In this proposal, we develop Bayesian methodology for high dimensional genomic data. The overarching theme in this proposal is that we develop several novel statistical methods for motif discovery in genomic sequence data. Chromatin Immunoprecipitation microarray (ChIP-chip) data allows the direct identification of transcription factor binding sites that are active in particular biological states. Jointly modeling array intensities and DNA sequence will lead to more accurate estimation of binding sites. We develop these joint models to account for multiple motifs and varied relationships between binding sites and array intensities. We also propose a novel joint model framework for direct estimation of a motif using gene expression and the DNA sequence that bypasses computationally expensive motif selection procedures. Chromatin structure, in the form of positioning of nucleosomes in DNA, has long been known to play a huge role in protein-DNA binding, however, a quantitative assessment of this role has not been available until very recently. Taking advantage of the increasing availability of accurate experimental data assessing chromatin features, we propose a novel Bayesian statistical model framework for improving motif detection through integration of nucleosome positioning and genomic sequence data. Alternative splicing of mRNA greatly expands the functional repertoire of many genes in the mammalian genome by including or excluding the exons making up the genetic coding sequence. Standard gene expression arrays fail to capture the variability of the exon composition of mRNA species, but rather give a crude measure of overall gene expression. We propose a method that detects over-representation of specific splice junctions in different biological states while adjusting for overall gene expression. The advent of high-throughput genomic technologies has ushered in a new data-driven era, allowing the ability to measure biological activity on a genome-wide scale. Chromatin Immunoprecipitation (ChIP), histone modification, and FAIRE for example are procedures that benefited from this technology, allowing one to determine relative enrichment for their isolated fragments genome wide. The recent development of Next generation sequencing (NGS) platforms offers greater dynamic range, resolution, and genomic coverage in measuring relative enrichment of DNA fragments compared to microarrays. We develop classes of statistical mixture models based on the zero-inflated negative binomial distribution to model such count data and develop an R software package called Zero-Inflated Negative Binomial Algorithm (ZINBA) to carry out the peak calling for a given dataset. 1

Public Health Relevance

We develop Bayesian methodology for high dimensional genomic data. The overarching theme in this proposal is that we develop several novel statistical methods for motif discovery in genomic sequence data. The proposed methodology has major applications in chronic diseases such as cancer, AIDS, cardiovascular disease, and environmental health. We will develop new statistical methods for ChIP-chip data, integrating chormatin structure into motif discovery, joint modeling of gene expression and sequence data, alternative mRNA splicing, and analysis of next generation sequencing (NGS) data. 1

National Institute of Health (NIH)
National Institute of General Medical Sciences (NIGMS)
Research Project (R01)
Project #
Application #
Study Section
Biostatistical Methods and Research Design Study Section (BMRD)
Program Officer
Brazhnik, Paul
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
University of North Carolina Chapel Hill
Biostatistics & Other Math Sci
Schools of Public Health
Chapel Hill
United States
Zip Code
Zhu, Hongtu; Khondker, Zakaria; Lu, Zhaohua et al. (2014) Bayesian Generalized Low Rank Regression Models for Neuroimaging Phenotypes and Genetic Markers. J Am Stat Assoc 109:997-990
DE Castro, Mário; Chen, Ming-Hui; Ibrahim, Joseph G et al. (2014) Bayesian Transformation Models for Multivariate Survival Data. Scand Stat Theory Appl 41:187-199
Rashid, Naim U; Sun, Wei; Ibrahim, Joseph G (2014) Some Statistical Strategies for DAE-seq Data Analysis: Variable Selection and Modeling Dependencies among Observations. J Am Stat Assoc 109:78-94
Zhang, Yuanye; Chen, Ming-Hui; Ibrahim, Joseph G et al. (2014) Bayesian gamma frailty models for survival data with semi-competing risks and treatment switching. Lifetime Data Anal 20:76-105
Lewis, Paul O; Xie, Wangang; Chen, Ming-Hui et al. (2014) Posterior predictive Bayesian phylogenetic model selection. Syst Biol 63:309-21
Chen, Qingxia; Ibrahim, Joseph G (2014) A note on the relationships between multiple imputation, maximum likelihood and fully Bayesian methods for missing responses in linear regression models. Stat Interface 6:315-324
Zeng, Donglin; Ibrahim, Joseph G; Chen, Ming-Hui et al. (2014) Multivariate recurrent events in the presence of multivariate informative censoring with applications to bleeding and transfusion events in myelodysplastic syndrome. J Biopharm Stat 24:429-42
Chen, Ming-Hui; Ibrahim, Joseph G; Zeng, Donglin et al. (2014) Bayesian design of superiority clinical trials for recurrent events data with applications to bleeding and transfusion events in myelodyplastic syndrome. Biometrics 70:1003-13
Chen, Ming-Hui; Ibrahim, Joseph G; Amy Xia, H et al. (2014) Bayesian sequential meta-analysis design in evaluating cardiovascular risk in a new antidiabetic drug development program. Stat Med 33:1600-18
Zhu, Hongtu; Ibrahim, Joseph G; Tang, Niansheng (2014) Bayesian Sensitivity Analysis of Statistical Models with Missing Data. Stat Sin 24:871-896

Showing the most recent 10 out of 60 publications