Personalis – Personalized medical platform for patients with autoimmune diseases (data hospital)

In 2015, 18 Phase 2/3 clinical trials in systemic lupus erythematosus were unsuccessful, probably as a result of patient stratification. This lack of stratification can lead to under-informed drug administration and possibly lack of patient response, particularly when we factor in the high diversity incurred either from molecular perspective to macro-factors such as comorbidities. In order to address this problem, we have built the first iteration of a decision-support system, the Personalis software. The Personalis platform is an EHR patient stratification tool for physicians that aims to facilitate and optimize patient stratification and management.

We are currently implementing Personalis as a web application. This will greatly facilitate deployment and device agnostic use of the software. Another added advantage of using a web-based framework is its compatibility with an agile development workflow. By iterating with our collaborators we have tailored the application around the syndication of EHR data and generation of ‘on-the-fly’ pivot tables. The tools in Personalis greatly facilitate the process of testing a theory based on user stratification. They allow to progressively gather observations from patient data and continuously assess whether the stratification of the population still holds (population vs individual) and/or how does the patient compare to current strata (individual vs population).

The unification of data import and visualization is an important aspect of collaboration and knowledge exchange. We aim to guide this system towards automatic hypothesis testing for continuous data flow. This process is imperative in order to keep up with the flux of modern data streams. This first iteration of the Personalis platform proves that it is possible to automate such a system in order for the application to guide physicians towards meaningful strata, instead of assessing or trying to identify them by matching observations.