Health Informatics Invited Speaker Series

submenu

The Invited Speaker Series is an extension of our Health Informatics Grand Rounds. In this series, top informaticians from around the country will share their current research with the University of Minnesota informatics community. This series is open to the public.

The series will generally be on Thursdays from 3:30-4:30, but dates and times may vary.

May 11, 2016 - Learning Latent Variable Causal Models

May 11, 2016 - Learning Latent Variable Causal Models

Erich Kummerfeld, MS, PhD
Postdoctoral Scholar, Department of Biomedical Informatics, University of Pittsburgh

Time: 12:00-1:00
Location: 2-520 Moos Tower

Abstract
Many scientific research programs aim to learn the causal structure of real world phenomena. Causal learning is made more difficult when important variables are latent, i.e. they are not directly observed. This talk presents and evaluates a new domain-general algorithm for discovering latent variables from observational data. Correctness conditions and potential applications of the algorithm are also discussed.

Bio
Erich Kummerfeld is a Postdoctoral Scholar in the Department of Biomedical Informatics at the University of Pittsburgh. He received his PhD in Logic, Computation and Methodology from Carnegie Mellon University's Department of Philosophy, where his dissertation developed new algorithms for discovering latent causal variables from observational data. The general aim of his research is to improve the procedures, methodologies, and inference tools used by scientists, with special attention to biomedical domains. His most recent work focuses on the development and evaluation of algorithms for discovering causal structures and latent measurement models.

Mar. 3, 2016 - Fine-Mapping Causal Variants in a Bayesian Framework Using Summary Statistics

Mar. 3, 2016 - Fine-Mapping Causal Variants in a Bayesian Framework Using Summary Statistics

Wenan Chen, PhD
Postdoctoral Fellow, Statistical Genetics and Genetic Epidemiology, Mayo Clinic

Time: 3:30-4:30
Location: 2-690 Moos Tower

Wenan ChenAbstract
Most current genome-wide association studies (GWAS) only report regions of association with traits or diseases, but the underlying genetic causal variants are still unclear. Methods of prioritizing plausible causal variants from associated regions are important for follow-up functional studies, which are both expensive and labor intensive. By showing that two fine- mapping methods are approximately equivalent to each other, we have developed a Bayesian fine-mapping method that requires only summary statistics, called CAVIARBF. Because functional annotations of DNA elements have been shown to be enriched in identified association regions from GWAS, and can be incorporated to improve the performance of fine mapping, we further propose a general Bayesian framework to provide a systematic way to incorporate functional annotations for fine-mapping causal variants. Simulation results show that the proposed method is the most accurate in identifying causal variants among different strategies and methods compared. I will also talk about our findings on applying our fine- mapping methods on two real data sets, a meta-analysis of GWAS data of high-density lipoprotein (HDL) cholesterol and gene expression quantitative trait loci (eQTL) study of normal prostate tissues.

Bio
Wenan Chen, PhD, is a postdoctoral fellow working with Dr. Daniel J. Schaid at Mayo Clinic in biostatistics with a focus in developing statistical and computational methods for genetic data analysis. Dr. Chen received his B.S. in Computer Information System from the Beijing Information Technology Institute in China and his PhD in Computer Science from Virginia Commonwealth University with a focus on biomedical image and signal processing, and the application and development of machine learning methods in medical outcome prediction.

Dr. Chen has worked on a broad range of problems in the field of statistical genetics, including genotype imputation in pedigree data, association test in admixed populations, and correcting for confounding factors in methylation data analysis. His current research interests center on developing methods for fine-mapping/prioritizing underlying causal genetic variants for functional follow-up. He has developed CAVIARBF, a fine-mapping tool used by researchers from different institutions, including the Wellcome Trust Sanger Institute. He is also interested in outcome prediction, such as disease risk prediction based on multiple layers of genetic, genomic, or epigenetic information using machine learning methods, and heritability estimation of different phenotypes. With his vast knowledge in computer science, statistics, and genetics, he is interested in tackling interdisciplinary problems arising in biology and medicine, disciplines in which respectable computational methods can make a big difference.

Download the recording.
Stream the recording.

Jan. 28, 2016 - Finding Orders in Heterogeneity: Studying X Chromosome Inactivation using Single Cell Transcriptome

Jan. 28, 2016 - Finding Orders in Heterogeneity: Studying X Chromosome Inactivation using Single Cell Transcriptome

Kyoung Jae Won, PhD
Research Assistant Professor of Genetics, Perelman School of Medicine, University of Pennsylvania

Time: 3:30-4:30
Location: 2-690 Moos Tower

Kyoung-Jae Won

Abstract
I studied X chromosome inactivation (XCI) during the human embryonic cell (hESC) differentiation using single-cell transcriptomic data. Clustering using single-cell transcriptomic data showed that the expression levels of the master regulator in ESCs were reduced drastically upon differentiation. Interestingly, the expression levels of XIST, a long intergenic non-coding RNA that controls XCI, were not correlated well with these master regulators. Classifying genes based on the XIST expression levels, we found a group of genes that are potentially regulated by XIST in ESCs. More importantly, I observed that many genes known to subject to XCI were not entirely inactivated in the inactivated X chromosome in a single-cell level. The analysis using single cell transcriptomic data expanded our knowledge about XCI.

As a second topic, I will introduce a computational approach to predict promoter-enhancer interactions using the transcription levels at enhancers (eRNA) as well as gene transcription levels. Applying this approach, I predicted the PPARγ target genes in adipocytes.

Bio
Kyoung-Jae Won is a Research Assistant Professor of Genetics at the University of Pennsylvania. He earned his BS and MS degree in Electronics from Chung-Ang University, Korea in 1996 and 1998, respectively. He received his PhD in Electronics at the University of Southampton, UK in 2005. During his PhD, he applied machine learning algorithms to learn the structure of hidden Markov models (HMMs) for biological sequence analysis. After spending one year at the Bioinformatics Center at the University of Copenhangen, Denmark, as a postdoctoral researcher, he moved to the University of California, San Diego (UCSD), where he developed computational algorithms to study gene regulation using epigenomic data. Since joining Penn in 2011, he has performed large-scale integrative analyses using various types of sequencing datasets.

Nov. 19, 2015 - Systematic Multi-scale Modeling and Analysis for Gene Regulation

Nov. 19, 2015 - Systematic Multi-scale Modeling and Analysis for Gene Regulation

Daifeng Wang, PhD
Associate Research Scientist, Molecular Biophysics and Biochemistry, Yale University

Time: 3:30-4:30
Location: W2-120 Weaver-Densford Hall

Abstract
The rapidly increasing quantity of Big Data offers novel and diverse resources to study biological and biomedical functions at the system level. Integrating and mining these various large-scale datasets is both a central priority and a great challenge for the field of informatics and necessitates the development of specialized computational approaches. In this talk, I will present several novel computational systems approaches in a multi-scale modeling framework to study gene expression and regulation with applications to cancer and developmental biology: 1) an algorithm to simultaneously cluster multi-layer networks such as gene co-expression networks across multiple species, which discovered novel human developmental genomic functions and behaviors; 2) a logic-circuit based method to identify the genome-wide cooperative logics among gene regulatory factors and pathways for the first time in cancers such as acute myeloid leukemia, which provided unprecedented insights into the gene regulatory logics in complex biological systems; 3) an integrated method using the state-space model and dimensionality reduction to identify principal temporal expression patterns driven by internal and external gene regulatory networks, which established an entirely new analytical platform to identify systematic and robust dynamic patterns from high dimensional, complex and noisy biomedical data. I also made these approaches available as general-purpose informatics tools. In addition, I will introduce some ongoing research projects and discuss the future directions where multi-scale approaches can make a significant impact in bioinformatics and biomedical informatics.

Bio
Daifeng Wang is an Associate Research Scientist at Yale University. He received his bachelor’s degree in Electronics and Information Engineering from Huazhong University of Science and Technology, Wuhan, China in 2004, and his Ph.D. in Electrical and Computer Engineering from the University of Texas at Austin in 2011. He joined Dr. Mark Gerstein’s Lab as a postdoctoral associate at Yale University in 2012. His research has focused on bioinformatics, genome informatics and biomedical informatics. He published his work in Nature, Genome Biology, PLOS Computational Biology, PNAS, IEEE/ACM Transactions on Computational Biology and Bioinformatics, PLOS ONE.

Nov. 12, 2015 - Integrative genomics towards precision medicine in relapsed ALL

Nov. 12, 2015 - Integrative genomics towards precision medicine in relapsed ALL

Jinhua Wang, PhD
Associate Professor of
Pediatrics, Faculty of the NYU Center for Health Informatics and Bioinformatics, and Associate Director of the Biomedical Informatics Core in the NYU Langone’s Laura and Isaac Perlmutter Cancer Center, NYU School of Medicine, New York University

Time: 3:30-4:30
Location: 2-520 Moos Tower

Abstract
The prognosis for pediatric acute lymphoblastic leukemia (ALL) has improved dramatically with cure rates approaching 90%. In spite of this progress, disease recurrence still presents a major challenge in about 10 to 20% of patients. Such patients have a dismal prognosis even with aggressive treatment strategies making relapsed ALL a common cause of cancer death in children. To better understand the underlying biological pathways that operate in vivo that lead to drug resistance and treatment failure, we developed an integrated genomic approaches using gene expression, copy number, methylation and next generation sequencing to compare diagnosis/relapse paired samples from patients enrolled on clinical trials. We have discovered a number of important genetic lesions with significant translational relevance that are associated with chemotherapy resistance.

Bio
Dr. Wang is an Associate Professor at the Department of Pediatrics, a faculty member of the NYU Center for Health Informatics and Bioinformatics, and the Associate Director of the Biomedical Informatics Core in the NYU Langone’s Laura and Isaac Perlmutter Cancer Center. Dr. Wang joined NYU Medical Center in 2007 as a junior faculty after a research scientist appointment in St Jude Children’s Research Hospital. Prior to that, Dr. Wang was a research postdoc fellow in Cold Spring Harbor Laboratory beginning in 2001. Dr. Wang got a BS in physics from Beijing University in 1996 and did his PhD training in computational biology and genomics research in the Institute of Biophysics at the Chinese Academy of Sciences in 2001. Dr. Wang also got an MBA from NYU Stern School of Business in 2011.

Nov. 5, 2015 - Bioinformatics in practice: mining “goldomics” for basic and translational science

Nov. 5, 2015 - Bioinformatics in practice: mining “goldomics” for basic and translational science

Steven S. Shen, MD, PhD
Associate Professor of Biochemistry and Molecular Pharmacology, Director for Genomics Bioinformatics of Genome Technology Center, NYU School of Medicine, New York University

Time: 3:30-4:30
Location: 2-520 Moos Tower

Dr. Steven Shen

Abstract
For many years, Dr. Shen and his research team have been working on developing computational approaches to mine and interpret genomic data for basic and translational research. This talk will focus on Dr. Shen’s published and ongoing studies. He will use some of his latest work to demonstrate how the genomics and bioinformatics are capable to be great tools for bench scientists to solve tough problems.

Bio
Dr. Shen is a director for genomics bioinformatics at NYUMC Genome Technology Center, an associate professor at the Department of Biochemistry and Molecular Pharmacology of NYUMC and a faculty member at the Center for Health Informatics and Bioinformatics. Before joining NYUMC, Dr. Shen was a research assistant professor at Boston University School of Medicine. He also spent seven years as a research scientist at Massachusetts Institute of Technology developing high throughput technology and computational methods and two years as a senior scientist at Helicos Biosciences developing single molecule sequencing technology. He graduated from China and did his postdoctoral training in Weill Medical College of Cornell University.

Sept. 17, 2015 - Navigating Big Data in Healthcare and Population Health

Sept. 17, 2015 - Navigating Big Data in Healthcare and Population Health

Adam Wilcox, PhD
Director, Medical Informatics, Intermountain Healthcare, Salt Lake City, Utah

Time: 3:30-4:30
Location: 2-530 Moos Tower
This Grand Rounds is co-sponsored by the School of Public Health

Adam Wilcox, PhD

Abstract
Over the past decade, there has been a significant change in the availability of health information along with increased urgency to use that data to improve the patient experience and population health while reducing healthcare costs. However, the effective application of that data towards this triple aim of health care remains elusive. Dr. Wilcox will discuss the challenges with using health information effectively, and present practical approaches that can improve the effectiveness and efficiency of healthcare while leveraging data and population health initiatives.

Bio
Adam Wilcox, PhD, is the Director of Medical Informatics at Intermountain Healthcare. He has spent over 15 years in clinical informatics and clinical research informatics, much of that in supporting comparative effectiveness and patient-centered outcomes research. At Intermountain, he leads efforts in applying health information technology to quality improvement processes, supports health IT applications to primary care, and leads Intermountain’s clinical decision support efforts. He also directs Intermountain’s analytic health repository, where he leads the development of a more research- accessible database extracted from electronic health records. Prior to his return to Intermountain, he was an associate professor in the Department of Biomedical Informatics at Columbia University, where he was the initial principal investigator for the Washington Heights/Inwood Informatics Infrastructure for Comparative Effectiveness Research (WICER) project. He also directed the clinical data warehouse, the clinical data repository and legacy electronic health record, a local health information exchange, and the informatics support for the Herbert Irving Comprehensive Cancer Center. He has broad experience in both applied and research informatics, and was the creator and director of the Research Methods in Informatics course at Columbia University. In 2015, Dr. Wilcox was appointed a member of the PCORI Methodology Committee. He is an elected fellow of the American College of Medical Informatics, is a senior editor for eGEMs, and serves on the Clinical Informatics Subcommittee for the American Board of Preventive Medicine, which administers the board examination for the clinical informatics subspecialty. He has authored over 100 book chapters, peer-reviewed articles and abstracts, and has presented at conferences and institutions across the country.

Download the Grand Rounds recording.
Stream the Grand Rounds recording.

Sept. 3, 2015 - Predictive and Causal Modeling in Health Sciences

Sept. 3, 2015 - Predictive and Causal Modeling in Health Sciences

Sisi Ma, MSc, MSc, PhD,
Research Scientist, Langone Medical Center, New York University

Time: 12:00-1:00
Location: 2-580 Moos Tower

Dr. Sisi Ma

With the rapid accumulation of high variety of large volume data in health science, there is an increasing demand for computational methods that can perform effective and systematic knowledge discovery. In this seminar, two broad categories of computational tools for knowledge discovery and their application in the domain of health science will be discussed. The first category is predictive modeling, which aims to discover predictive (prognostic, forecasting) and diagnostic (pattern recognition) knowledge from data. The predictive models of a particular public health event or a particular disease can lead to screening tests, diagnostic and prognostic assays. The second category is causal modeling, which aims to discover the underlying mechanisms driving the behavior of a system. The causal models of a particular disease can help experimental scientists and clinicians to identify disease treatment targets or health system interventions. The general analytical frameworks and the operating principles of the predictive modeling and causal modeling will be introduced, along with case study examples of their applications in several health science discovery settings.

PDF icon Dr. Ma's Slides.