Discussions of competitive advantage in AI drug discovery often turn to how much proprietary data a company holds.
Yet in this issue, companies working in drug discovery are themselves funding an effort to build open datasets together. Isomorphic Labs is one of them. Investing in a shared foundation and building a distinct business on it need not be contradictory. Participation announcement ↗
It helps to distinguish two kinds of data.
One is data used to learn how cells and the immune system behave. Many researchers and companies can use it repeatedly, adding to a shared body of knowledge.
The other is data needed to understand what is happening in the patient who is about to be treated.
However good an antigen-selection model may be, it cannot on its own reveal which mutations are present in that patient’s tumour. Personalised cancer vaccines still require someone to analyse the patient’s sample and pass the results into the design process. The collaboration with Tempus can be read as a concrete step towards providing that service at commercial scale. Collaboration announcement ↗
This suggests that more open data may change where value is created, rather than reduce the value of all data equally.
General knowledge can be shared, while a business is built around the measurements needed for each patient and the analyses that connect them to treatment design. The work arises each time a new patient starts treatment, rather than ending with a one-off sale of historical data.
Taking on that work does not guarantee high margins. If many companies can provide it, prices will come under pressure. If the work required for personalisation grows in proportion to the number of tests, profits may be difficult to increase.
Even so, we can ask a slightly different question when evaluating a company. Beyond how many records it holds, what new work does it receive when one more patient is treated?
Shared knowledge can support the provision of information needed for decisions and manufacturing for each patient. These two announcements point to ways of building a life-sciences business that extend beyond keeping data exclusive.