What gets me up in the morning is wanting to cure cancer and positively impact lives through the use of data and analytics. Our capabilities to try and test formulations exceed our ability to effectively analyze the results. Our use of real world evidence across populations in the 100s of millions allow us to potentially detect conditions that we have cured but not yet proven the cure.
For me this is essentially a classification problem. Part of the classification problem involves micro-populations. Another part of the classification problem involves meta-features.
Micro-populations are small cohorts differentiated by a bio-marker or a drug outcome. One micro-population taking a given drug might have very positive outcomes.
A different micro-population might have no positive outcomes and tend to have adverse outcomes. It is understanding the differences between these micro-populations that allows us to more fully understand a disease and how to cure.
As important as micro-populations is the notion of meta-features. In the same way as micro-populations allow precisions to be increased, meta-features allow the treatments, and drugs have a wide range of precisions. In some cases, a more generalized classifier of the treatment might be used. In other cases, such as a drug cocktail a more specific classifier needs to be used.