Postdoctoral Fellow, Biomedical Informatics and Medical Education
UW Department of Biomedical Informatics and Medical Education
Sean Mooney, Biomedical Informatics and Medical Education
Kari Stephens, Psychiatry and Behavioral Sciences
Ph.D. Candidate, Informatics, Indiana University
M.S., Bioinformatics, Indiana University, 2010
B.E., Biotechnology, People’s Education Society Institute of Technology, India, 2008
Vikas Rao Pejaver is interested in the development and application of machine learning methods to medical genomics and proteomics problems, with a particular emphasis on the integration of heterogeneous data from disparate sources. At Indiana University, he developed methods to predict over 50 different types of structurally and functionally important sites in proteins and integrated them into the prediction of disease-causing mutations.
A key challenge here was the identification of a common set of features for repeated use in multiple related prediction tasks, based on domain knowledge. He is particularly interested in automating this process to enable the discovery of previously unknown predictive features and the generation of provocative hypotheses to explain the resulting predictions.
At the University of Washington, his research encapsulates all of these aspects through the development of methods to learn patterns in population-scale electronic medical record (EMR) data sets. Specifically, he plans to use multimodal deep learning approaches for the extraction of the different types of previously untapped information present in EMRs (e.g. unstructured text, image data, among others) and their conversion into unified representations for a variety of prediction tasks related to decision-making in the clinic and basic research on the factors that influence health and disease (e.g. readmission risk prediction, many-to-many gene-disease relationship prediction).
His broader research interests also include other topics in bioinformatics, computational biology and machine learning, such as genome interpretation, post-translational modifications of proteins, intrinsically disordered regions, ensemble learning and model evaluation methods, among others.