Robert F. Murphy

Center for Bioimage Informatics, Departments of Biological Sciences,
Biomedical Engineering and Machine Learning, and Molecular Biosensor and Imaging Center,
Carnegie Mellon University, Pittsburgh, Pennsylvania

Biography:

Robert F. Murphy earned an A. B. in Biochemistry from Columbia College in 1974 and a Ph.D. in Biochemistry from the California Institute of Technology in 1980. He was a Damon Runyon-Walter Winchell Cancer Foundation postdoctoral fellow with Dr. Charles R. Cantor at Columbia University from 1979 through 1983, after which he became an Assistant Professor of Biological Sciences at Carnegie Mellon University in Pittsburgh, Pennsylvania. He received a Presidential Young Investigator Award from the National Science Foundation shortly after joining the faculty at Carnegie Mellon in 1983, was named a Fellow of the American Institute for Medical and Biological Engineering and a Senior Member of IEEE in 2007. He has received research grants from the National Institutes of Health, the National Science Foundation, the American Cancer Society, the American Heart Association, the Arthritis Foundation, and the Rockefeller Brothers Fund. He has co-edited two books and two special journal issues on “Cell and Molecular Imaging,” and published over 130 research papers. He is currently Professor of Biological Sciences, Biomedical Engineering, and Machine Learning, and Director (with Jelena Kovacevic) of the Center for Biomedical Image Informatics at Carnegie Mellon. He also directs (with Ivet Bahar) the joint CMU-Pitt Ph.D. Program in Computational Biology. The focus of his career has been on combining fluorescence-based cell measurement methods with quantitative and computational methods. His group at Carnegie Mellon did extensive work on the application of flow cytometry to analyze endocytic membrane traffic beginning in the early 1980’s and pioneered the application of machine learning methods to high-resolution fluorescence microscope images depicting subcellular location patterns in the mid 1990’s. This work led to the development of the first systems for automatically recognizing all major organelle patterns in 2D and 3D images. His leadership experience includes developing the first formal undergraduate program in computational biology in 1987 and founding the Merck Computational Biology and Chemistry program at Carnegie Mellon in 1999. These program were important forerunners to the establishment in 2005 of a Ph.D. program in computational biology that is run jointly with the University of Pittsburgh. This program was recently chosen as one of only ten awardees by the new HHMI-NIBIB Interfaces Initiative (Dr. Murphy is principal investigator on that grant). He has played a significant role in outreach for undergraduates, directing a summer research program for students underrepresented in science and engineering since 1996. With B.S. Manjunath at the University of California, Santa Barbara he is leading a major effort to create a new multi-institution, NSF-funded $9.4 million program in bioimage informatics that is bridging biology, cytometry, engineering and computer science. His group is responsible for providing image informatics tools for the NIH-funded Technology Center for Networks and Pathways (led by Alan Waggoner at Carnegie Mellon University and Simon Watkins at the University of Pittsburgh) and for providing structured, image-based information on subcellular location for the National Center for Integrative Biomedical Informatics (led by Brian Athey at the University of Michigan). He has also played leadership roles in recreation and athletics organizations in his community

Talk Title: Automated Modeling of Subcellular Patterns for Systems Biology

Abstract: The goal of systems biology is to build accurate models of entire eukaryotic organisms. Simply identifying which of tens of thousands of proteins are expressed in each of over a hundred animal cell types is a daunting task. To meaningfully understand the behavior of even a single cell type, we need information on not only which proteins are expressed but within what subcellular structures they are found. Information on subcellular location is usually obtained by visual examination of microscope images, but the scale of the problem and the need for objective and sensitive determinations argues against the visual approach. Our group has pioneered the application of machine learning methods to this problem, and demonstrated that automated methods can discriminate subcellular patterns that cannot be distinguished by visual examination. These methods are based on sets of features that capture the essence of patterns without being overly sensitive to cell size, shape and orientation. We have also demonstrated that cluster analysis can be used to group highly similar patterns without human assistance, and that proteins that humans lump together (such as different subpatterns of cytoplasmic proteins) are properly separated. Once such “protein subcellular location families” are built we need methods to communicate information on the pattern observed within each. Each pattern can be highly variable and may be represented by images of hundreds to thousands of cells. We have proposed building of generative models as a solution to this problem. We have obtained encouraging results on using a series of conditional shape and distribution models to summarize subcellular patterns and have developed an XML specification for exchanging them. The models allow images (distributions) to be generated that reflect the statistical variation in pattern within a particular location family and can be expected to be useful in systems biology simulations of cell behavior.