artificial Intelligence. I looked at Novartis’ innovation called the Sense Bridge – a room with six screens lit up with colored dots, bars, curves and maps. Tables and chairs are set in a semicircle around the screens. The idea is to give the employees not bare numbers, but to present the meaning in the numbers. That’s why the room is called “Sense Bridge”. But why bridge? I ask. Should the data bridge something? It’s more like Star Trek, I got in response.
Over the last two decades, Novartis has collected two million patient years of data from clinical trials. Only a few competitors are able to demonstrate such a thing even approximately. The Novartis Institute for BioMedical Research (NIBR) has about 15 petabytes of data: scans, videos, chemical data, study results, data on 1.5 million molecular compounds. If you had the idea of storing them on DVDs, you would need about three million of them. The data from the Sense Bridge comes from nine different sources.
But what does the Sense Bridge actually do?
The visualization provides an overview of all clinical trials conducted by Novartis worldwide. Not all are in the system yet, some are missing, but there are many. 440 intervention studies testing new drugs with patients. It takes an average of 10-12 years for a drug to go through clinical trials and each one engulfs nearly two billion dollars on average. 85-90% of all studies do not end in an approval to enter the market.
The system shows the current status of each study and the artificial intelligence predicts the further course. There are five milestones for each study: the first patient comes to the study center; the last patient starts the first treatment, the database is locked because all data is collected, the trial report is ready and the drug is approved. The milestones have colors: red means a high risk that the timetable is not adhered to. With yellow the risk is lower, with green the study is on schedule.
Everything runs automatically – the AI retrieves the data and recalculates the risks with each input – it learns every minute. If a study does not go as planned, for example in Argentina, because the regulatory authorities there operate too slowly, the AI system can recommend that some of the patients be relocated to Brazil. The AI knows for which type of active ingredient which country or region or population group is the most suitable. It calculates this from historical data, creates a trial footprint optimizer.
The pharmaceutical companies have understood that patients can be subdivided into so-called subgroups, where therapy is more effective than other people with the same disease. Tracking down such a group is therefore essential. An arbitrary group is guaranteed to delay.
How well the Sense Bridge works, and what comes out in the end for the patients, I can not judge yet. At Novartis, all people I have talked with assure me that they have patients in mind, and I believe them, but company dynamics sometimes overwrite the intentions of employees. As a patient, I only show up as a number, but many people would suffer very much or sometimes not even be alive without the products that are being developed here. So it’s a contradictory feeling to be a number.
The visit to Basel resulted in an article for brand eins: Kampf um die Patientendaten