An exploration of therapeutic impact on human health.
Researchers, scientists, and professionals take on an important, yet challenging, task when developing therapeutics. However, the task is made even more difficult when the results from a study are in one large Excel table and the supporting information is scattered across various publications, knowledge bases, and spreadsheets.
Our challenge was to create a set of visualizations that communicated what happened in a study as well as what those results mean. The visualization had to be dynamic, to work with any study results data. For example, if a set of metabolites significantly changed, we wanted to call that out, but also relay how those changes impact the host. Additionally, we wanted to create an inviting experience that made scientific information feel approachable. The tool should allow researchers to answer the questions of “does this drug impact the microbiome?” and “if it does impact the microbiome, what do I do next? How do I use this information to structure trials in the future?”
We had two different datasets powering the tool, both focusing on a set of 170 metabolites. The first dataset contained study results and through using p-values and fold change, we were able to see which metabolites had increased/decreased activities. The second dataset was used to overlay additional context, including pathways, diseases, and other relevant biological information associated with each metabolite. Using these two datasets together allowed for the ability to answer questions such as “did the metabolites associated with cancer increase or decrease?” Seeing all cancer related metabolites decrease, could be a sign that a cancer drug was working.
We met with subject matter experts - from marketers who sell the Discovery Panel to scientists who interpret the results - to understand how they think about the Microbiome. We determined the two audience member types: those with clear hypotheses to prove and those who are gathering information in order to create future hypotheses.
Then we performed quantitative analysis in order to apply what we had learned from our interviews to real data. We generated our own insights and used the results to ask additional questions of scientists. If we found an interesting data point, we wanted to ensure that it was worth calling out in our visualization (and wasn’t common knowledge). We also used this time to categorize data points as primary versus secondary and think about scales and pre-attentive attributes we could use to convey insight. We wanted to stay within scientific norms, but also wanted to challenge them, where possible.