Structural Genomics Approach to Disease Vector Control

Presented by: Yuan-Ping Pang

ABSTRACT NOT YET AVAILABLE

Dr. Pang received his B.S. degree in physical chemistry at Amoy University in China, his M.S. degree training in neuroscience and biochemistry at the Shanghai Institute of Physiology in China (with Li-Jun Chen), and his Ph.D. degree in synthetic chemistry at the University of Pittsburgh (with Alan P. Kozikowski). He then embarked on a one-year sabbatical study in computational chemistry at the University of California, San Francisco (with the late Peter A. Kollman). Since 1991, he has been working on the development and application of special-purpose computer hardware and software as well as new methods for just-in-time drug discovery and protein folding at the Mayo Clinic, and he is currently a professor of biophysics and pharmacology and the director of the Computer-Aided Molecular Design Laboratory of the Mayo Clinic. He has published 126 peer-reviewed original articles with an H-index of 32 and holds 14 active patents and 8 pending patents. His research is supported primarily by the Defense Advanced Research Projects Agency (DARPA), the U.S. Army Medical Research Material Command (USAMRMC), the Army Research Office (ARO), the National Institutes of Health (NIH), the U.S. Department of Agriculture (USDA), the High Performance Computing Modernization Office (HPCMO), the State of Minnesota, and the Mayo Foundation for Medical Education and Research. 

Computational Collaboratives: Connected Thinking to Correlate Genomic Variations with Cancer Risk and Outcomes

Presented by: Brian Van Ness

One of the important concepts that has emerged from the human genome project is that disease risk and clinical outcomes can be influenced by individual genetic background. Thus, while many diseases may have their unique genetic signatures, individual patient outcomes are dependent on heritable variations (SNPs) in a wide variety of genes and pathways affecting cellular functions and drug responses. Moreover, genetic variations in such global functions such as inflammation, immunity, and cellular signaling in the tumor microenvironment can have an impact on diverse clinical responses.   Our lab has developed a targeted SNP panel to determine genetic variations among cancer patients (eg. multiple myeloma, lung cancer) that focusses on genetic variations in genes within key cellular pathways. Because clinical trials often have limited sample size, we needed to develop computational approaches that go beyond the high error association analysis of univariate rank ordering. Indeed, most complex traits, including cancer risk and therapy outcomes, certainly result more from complex genetic interactions than single genetic drivers.   We addressed this by working with the Computational Science group (Vipin Kumar). Using some established methods in Support Vector Machines we developed significant predictive algorithms of survival. Further, novel methods in pathway intersections and "p-value jumps" provided clusters of SNPs that are associated with cancer risk and clusters associated with survival predictions far above random permutation testing.   The connected thinking in a project like this starts with clinical expertise in developing the clinical trials; which feeds the basic science to develop genetic approaches to assess variations in responses; followed by challenges met in analysis by computational expertise; and biologic insights developed by the team. Discussion will also include how this approach can incorporated in to research models we are developing for Healthcare IP2 - Informative, Personalized, and Predictive.

The research in the Van Ness lab is directed at defining genetic deregulation that contributes to lymphoid malignancies, particularly multiple myeloma. Multiple myeloma results from plasma cell expansion in the bone marrow, and unfortunately is very hard to treat. This difficulty comes in part from the variability in genetic and signaling pathways that are deregulated in the plasma cells as well as the cells in the bone marrow microenvironment. The lab is developing both cell lines and mouse models to explore how different genes can influence disease progression and therapeutic response. The lab is identifying some of the complexity of gene expression through microarray expression profiling; and we have undertaken a collaborative project to target gene deregulation that will contribute to models of plasma cell malignancy in the mouse. The Van Ness lab is also working with both national and international clinical groups to correlate genetic defects with disease outcome and response to different therapies. The ultimate goal is to contribute to genetic characterization of patients that will direct individualized therapy.

Identifying Chromosomal Features Affecting Targeting of Site Specific Events in S. Cerevisiae 

Presented by: Joshua Baller

The yeast Saccharomyces cerevisiae is host to a number of transposons, small sequences of DNA that replicate and move within the chromosomes of their host. Two such transposons Ty1 and Ty5 (Tys), are members of the retrotransposon class of transposons. A signature of retrotransposons is the presence of reverse-transcriptase and integrase genes in their coding sequence. Reverse-transcriptase generates a cDNA copy of the transposon transcript while integrase inserts the cDNA copy into the genome. Previous studies have shown that Ty5 integrase interacts with the heterochromatin protein Sir4, creating a preference for insertion into Sir4 covered regions of the genome. Likewise Ty1 is hypothesized to interact with a component of the PolIII transcription complex resulting in insertion near PolIII genes. In both cases the observed distribution is not uniform over the suspected distribution of the interacting chromatin. This suggests the existence of secondary factors influencing insertion site preference. To identify these factors we have applied machine learning approaches. In the case of Ty5 we applied log linear classifiers to identify telomeres, Y’s and the area surrounding ARS consensus sequences as transposition hotspots. Additionally, we identified nucleosomes and ORFs as areas of decreased transposition. To validate our classification accuracy we used ROC analysis under 5-fold cross-validation. For Ty1 we applied regression models to relate various features to insertion frequency at PolIII genes. This work has verified a slight correlation with PolIII machinery. The features used by the classifier to discriminate between the two sets provide candidates for further benchtop research. Future work will improve feature selection and regularization in order to reduce the number of features identified to only a few key features.

Discovering Breast Cancer Biomarkers in Saliva: Quantifying Low Abundance Proteins from Dynamic Range Compressed Samples

Presented by: Sue Van Riper

Breast cancer ranks second as a cause of cancer death in women.   Successfully managing and curing breast cancer requires early detection and diagnosis through palpitation, mammography and biomarkers. An ideal vessel for protein biomarkers is saliva; it is easily collected, non-invasive, abundant, and particularly important to patient groups in which it can be difficult to conduct other diagnostic tests.  Unfortunately, saliva’s wide dynamic range in protein abundance impedes shotgun proteomics’ ability to detect and quantify low-abundance-but-potentially-indicative proteins. We recently showed that hexapeptide bead dynamic range compression (DRC) integrated into our proteomics workflow increases sensitivity and allows detection of low abundance proteins in saliva. Building on this work, we next determined whether we could quantify abundance levels of proteins in DRC saliva in the context of breast cancer biomarker discovery. Pooled DRC saliva samples from healthy and metastatic breast cancer patients were trypsin digested, labeled with stable isotope tags (newly introduced mTRAQ heavy and light labels), orthogonally fractionated in three steps, and analyzed on an LTQ-Orbitrap mass spectrometer. As with many new technologies, introduction of new biochemical methods (mTRAQ) into our workflow necessitated corresponding changes in our computational proteomics software. We tackled this problem in true interdisciplinary fashion, accommodating mTRAQ by integrating relational database approaches, quantification software (ASAPRatio), and validation software (Mayu) into a new quantitative shotgun proteomics workflow. Using this novel workflow, our preliminary results show successful identification and quantification of mTRAQ labeled proteins in DRC saliva and reveal notable (> 2x) abundance changes in several proteins. Of these proteins, some are associated with existing serum/breast tissue oncogenes and biomarkers, and others are novel in that they are not known as diagnostic for breast cancer. While our work to date has been on late stage metastatic breast cancer saliva, these preliminary results suggest that it is possible to discover early stage salivary biomarkers and justify continued studies seeking to identify salivary biomarkers for early detection of breast cancer. 

A native Minnesotan, Susan Van Riper is a 3^rd year BICB PhD graduate student with a research focus in computational proteomics. After receiving her B.S. in Computer Science from Winona State University, she held several technical and management positions during a twenty year career in industry. She returned to academia, receiving a M.S. in Software Engineering from the University of Minnesota in 2007. During her matriculation, she became interested in medical diagnostics and subsequently became a full-time PhD student applying computational and database algorithms to biological problems. Working with Professors John Carlis and Tim Griffin, her current research focuses on novel relational algebra extensions for de novo identification and quantification algorithms in ultra high resolution mass spectrometry based proteomics in the context of discovering novel diagnostic biomarkers in saliva.