GET STARTED
1
Request Info
2
Visit
3
Apply

Dr. J. Sunil Rao to Begin National Science Foundation Project on Subject-level Prediction and Application

According to a study published in the Journal of the American Statistical Association, a myriad of practical problems can be linked to prediction, where the main interest is at the subject or sub-population level. Personalized medicine, for example, is a type of subject-level prediction, where doctors select treatments that can help patients based on the genetic understanding of their disease. Sub-population levels can focus on small communities, rather than a population-level based approach.

In 2015, J. Sunil Rao, Ph.D., interim chair and professor at the Department of Public Health Sciences, and colleagues developed a method called the classified mixed model prediction (CMMP). The goal of the CMMP was to create a match between a group or dataset, with known groups or clusters, and once a match was built, the traditional mixed model prediction method could be utilized to make accurate predictions.

“Any place where you could envision employing a precision medicine type prediction is where CMMP could help. Interesting examples include looking for novel drug targets based upon drug-gene mutation interactions, looking for interesting matches in an electronic health record that might provide clues about potential drug repositioning targets for patients who have failed standard therapies, trying to predict expensive genomic marker values based upon more commonly collected ones and also dealing with the fact that genomic databases are not racially or ethnically diverse,” Dr. Rao said.

While CMMP is the first work to layout the framework under which predictions using a precision medicine framework can work, the scenarios under which it currently operates, however, are limited and many real-life situations fall outside its scope.

Various fields, such as business, social sciences, and health sciences, also need more predictions of health outcomes for new patients or perhaps, for example, a prediction of a new school’s response to efforts in educating children about smoking prevention.

Considering these reasons, Dr. Rao will serve as principal investigator on a project that will make methodological advances to CMMP into other types of subject-level prediction problems, as well as to develop new inferential methods to add a measure of confidence in the predictions. Dr. Rao will work with Jiming Jiang, Ph.D., from the University of California at Davis, as well as with Thuan Nguyen, Ph.D., from Oregon Health and Science University.

“CMMP in its original guise was designed for slightly more constrained situations like knowing subpopulation membership apriori or not having a huge number of predictors to model. Many interesting problems in health that involve precision medicine will fall under these more complicated situations. In this project, we extended the theory to accommodate a wider range of precision medicine problems. This now includes the use of genomic information as well,” added Dr. Rao.

Dr. Rao and colleagues will learn how to deal with training data with unknown grouping, as well as with sparse, high dimensional covariates and provide accurate measures of uncertainty for CMMP-type prediction.

“The solutions will all involve developing new theory, such as estimation techniques, study their optimality properties, and develop new computational algorithms, run extensive simulations and then work with two subject matter collaborators in testing the methods out on real data,” Dr. Rao said.

One of the areas that will be investigated in this project is precision medicine and health disparities focusing on the prediction of epigenetic markers using high dimensional genotype profiles. The other will involve an area of family economics. They will use a large survey of data from China and focus on households as a primary interest.

“Both applications will leverage important collaborations with practitioners and thus increase the impact of the work,” Dr. Rao said.

Written by Amanda Torres
Published on October 2, 2019