The visualization of probabilistic results from consumer genetic testing for ethnicity at AncestryDNA. R. E. Curtis, K. H. Freestone, M. J. Barber, J. M. Callaway, K. Noto, Y. Wang, C. A. Ball, K. G. Chahine AncestryDNA, Provo, US.
An important, but often overlooked, challenge in consumer genetics is the design of engaging and informative data visualization strategies that help consumers understand and fully appreciate the results of their genetic tests. Successful data visualizations that allow accurate and meaningful interpretation of consumer genetic test results must (1) help consumers overcome erroneous preconceptions or assumptions, (2) bridge gaps in the genetics education of consumers and (3) communicate the uncertainty associated with probabilistic predictions to consumers who may not have a strong understanding of statistics. Conveying uncertainty associated with predictions is crucial because many aspects of genetic science rely on probabilistic theory, including identity by descent (i.e., cousin matching), admixture predictions (i.e., ethnicity), and relative disease risk. If the stochastic nature of genetic algorithms is not properly conveyed in the visualization, na´ve users often will either whole-heartedly accept the results as ground truth or dismiss them altogether. At AncestryDNA, we have considered this problem in the context of admixture prediction based on autosomal SNP testing. Over the past year, we have delivered over 125,000 admixture predictions to consumers across the US and conducted multiple surveys to assess our customers satisfaction with and comprehension of those predictions. User surveys revealed several areas for improvement: some consumers had incorrect assumptions about the test; many consumers were confused about genetic concepts and most consumers were unable to understand the probabilistic nature of the predictions. In response to our consumer reactions, we have developed new visualization tools that attempt to address the types of shortcoming described above and to create consumer-friendly approaches to describe statistical concepts such as confidence intervals, sensitivity, positive predictive value, and standard deviations. As part of the natural life cycle of iterative consumer-facing improvements, we survey and evaluate our user interfaces using systematic questionnaires and focus groups to judge which tools most effectively convey the concept of genetics and probabilistic theory to a largely na´ve consumer population. We will present our findings based on the conviction that the principles we have applied in our iterative development, testing and refinement of user experiences can also extend to other aspects of consumer genetics.
You may contact the first author (during and after the meeting) at