Computing That Serves

Uncertainty Quantification for Biomedical Decision Making: A Critical Informatics Contribution to Precision Medicine


Thursday, September 8, 2016 - 11:00am


Julio Facelli


Christophe Giraud-Carrier

Julio C. Facelli
Thursday, September 8, 2016
11:00am 1170 TMCB

There is concern about the lack of reproducibility of biomedical studies, but the research community has not taken advantage of formal Uncertainty Quantification (UQ) methods to better understand this issue. Here we show the importance of UQ in Translational Science and Precision Medicine using prediction modeling in cancer. This presentation describes the use of UQ Monte Carlo (MC) methods in biomedical research with applications to: uncertainty in breast cancer classification and family history risk assessment.
Breast cancer subtyping uses gene expression profiles to identify subtypes defining treatment protocols. Our work shows that there is a substantial uncertainty in this classification. For instance, the classification of the 100,000 MC simulated replicas of a HER2-enriched sample, from the GEICAN study, show that as many as 38.7% are misclassified. Cancer prevention guidelines use estimated lifetime risk to provide advanced screening. Often risk is calculated using family history tools such as, BRCAPRO. Using an AHRQ report on self-reported family history inaccuracies and MC simulations we found that up to 47.8% high-risk pedigrees can be classified as low-risk and 6% low-risk pedigrees can be classified as high-risk.
The results presented here show that UQ methods can be applied to biomedical sciences. UQ provides useful clinical and translational information, and arguable UQ should become a common tool in translational science and precision medicine because UQ methods can provide a better understanding of the underlying factors leading to the lack of reproducibility.
This work has been supported by the University of Utah Program in Personalized Health, the Utah Center for Clinical and Translational Science funded by NCATS award 1ULTR001067 and the NLM Training grant T15-LM007124-18. Computer resources were provided by the University of Utah Center for High Performance Computing



Dr. Facelli was born in Buenos Aires, Argentina and attended the University of Buenos Aires where he got his Ph.D in physics in 1982. In 1993 he did post-doctoral research at the University of Arizona and the following year he joined the University of Utah. At the University of Utah he was the Director of the Center for High Performance Computing from 1995 to 2013 and he is currently, Professor and Vice Chair of the Department of Biomedical Informatics, Associate Director for Biomedical Informatics, Center for Clinical and Translational Science, Adjunct Professor of Chemistry and Physics and member of the Institute for Clean and Safe Energy and the Utah Nano Science Institutes. Dr. Facelli was elected fellow of the American College of Medical Informatics (ACMI) in 2014. Dr. Facelli has been involved in numerous computer and network related research projects and in many University and national committees dealing with Information Technology. He has extensive expertise in computational sciences, parallel and distributed computing and advance network applications. Dr. Facelli is co-author of more than 200 international per review publications and his research has been funded by NSF, NIH and DOE. Dr. Facelli served as Chair of the Coalition for Scientific Computing (CASC) during 2003 and 2004. He is referee for numerous international publications and funding agencies and has participated in advisory panels at NSF and NIH. He has taught classes in Physics, Chemistry, Computational Sciences, Telecommunications and Medicine. His current research interest are in parallel and distributed computing applications in biomedical informatics.