Mapping the Analytics Continuum in Life Sciences

by Christopher Waller, VP, Chief Scientist, EPAM United States

October 31, 2017

Before getting into what I call the Analytics Continuum in Life Sciences, let me introduce myself. I’m Christopher Waller, and in my role as a business consultant for the Life Sciences vertical at EPAM, I am tasked with bridging product strategy and roadmap development within specific engagements at large industry accounts to accelerate business transformation.

My areas of expertise include Life Sciences and Healthcare within pharmaceutical and biotechnology settings, and I draw heavily on my past experience as an Executive Director of a Scientific Modeling Platform and a Senior Director of Healthcare Informatics to provide clients the very best in data science consulting.

With the intro out of the way, let’s talk about employing data science in Life Sciences. There’s a lot of talk about how, when and why we should use analytics to make better decisions across every part of the business, including R&D. So, how can we best harness the technical skills of data scientists to make more breakthrough discoveries and increase operational efficiency and profitability?

The first step to improving your organization’s data science productivity could be simply recognizing the natural order of what I call the analytics continuum in Life Sciences. Just like any story has a beginning, middle and end, so do the many different functions of analytics. What follows is a simple hierarchical view into the world of analytics.

The Entry Point: Descriptive Analytics

Descriptive Analytics approaches provide hindsight into situations by answering questions such as “What happened?” and “How did it happen?” These are the most fundamental approaches and generally represent a logical starting point for data science practitioners learning analytics:

  • Standard Reports and Dashboards: Tools such as SpotFire, Tableau, etc. are routinely used here.
  • Ad hoc and Custom Reports: Again, SpotFire and Tableau are routinely used, but this time along with more advanced Business Intelligence reporting tools like Crystal Reports and bespoke applications.


Complexity Rising: Predictive Analytics

Predictive Analytics approaches provide insight into situations and involve the necessary process of hypothesis generation and testing. Questions such as “What is the problem?”, “Is there a pattern?”, “What is a good question to ask?”, and “When is action needed?” generally form the basis for:

  • Enquiry Analytics
  • Data Exploration and Mining (Analysis, Visualization, Query, Drill-down, Alerts)


Experts Only: Prescriptive Analytics

Prescriptive Analytics represents the most highly evolved and complex set of techniques in this continuum. Asking questions like “Is my hypothesis correct?” and “What is the cause?” lead experienced data scientists to employ:

  • Statistical and Mathematical Analysis: Standard packages from R and SAS along with more mundane tools such as Excel provide solutions for data scientists in this regard.

Finally, “What will happen if…?”, “What’s the best choice?”, “What are the alternatives?”, and “What should I do?” provide the intellectual foundations for:

  • Predictive Modeling, Simulation and Optimization: A wide variety of tools (open source, commercial and custom-developed) are used here. Common ones include R and SAS for generic model building, etc., but there are also more domain-specific tools such as NONMEM and NLME for simulation and optimization.


Applying the Analytics Continuum

Now that we’ve mapped the analytics continuum, we can talk about how it works in practice. In the next installment of this blog series, we’ll take a look specifically at the practice and current state of analytics in Life Sciences with a special emphasis on pharmaceutical research and development.