|
Events CCMB Seminar Series 2007-2008 _______________________________________________________ Events
Wednesday, October 31st, 2007
Abstract:
Wednesday, September 26th, 2007
Abstract:
Abstract: The DNA microarray technology has been widely used in measuring gene expression levels in biomedical research. The potential ability of monitoring tens of thousands of genes simultaneously makes the microarray approach an efficient tool. However, the raw measurements are noisy. The data is a combination of specific binding and background noise. A considerable proportion of genes appear constantly expressed because the variation in signal is small and seemingly within the variation of background noise alone. The dynamic range of the gene expression level is thus limited to highly expressed genes. This is a crucial limit because often more than half of the genes in an experiment do not reach the level that differential expression can be reliably detected. Our previous work has shown that probes have sequence specific background levels and probe specific background adjustment can improve the dynamic range. Since large number of observations on the same probe is usually not available in a single study, information is borrowed across probes within each sample to estimate the probe specific background. The accumulation of microarray data in public data depository has enabled us to approach the problem from a different angle. With a large number of samples, we now observe data on the same probe across a large variety of experimental conditions. We start with a database of human gene expression arrays of 500 samples. Based on the expression profile on each gene across experiments, we evaluate each probe's tendency to non-specific binding and ability to track target expression variation. We demonstrate further extension of the dynamic range for gene expression measures. Wednesday,
January 23rd, 2008
Abstract: Peptide identification by tandem mass spectrometry is the dominant proteomics workflow for protein characterization in complex samples. Coupled with well established high-throughput proteomics workflows, tandem mass spectrometry search engines make identifying the major constituent proteins in clinical samples straightforward. Driven by increasingly sensitive protein chemistry protocols and mass spectrometers, and a new perspective on the importance of alternative splicing and coding SNP protein isoforms, however, the shortcomings of the existing tools are becoming more and more apparent. We use a variety of computational techniques to improve the reliability of peptide identification analyses, as we seek to address the limitations of current tandem mass spectrometry search tools. First, we aggressively enumerate an inclusive set of potential peptide sequences from transcript evidence, particularly ESTs, to ensure that evidence of novel, unexpected, or unannotated protein isoforms is not missed by tandem mass spectrometry search engines. We use a novel compression technique to ensure that the resulting sequence database can be searched quickly and easily using existing tools, and demonstrate that novel peptides, representing coding SNPs, alternative splicing, and novel mutations, can be observed in publicly available datasets. Second, we apply hidden Markov models to spectral matching of tandem mass spectra of previously identified peptides, improving on the sensitivity and specificity of peptide identification by sequence database search engines and traditional spectral matching techniques. Lastly, we post-process search engine peptide identification results using an unsupervised, model-free, result-combining machine-learning approach that achieves superior sensitivity and specificity than either result combining or machine learning alone. Using this technique on datasets derived from standard protein mixtures, we demonstrate that the performance of the commercial search engine Mascot can be bested by combining the results of two open-source search engines, X!Tandem and OMSSA; but that using all three search engines is better still. Such a reliable, sensitive, and specific peptide identification analysis platform has the potential to not only explore a largely untapped source of potential cancer biomarkers from clinical cancer samples and cancer cell-lines, but also to inform functional genomics and genome annotation. The characterization of expressed proteins using tandem mass-spectrometry provides direct evidence for the amino-acid sequence of functional proteins and their isoforms, evidence which is not available using other high-throughput experimental techniques. We conclude with a discussion of unconventional experimental workflows for peptide identification and their potential to inform functional genome annotation. Wednesday,
March 5th, 2008
Abstract: Dr. Collins will present on the use of array comparative genomic hybridization (aCGH) and end sequence profiling (ESP) for the development of novel biomarkers and therapeutic targets for cancer. Prostate cancer is amongst the prevalent cancers in the Western world and is increasing in incidence. PSA screening for prostate cancer has resulted in stage migration so that increasingly tumors are detected at an earlier stage. Nonetheless the percentage of men diagnosed with tumors at an intermediate risk of progressing to metastasis has remained relatively constant at ~ 25% and there is little change in the outcome statistics for this group. Therefore, it is critical that biomarkers be developed that can identify men in this group who can safely delay or avoid definitive therapy, alleviating the problem of over treatment. Dr. Collins laboratory has employed unique tumor cohorts and aCGH to identify genome based biomarkers that may be capable of dichotomizing this group of patients. To refine and advance this assay to the clinic whole genome amplification methods are being developed to enable analysis of tumor biopsies on Agilent oligonucleotide arrays. Dr. Collins will review progress and discuss the challenges of this type of research. A problem with array-based analyses is that it is very difficult for them to detect tumor specific biomarkers and drug targets such as the BCR-ABL fusion in CML. In addition, they are blind to heterogeneity. To overcome these limitations Drs. Collins and Volik invented ESP with the explicit goals of determining the structural organization of tumor genomes and transcriptomes. Progress made in the application of ESP to breast cancer cell lines and multiple tumor types will be presented. Specifically, data will be presented showing the structural organization of tumor genomes, the frequency and spectrum of mutations, molecular heterogeneity, validation of genome breakpoints, and detection of tumor-specific fusion genes and transcripts. In addition, ideas will be explored for a genome project based on ESP, and advancement of ESP to the clinic using next generation sequencing technologies. Wednesday,
February 13th, 2008
Abstract: I'll
talk about three areas of evolutionary biology
using a combination of statistics and discrete
math: viral population diversity, the evolution
of drug resistance, and phylogenetics. Next, I will describe a new class of graphical models inspired by poset theory that describe the accumulation of (genetic) events with constraints on the order of occurrence. Applications of these models include calculating the risk of drug resistance in HIV and understanding cancer progression. Finally, I'll describe a polyhedral
method for determining the sensitivity of phylogenetic
algorithms to changes in the parameters. We will
analyze several datasets where small changes in
parameters lead to completely different trees and
see how discrete geometry can be used to average
out the uncertainty in parameter choice.
Monday, March
3rd, 2008
Abstract: Tiling DNA microarrays extend current microarray technology by probing the non-repeat portion of a genome at regular intervals in an unbiased fashion. A fundamental problem in the analysis of these data is the detection of genomic regions that are differently transcribed across multiple conditions. We propose a linear time algorithm based on segmentation techniques and linear modeling that can work at a user-selected false discovery rate. It also attains a four-fold sensitivity gain over the only competing algorithm when applied to a whole genome transcription data set spanning the embryonic development of Drosophila melanogaster. Monday, March
10, 2008 Hosted by: Sorin Istrail To receive CCMB seminar announcements by email, sign up for the computational biology mailing list by sending email to listserv@listserv.brown.edu with the message body "subscribe computational-biology".
Abstract: Epistasis means that the functional consequence of mutations varies with genetic background, and the evolutionary consequences of epistasis are profound in asexual populations because a novel mutation’s fate is then determined by its fitness effect on only one of many alternative genetic backgrounds. This possibility motivates interest in Sewall Wright’s adaptive landscape, the projection of genotypic fitness values over a discrete, multidimensional nucleotide sequence space. In this framework, populations follow a temporal succession of individual points through this space determined by the interplay between mutational pressure, the local selective gradient defined by the landscape, and stochastic loss of novel genotypes. Several recent empirical characterizations of small regions of this landscape have demonstrated that in asexual populations functional epistasis i) sharply limits the number of mutational trajectories to high-fitness genotypes that are selectively accessible and ii) gives rise to a very sharp non-uniform probability distribution among selectively accessible trajectories. To date however, the absence of an analogous formal framework in which to characterize selectively accessible recombinational trajectories has limited understanding of this problem in sexual populations. I will describe a novel definition of the adaptive landscape appropriate for this problem: the vector field reflecting the joint pressures of mutation, selection and recombination over a continuous multidimensional space that represents both allele frequencies and linkage disequilibrium among alleles. Populations are again regarded as occupying a temporal succession of points in the underlying space, and I will illustrate the potential of this approach by describing how recombination influences a population’s evolutionary trajectory single- and multi-peaked fitness models. Wednesday,
March 19th, 2008
Abstract: The question how genetic variation and personal health are linked is one of the compelling puzzles facing scientists today. The ultimate goal is to exploit human variability to find genetic causes for multi-factorial diseases such as cancer and coronary heart disease. Recent technology improvement enables the typing of millions of single nucleotide polymorphisms (SNPs) for a large number of individuals. Consequently, there is a great need for efficient and accurate computational tools for rigorous and powerful analysis of these data. In my talk I am going to concentrate on two computational problems, which are an essential step in studying the data obtained by this technology: Accurate and efficient significance testing with a correction for population stratification and estimating local ancestries in admixed populations. Wednesday,
March 31, 2008
Abstract: Because binding of RNA polymerase (RNAP) to misplaced sites could compromise the efficiency of transcription, natural selection for the optimization of gene expression should regulate the distribution of DNA motifs capable of RNAP-binding across the genome. We have analyzed the distribution of –10 promoter motifs recognized by the s70 subunit of RNAP in 42 bacterial genomes. We show that selection on these motifs operates across the genome, maintaining an over-representation of –10 motifs in regulatory sequences while eliminating them from the nonfunctional and, in most cases, from the protein coding regions. In some genomes, however, –10 sites are over-represented in the coding sequences; these sites could induce pauses effecting regulatory roles throughout the length of a transcriptional unit. For nonfunctional sequences, the extent of motif under-representation varies across genomes in a manner that broadly correlates with the number of tRNA genes, a good indicator of translational speed and growth rate. This suggests that minimizing the time invested in gene transcription is an important selective pressure against spurious binding. However, selection against spurious binding is also detectable in the reduced genomes of host-restricted bacteria that grow at slow rates, indicating that components of efficiency other than speed may also be important. Minimizing the number of RNAP molecules per cell required for transcription, and the corresponding energetic expense, may be most relevant in slow growers. These results indicate that genome-level properties affecting the efficiency of transcription and translation can respond in an integrated manner to optimize gene expression. The detection of selection against promoter motifs in nonfunctional regions also indicates that no sequence may evolve free of selective constraints, at least in the relatively small and unstructured genomes of bacteria. Wednesday,
April 2, 2008
Abstract: Standard amino acid substitution matrices are constructed as log-odds ratios from large collections of alignments of related proteins. Any such collection has an implicit "standard" set of amino acid background frequencies. The matrices produced, however, often are used to compare proteins with quite non-standard amino acid compositions. We argue on theoretical grounds that this is inappropriate, and have described a method for transforming a standard matrix into one appropriate for comparing proteins with any non-standard compositions. Compositionally-adjusted matrices yield improved results from the twin perspectives of alignment score and alignment quality when proteins with strongly biased compositions are compared. To what extent are such adjusted matrices of utility for general purpose protein database searches? Using standard test platforms, we compared a standard matrix to compositionally-adjusted matrices, with relative entropy left unconstrained, or constrained in various ways. We found that constraining the relative entropy of the compositionally adjusted matrix to a fixed value in the new compositional context generally produced the best results. We also found that if the sequences compared are not known to have strong compositional biases, then it is still on average advantageous to use an adjusted matrix when the sequences satisfy certain simple length or compositional inequalities. Applying these findings to general-purpose database searches can lead to a significant improvement in retrieval performance, with a minimal increase in execution time. Wednesday,
April 9th, 2008
Abstract: Apart
from a relatively small number of variants the
genomes of two individuals are identical. These
small differences explain much the human diversity. The
most common form of variation between individuals
are single nucleotide polymorphisms (SNPs), representing
a single letter change in an otherwise conserved
sequence. In the past couple of years, tremendous
progress has been made in identifying the genetic
causes of a number of disease and phenotypic traits
using arrays, such as those made by Illumina, that
simultaneously assay a large number of SNPs. Wednesday,
April 16th, 2008
Abstract: Pattern discovery is a ubiquitous problem in many disciplines. It is especially prominent in recent years due to our greatly improved data-generation capabilities in science and technologies. The method I present here is motivated by the "motif-finding" and "module-finding" problems in biology, i.e., to find sequence patterns (i.e., "words") that seem to appear more frequent than usual in a given set of text sequences (i.e., sentences) and to find which of these "words" tend to co-occur in a sentence. A challenge in the motif-finding problem is that there are no spacings and punctuations between the words and the dictionary of "words" is unknown to us. Existing methods are mostly "bottom-up" approaches, i.e., to build up the dictionary starting with single-letter words and then concatenate some existing words that appear to occur next to each other in sentences more frequently than chance. Our new approach is a top-down strategy, which uses a tree structure to represent the relationship among all possible existing words and uses the EM algorithm to estimate the usage frequency of each word. It automatically trims down most of the incorrect "words" by letting their usage frequencies converge to zero. The module-finding problem is closely related to the well-known "market basket" problem, in which one attempts to mine association rules among the items in a supermarket based on customers' transaction records. It is also related to the two-way clustering problem. In this problem, we assume that the words are given, and our goal is to find subsets of words that tend to co-occur in a sentence. We call the set of co-occurring words (not necessarily orderly) a "theme" or a "module". We can generalize the dictionary model to the "theme"-model and use a similar EM-strategy to infer these themes. I will demonstrate its applications in a few examples including an analysis of chinese medicine prescriptions and an analysis of a chinese novel. This is based on a joint work with Ke Deng and Zhi Geng.Wednesday,
April 23rd 2008
Joe W. Gray, Ph.D., is Associate Laboratory Director for Life and Environmental Sciences and Life Sciences Division Director at the Lawrence Berkeley National Laboratory (LBNL). He is also Adjunct Professor of Laboratory Medicine at the University of California, San Francisco (UCSF) and program leader for the Breast Oncology Program in the UCSF Comprehensive Cancer Center. Dr. Gray's current research program focuses on in molecular analysis technology, identification of genomic aberrations that contribute to cancer pathophysiology, development of efficient strategies for enhanced marker guided cancer therapy –especially for breast and ovarian cancer and early breast cancer detection. His work is described in more than 330 publications and 50 patents. Major awards include the Radiation Research Society Research Award (1985), the E.O. Lawrence Award from the US Department of Energy (1986), Election as a Fellow of the American Association for the Advancement of Science (1996), the Curt Stern Award from the American Society for Human Genetics (2001), an Honorary Doctorate from the University of Tampere, Tampere, Finland (2005) a DOD Innovator Award (2007) and the Brinker Award (2007). Dr. Gray earned a Ph.D. in physics from Kansas State University. Wednesday,
April 30th, 2008
Abstract: One of the main challenges of systems biology is to cope with the overwhelming amount and diversity of omics data and a key problem is that of 'candidate gene prioritization' - i.e., selecting among a large list of candidate genes those that are most promising for further biological validation. We present ENDEAVOUR, a generic computational strategy to prioritize candidate genes based on their similarity (across multiple types of data, including sequence, expression, literature, annotation, etc.) to a set of genes already implicated in the process under scrutiny. We first validate the overall performance through a statistical cross-validation of 29 diseases and 3 biological pathways. Next, we validate a novel candidate for DiGeorge syndrome in a zebrafish model. Finally, we present an alternative machine learning strategy for gene prioritization using kernel methods. The key advantage of kernel methods in this context is that they provide an elegant framework for the fusion of data - by relying only on positive semi-definite kernel similarity matrices for the representation of heterogeneous data sources. Kernel-based novelty detection outperforms our previous method on our disease gene benchmark. Bio: My research interest falls in the broad field of bioinformatics and, more specifically into what I call Computational Systems Biomedicine, which is the application of computational methods in Systems Biology towards the understanding and modulation of developmental and pathological processes relevant to human health. The area of application in which I am currently most active is diagnosis and gene discovery in congenital disorders. Tuesday, July
29th, 2008 Dr. Moreau will be on campus Tuesday, July 29th. Individuals interested in meeting privately with him are encouraged to contact Louise Patterson at Louise_Patterson@Brown.edu or 863-3178.
_______________________________________________________ Events |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
