When to use Computer-aided drug discovery, Data analytics, and Informatics

What do these terms mean? We have seen an explosion in the types and volumes of available data in the biotech and pharmaceutical industries. This data is generated from many sources, including R&D efforts, patients, caregivers. Imagine a future where predictive modeling of biological processes and drugs becomes so sophisticated that new potential candidate molecules have a high probability of being successfully developed into drugs that act on biological targets safely and effectively. This is the dream, and yet getting there is made difficult by the barriers between the “big data” being generated and the “predictive analytics” on the other side. I’ve used these two terms purposefully because this is a well-known problem in the data science field: unstructured “big data” is just not very useful.

The key is to capture data electronically. This way it can flow easily between functions and it’s format is consistent, which is essential for powering real-time and predictive analytics.

“Data are not taken for museum purposes” - W. Edwards Denning

Very often scientific research data is generated to answer a one-off question, without any thought given to how it will be used. Despite how terrible this sounds to an informaticist, there are many one-time-only questions in research where it doesn’t make sense to invest much effort in data capture! However this should not be used as an excuse to never do it. Once a scientist gets to a point where they are asking a question daily or weekly they should consider automation to save everyone time. “Sharpen your saw”. Try to define the scope of the question (what variables you might tweak) and decide if your analytics setup must support daily or weekly decision-making. The more detail they can give you, the more likely you will get an interface that suits your needs.