Wing Hung Wong

The Calculation of Posterior Distributions by Data Augmentation

Abstract The idea of data augmentation arises naturally in missing value problems, as exemplified by the standard ways of filling in missing cells in balanced two-way tables. Thus data augmentation refers to a scheme of augmenting the observed data so as to make it more easy to analyze. This device is used to great advantage by the EM algorithm (Dempster, Laird, and Rubin 1977) in solving maximum likelihood problems. In situations when the likelihood cannot be approximated closely by the normal likelihood, maximum likelihood estimates and the associated standard errors cannot be relied upon to make valid inferential statements. From the Bayesian point of view, one must now calculate the posterior distribution of parameters of interest. If data augmentation can be used in the calculation of the maximum likelihood estimate, then in the same cases one ought to be able to use it in the computation of the posterior distribution. It is the purpose of this article to explain how this can be done. The basic idea ...

Classification of human lung carcinomas by mRNA expression profiling reveals distinct adenocarcinoma subclasses

We have generated a molecular taxonomy of lung carcinoma, the leading cause of cancer death in the United States and worldwide. Using oligonucleotide microarrays, we analyzed mRNA expression levels corresponding to 12,600 transcript sequences in 186 lung tumor samples, including 139 adenocarcinomas resected from the lung. Hierarchical and probabilistic clustering of expression data defined distinct subclasses of lung adenocarcinoma. Among these were tumors with high relative expression of neuroendocrine genes and of type II pneumocyte genes, respectively. Retrospective analysis revealed a less favorable outcome for the adenocarcinomas with neuroendocrine gene expression. The diagnostic potential of expression profiling is emphasized by its ability to discriminate primary lung adenocarcinomas from metastases of extra-pulmonary origin. These results suggest that integration of expression profile data with clinical parameters could aid in diagnosis of lung cancer patients.

Extensive and divergent circadian gene expression in liver and heart

Many mammalian peripheral tissues have circadian clocks; endogenous oscillators that generate transcriptional rhythms thought to be important for the daily timing of physiological processes. The extent of circadian gene regulation in peripheral tissues is unclear, and to what degree circadian regulation in different tissues involves common or specialized pathways is unknown. Here we report a comparative analysis of circadian gene expression in vivo in mouse liver and heart using oligonucleotide arrays representing 12,488 genes. We find that peripheral circadian gene regulation is extensive (≥8–10% of the genes expressed in each tissue), that the distributions of circadian phases in the two tissues are markedly different, and that very few genes show circadian regulation in both tissues. This specificity of circadian regulation cannot be accounted for by tissue-specific gene expression. Despite this divergence, the clock-regulated genes in liver and heart participate in overlapping, extremely diverse processes. A core set of 37 genes with similar circadian regulation in both tissues includes candidates for new clock genes and output genes, and it contains genes responsive to circulating factors with circadian or diurnal rhythms.

Genome-wide expression analysis reveals dysregulation of myelination-related genes in chronic schizophrenia

Neuropathological and brain imaging studies suggest that schizophrenia may result from neurodevelopmental defects. Cytoarchitectural studies indicate cellular abnormalities suggestive of a disruption in neuronal connectivity in schizophrenia, particularly in the dorsolateral prefrontal cortex. Yet, the molecular mechanisms underlying these findings remain unclear. To identify molecular substrates associated with schizophrenia, DNA microarray analysis was used to assay gene expression levels in postmortem dorsolateral prefrontal cortex of schizophrenic and control patients. Genes determined to have altered expression levels in schizophrenics relative to controls are involved in a number of biological processes, including synaptic plasticity, neuronal development, neurotransmission, and signal transduction. Most notable was the differential expression of myelination-related genes suggesting a disruption in oligodendrocyte function in schizophrenia.

Model-based analysis of oligonucleotide arrays: model validation, design issues and standard error application

BackgroundA model-based analysis of oligonucleotide expression arrays we developed previously uses a probe-sensitivity index to capture the response characteristic of a specific probe pair and calculates model-based expression indexes (MBEI). MBEI has standard error attached to it as a measure of accuracy. Here we investigate the stability of the probe-sensitivity index across different tissue types, the reproducibility of results in replicate experiments, and the use of MBEI in perfect match (PM)-only arrays.ResultsProbe-sensitivity indexes are stable across tissue types. The target genes presence in many arrays of an array set allows the probe-sensitivity index to be estimated accurately. We extended the model to obtain expression values for PM-only arrays, and found that the 20-probe PM-only model is comparable to the 10-probe PM/MM difference model, in terms of the expression correlations with the original 20-probe PM/MM difference model. MBEI method is able to extend the reliable detection limit of expression to a lower mRNA concentration. The standard errors of MBEI can be used to construct confidence intervals of fold changes, and the lower confidence bound of fold change is a better ranking statistic for filtering genes. We can assign reliability indexes for genes in a specific cluster of interest in hierarchical clustering by resampling clustering trees. A software dChip implementing many of these analysis methods is made available.ConclusionsThe model-based approach reduces the variability of low expression estimates, and provides a natural method of calculating expression values for PM-only arrays. The standard errors attached to expression values can be used to assess the reliability of downstream analysis.

FoxOs Are Lineage-Restricted Redundant Tumor Suppressors and Regulate Endothelial Cell Homeostasis

Activated phosphoinositide 3-kinase (PI3K)-AKT signaling appears to be an obligate event in the development of cancer. The highly related members of the mammalian FoxO transcription factor family, FoxO1, FoxO3, and FoxO4, represent one of several effector arms of PI3K-AKT signaling, prompting genetic analysis of the role of FoxOs in the neoplastic phenotypes linked to PI3K-AKT activation. While germline or somatic deletion of up to five FoxO alleles produced remarkably modest neoplastic phenotypes, broad somatic deletion of all FoxOs engendered a progressive cancer-prone condition characterized by thymic lymphomas and hemangiomas, demonstrating that the mammalian FoxOs are indeed bona fide tumor suppressors. Transcriptome and promoter analyses of differentially affected endothelium identified direct FoxO targets and revealed that FoxO regulation of these targets in vivo is highly context-specific, even in the same cell type. Functional studies validated Sprouty2 and PBX1, among others, as FoxO-regulated mediators of endothelial cell morphogenesis and vascular homeostasis.

Sequential Imputations and Bayesian Missing Data Problems

Abstract For missing data problems, Tanner and Wong have described a data augmentation procedure that approximates the actual posterior distribution of the parameter vector by a mixture of complete data posteriors. Their method of constructing the complete data sets is closely related to the Gibbs sampler. Both required iterations, and, similar to the EM algorithm, convergence can be slow. We introduce in this article an alternative procedure that involves imputing the missing data sequentially and computing appropriate importance sampling weights. In many applications this new procedure works very well without the need for iterations. Sensitivity analysis, influence analysis, and updating with new data can be performed cheaply. Bayesian prediction and model selection can also be incorporated. Examples taken from a wide range of applications are used for illustration.

An integrated software system for analyzing ChIP-chip and ChIP-seq data

We present CisGenome, a software system for analyzing genome-wide chromatin immunoprecipitation (ChIP) data. CisGenome is designed to meet all basic needs of ChIP data analyses, including visualization, data normalization, peak detection, false discovery rate computation, gene-peak association, and sequence and motif analysis. In addition to implementing previously published ChIP–microarray (ChIP-chip) analysis methods, the software contains statistical methods designed specifically for ChlP sequencing (ChIP-seq) data obtained by coupling ChIP with massively parallel sequencing. The modular design of CisGenome enables it to support interactive analyses through a graphic user interface as well as customized batch-mode computation for advanced data mining. A built-in browser allows visualization of array images, signals, gene structure, conservation, and DNA sequence and motif information. We demonstrate the use of these tools by a comparative analysis of ChIP-chip and ChIP-seq data for the transcription factor NRSF/REST, a study of ChIP-seq analysis with or without a negative control sample, and an analysis of a new motif in Nanog- and Sox2-binding regions.

A genomic storm in critically injured humans

Critical injury in humans induces a genomic storm with simultaneous changes in expression of innate and adaptive immunity genes.

p53 and Pten control neural and glioma stem/progenitor cell renewal and differentiation.

Glioblastoma (GBM) is a highly lethal brain tumour presenting as one of two subtypes with distinct clinical histories and molecular profiles. The primary GBM subtype presents acutely as a high-grade disease that typically harbours mutations in EGFR, PTEN and INK4A/ARF (also known as CDKN2A), and the secondary GBM subtype evolves from the slow progression of a low-grade disease that classically possesses PDGF and TP53 events. Here we show that concomitant central nervous system (CNS)-specific deletion of p53 and Pten in the mouse CNS generates a penetrant acute-onset high-grade malignant glioma phenotype with notable clinical, pathological and molecular resemblance to primary GBM in humans. This genetic observation prompted TP53 and PTEN mutational analysis in human primary GBM, demonstrating unexpectedly frequent inactivating mutations of TP53 as well as the expected PTEN mutations. Integrated transcriptomic profiling, in silico promoter analysis and functional studies of murine neural stem cells (NSCs) established that dual, but not singular, inactivation of p53 and Pten promotes an undifferentiated state with high renewal potential and drives increased Myc protein levels and its associated signature. Functional studies validated increased Myc activity as a potent contributor to the impaired differentiation and enhanced renewal of NSCs doubly null for p53 and Pten (p53-/- Pten-/-) as well as tumour neurospheres (TNSs) derived from this model. Myc also serves to maintain robust tumorigenic potential of p53-/- Pten-/- TNSs. These murine modelling studies, together with confirmatory transcriptomic/promoter studies in human primary GBM, validate a pathogenetic role of a common tumour suppressor mutation profile in human primary GBM and establish Myc as an important target for cooperative actions of p53 and Pten in the regulation of normal and malignant stem/progenitor cell differentiation, self-renewal and tumorigenic potential.