A Predictive Framework for Integrating Disparate Genomic Data Types Using Sample-Specific Gene Set Enrichment Analysis and Multi-Task Learning

Understanding the root molecular and genetic causes driving complex traits is a fundamental challenge in genomics and genetics. Numerous studies have used variation in gene expression to understand complex traits, but the underlying genomic variation that contributes to these expression changes is not well understood. In this study, we developed a framework to integrate gene expression and genotype data to identify biological differences between samples from opposing complex trait classes that are driven by expression changes and genotypic variation. This framework utilizes pathway analysis and multi-task learning to build a predictive model and discover pathways relevant to the complex trait of interest. We simulated expression and genotype data to test the predictive ability of our framework and to measure how well it uncovered pathways with genes both differentially expressed and genetically associated with a complex trait. We found that the predictive performance of the multi-task model was comparable to other similar methods. Also, methods like multi-task learning that considered enrichment analysis scores from both data sets found pathways with both genetic and expression differences related to the phenotype. We used our framework to analyze differences between estrogen receptor (ER) positive and negative breast cancer samples. An analysis of the top 15 gene sets from the multi-task model showed they were all related to estrogen, steroids, cell signaling, or the cell cycle. Although our study suggests that multi-task learning does not enhance predictive accuracy, the models generated by our framework do provide valuable biological pathway knowledge for complex traits.     

Nonparametric methods for identifying differentially expressed genes in microarray data

MOTIVATION: Gene expression experiments provide a fast and systematic way to identify disease markers relevant to clinical care. In this study, we address the problem of robust identification of differentially expressed genes from microarray data. Differentially expressed genes, or discriminator genes, are genes with significantly different expression in two user-defined groups of microarray experiments. We compare three model-free approaches: (1). nonparametric t-test, (2). Wilcoxon (or Mann-Whitney) rank sum test, and (3). a heuristic method based on high Pearson correlation to a perfectly differentiating gene ('ideal discriminator method'). We systematically assess the performance of each method based on simulated and biological data under varying noise levels and p-value cutoffs. RESULTS: All methods exhibit very low false positive rates and identify a large fraction of the differentially expressed genes in simulated data sets with noise level similar to that of actual data. Overall, the rank sum test appears most conservative, which may be advantageous when the computationally identified genes need to be tested biologically. However, if a more inclusive list of markers is desired, a higher p-value cutoff or the nonparametric t-test may be appropriate. When applied to data from lung tumor and lymphoma data sets, the methods identify biologically relevant differentially expressed genes that allow clear separation of groups in question. Thus the methods described and evaluated here provide a convenient and robust way to identify differentially expressed genes for further biological and clinical analysis.     

Capturing heterogeneity in gene expression studies by surrogate variable analysis

It has unambiguously been shown that genetic, environmental, demographic, and technical factors may have substantial effects on gene expression levels. In addition to the measured variable(s) of interest, there will tend to be sources of signal due to factors that are unknown, unmeasured, or too complicated to capture through simple models. We show that failing to incorporate these sources of heterogeneity into an analysis can have widespread and detrimental effects on the study. Not only can this reduce power or induce unwanted dependence across genes, but it can also introduce sources of spurious signal to many genes. This phenomenon is true even for well-designed, randomized studies. We introduce "surrogate variable analysis" (SVA) to overcome the problems caused by heterogeneity in expression studies. SVA can be applied in conjunction with standard analysis techniques to accurately capture the relationship between expression and any modeled variables of interest. We apply SVA to disease class, time course, and genetics of gene expression studies. We show that SVA increases the biological accuracy and reproducibility of analyses in genome-wide expression studies.     

Network Analysis of Differential Expression for the Identification of Disease-Causing Genes

Genetic studies (in particular linkage and association studies) identify chromosomal regions involved in a disease or phenotype of interest, but those regions often contain many candidate genes, only a few of which can be followed-up for biological validation. Recently, computational methods to identify (prioritize) the most promising candidates within a region have been proposed, but they are usually not applicable to cases where little is known about the phenotype (no or few confirmed disease genes, fragmentary understanding of the biological cascades involved). We seek to overcome this limitation by replacing knowledge about the biological process by experimental data on differential gene expression between affected and healthy individuals. Considering the problem from the perspective of a gene/protein network, we assess a candidate gene by considering the level of differential expression in its neighborhood under the assumption that strong candidates will tend to be surrounded by differentially expressed neighbors. We define a notion of soft neighborhood where each gene is given a contributing weight, which decreases with the distance from the candidate gene on the protein network. To account for multiple paths between genes, we define the distance using the Laplacian exponential diffusion kernel. We score candidates by aggregating the differential expression of neighbors weighted as a function of distance. Through a randomization procedure, we rank candidates by p-values. We illustrate our approach on four monogenic diseases and successfully prioritize the known disease causing genes.     

Detecting differentially expressed genes by relative entropy

DNA microarray experiments have generated large amount of gene expression measurements across different conditions. One crucial step in the analysis of these data is to detect differentially expressed genes. Some parametric methods, including the two-sample t-test (T-test) and variations of it, have been used. Alternatively, a class of non-parametric algorithms, such as the Wilcoxon rank sum test (WRST), significance analysis of microarrays (SAM) of Tusher et al. (2001), the empirical Bayesian (EB) method of Efron et al. (2001), etc., have been proposed. Most available popular methods are based on t-statistic. Due to the quality of the statistic that they used to describe the difference between groups of data, there are situations when these methods are inefficient, especially when the data follows multi-modal distributions. For example, some genes may display different expression patterns in the same cell type, say, tumor or normal, to form some subtypes. Most available methods are likely to miss these genes. We developed a new non-parametric method for selecting differentially expressed genes by relative entropy, called SDEGRE, to detect differentially expressed genes by combining relative entropy and kernel density estimation, which can detect all types of differences between two groups of samples. The significance of whether a gene is differentially expressed or not can be estimated by resampling-based permutations. We illustrate our method on two data sets from Golub et al. (1999) and Alon et al. (1999). Comparing the results with those of the T-test, the WRST and the SAM, we identified novel differentially expressed genes which are of biological significance through previous biological studies while they were not detected by the other three methods. The results also show that the genes selected by SDEGRE have a better capability to distinguish the two cell types.     

Aberrant Gene Expression in Humans

Gene expression as an intermediate molecular phenotype has been a focus of research interest. In particular, studies of expression quantitative trait loci (eQTL) have offered promise for understanding gene regulation through the discovery of genetic variants that explain variation in gene expression levels. Existing eQTL methods are designed for assessing the effects of common variants, but not rare variants. Here, we address the problem by establishing a novel analytical framework for evaluating the effects of rare or private variants on gene expression. Our method starts from the identification of outlier individuals that show markedly different gene expression from the majority of a population, and then reveals the contributions of private SNPs to the aberrant gene expression in these outliers. Using population-scale mRNA sequencing data, we identify outlier individuals using a multivariate approach. We find that outlier individuals are more readily detected with respect to gene sets that include genes involved in cellular regulation and signal transduction, and less likely to be detected with respect to the gene sets with genes involved in metabolic pathways and other fundamental molecular functions. Analysis of polymorphic data suggests that private SNPs of outlier individuals are enriched in the enhancer and promoter regions of corresponding aberrantly-expressed genes, suggesting a specific regulatory role of private SNPs, while the commonly-occurring regulatory genetic variants (i.e., eQTL SNPs) show little evidence of involvement. Additional data suggest that non-genetic factors may also underlie aberrant gene expression. Taken together, our findings advance a novel viewpoint relevant to situations wherein common eQTLs fail to predict gene expression when heritable, rare inter-individual variation exists. The analytical framework we describe, taking into consideration the reality of differential phenotypic robustness, may be valuable for investigating complex traits and conditions.     

Normalization of microarray data using a spatial mixed model analysis which includes splines

MOTIVATION: Microarray experiments with thousands of genes on a slide and multiple slides used in any experimental set represent a large body of data with many sources of variation. The identification of such sources of variation within microarray experimental sets is critical for correct deciphering of desired gene expression differences. RESULTS: We describe new methods for the normalization using spatial mixed models which include splines and analysis of two-colour spotted microarrays for within slide variation and for a series of slides. The model typically explains 45-85% of the variation on a slide with only approximately 1% of the total degrees of freedom. The results from our methods compare favourably with those from intensity dependent normalization loess methods where we accounted for twice as much uncontrolled and unwanted variation on the slides. We have also developed an index for each EST that combines the various measures of the differential response into a single value that researchers can use to rapidly assess the genes of interest.     

Gene selection for oligonucleotide array: an approach using PM probe level data

MOTIVATION: Analysis of oligonucleotide array data, especially to select genes of interest, is a highly challenging task because of the large volume of information and various experimental factors. Moreover, interaction effect (i.e. expression changes depend on probe effects) complicates the analysis because current methods often use an additive model to analyze data. We propose an approach to address these issues with the aim of producing a more reliable selection of differentially expressed genes. The approach uses the rank for normalization, employs the percentile-range to measure expression variation, and applies various filters to monitor expression changes. RESULTS: We compare our approach with MAS and Dchip models. A data set from an angiogenesis study is used for illustration. Results show that our approach performs better than other methods either in identification of the positive control gene or in PCR confirmatory tests. In addition, the invariant set of genes in our approach provides an efficient way for normalization.     

Estimating the false discovery rate using nonparametric deconvolution

Given a set of microarray data, the problem is to detect differentially expressed genes, using a false discovery rate (FDR) criterion. As opposed to common procedures in the literature, we do not base the selection criterion on statistical significance only, but also on the effect size. Therefore, we select only those genes that are significantly more differentially expressed than some f-fold (e.g., f = 2). This corresponds to use of an interval null domain for the effect size. Based on a simple error model, we discuss a naive estimator for the FDR, interpreted as the probability that the parameter of interest lies in the null-domain (e.g., mu < log(2)(2) = 1) given that the test statistic exceeds a threshold. We improve the naive estimator by using deconvolution. That is, the density of the parameter of interest is recovered from the data. We study performance of the methods using simulations and real data.     

Estimating cell type composition using isoform expression one gene at a time

Human tissue samples are often mixtures of heterogeneous cell types, which can confound the analyses of gene expression data derived from such tissues. The cell type composition of a tissue sample may itself be of interest and is needed for proper analysis of differential gene expression. A variety of computational methods have been developed to estimate cell type proportions using gene-level expression data. However, RNA isoforms can also be differentially expressed across cell types, and isoform-level expression could be equally or more informative for determining cell type origin than gene-level expression. We propose a new computational method, IsoDeconvMM, which estimates cell type fractions using isoform-level gene expression data. A novel and useful feature of IsoDeconvMM is that it can estimate cell type proportions using only a single gene, though in practice we recommend aggregating estimates of a few dozen genes to obtain more accurate results. We demonstrate the performance of IsoDeconvMM using a unique data set with cell type–specific RNA-seq data across more than 135 individuals. This data set allows us to evaluate different methods given the biological variation of cell type–specific gene expression data across individuals. We further complement this analysis with additional simulations.     

RUV-III-NB: normalization of single cell RNA-seq data

Normalization of single cell RNA-seq data remains a challenging task. The performance of different methods can vary greatly between datasets when unwanted factors and biology are associated. Most normalization methods also only remove the effects of unwanted variation for the cell embedding but not from gene-level data typically used for differential expression (DE) analysis to identify marker genes. We propose RUV-III-NB, a method that can be used to remove unwanted variation from both the cell embedding and gene-level counts. Using pseudo-replicates, RUV-III-NB explicitly takes into account potential association with biology when removing unwanted variation. The method can be used for both UMI or read counts and returns adjusted counts that can be used for downstream analyses such as clustering, DE and pseudotime analyses. Using published datasets with different technological platforms, kinds of biology and levels of association between biology and unwanted variation, we show that RUV-III-NB manages to remove library size and batch effects, strengthen biological signals, improve DE analyses, and lead to results exhibiting greater concordance with independent datasets of the same kind. The performance of RUV-III-NB is consistent and is not sensitive to the number of factors assumed to contribute to the unwanted variation.     

Replication, Variation and Normalisation in Microarray Experiments

INTRODUCTION: Microarray experiments often have complex designs that include sample pooling, biological and technical replication, sample pairing and dye-swapping. This article demonstrates how statistical modelling can illuminate issues in the design and analysis of microarray experiments, and this information can then be used to plan effective studies. METHODS: A very detailed statistical model for microarray data is introduced, to show the possible sources of variation that are present in even the simplest microarray experiments. Based on this model, the efficacy of common experimental designs, normalisation methodologies and analyses is determined. RESULTS: When the cost of the arrays is high compared with the cost of samples, sample pooling and spot replication are shown to be efficient variance reduction methods, whereas technical replication of whole arrays is demonstrated to be very inefficient. Dye-swap designs can use biological replicates rather than technical replicates to improve efficiency and simplify analysis. When the cost of samples is high and technical variation is a major portion of the error, technical replication can be cost effective. Normalisation by centreing on a small number of spots may reduce array effects, but can introduce considerable variation in the results. Centreing using the bulk of spots on the array is less variable. Similarly, normalisation methods based on regression methods can introduce variability. Except for normalisation methods based on spiking controls, all normalisation requires that most genes do not differentially express. Methods based on spatial location and/or intensity also require that the nondifferentially expressing genes are at random with respect to location and intensity. Spotting designs should be carefully done so that spot replicates are widely spaced on the array, and genes with similar expression patterns are not clustered. DISCUSSION: The tools for statistical design of experiments can be applied to microarray experiments to improve both efficiency and validity of the studies. Given the high cost of microarray experiments, the benefits of statistical input prior to running the experiment cannot be over-emphasised.     

Mixture models with multiple levels, with application to the analysis of multifactor gene expression data

Model-based clustering is a popular tool for summarizing high-dimensional data. With the number of high-throughput large-scale gene expression studies still on the rise, the need for effective data- summarizing tools has never been greater. By grouping genes according to a common experimental expression profile, we may gain new insight into the biological pathways that steer biological processes of interest. Clustering of gene profiles can also assist in assigning functions to genes that have not yet been functionally annotated. In this paper, we propose 2 model selection procedures for model-based clustering. Model selection in model-based clustering has to date focused on the identification of data dimensions that are relevant for clustering. However, in more complex data structures, with multiple experimental factors, such an approach does not provide easily interpreted clustering outcomes. We propose a mixture model with multiple levels, , that provides sparse representations both "within" and "between" cluster profiles. We explore various flexible "within-cluster" parameterizations and discuss how efficient parameterizations can greatly enhance the objective interpretability of the generated clusters. Moreover, we allow for a sparse "between-cluster" representation with a different number of clusters at different levels of an experimental factor of interest. This enhances interpretability of clusters generated in multiple-factor contexts. Interpretable cluster profiles can assist in detecting biologically relevant groups of genes that may be missed with less efficient parameterizations. We use our multilevel mixture model to mine a proliferating cell line expression data set for annotational context and regulatory motifs. We also investigate the performance of the multilevel clustering approach on several simulated data sets.     

Utilizing logical relationships in genomic data to decipher cellular processes

The wealth of available genomic data has spawned a corresponding interest in computational methods that can impart biological meaning and context to these experiments. Traditional computational methods have drawn relationships between pairs of proteins or genes based on notions of equality or similarity between their patterns of occurrence or behavior. For example, two genes displaying similar variation in expression, over a number of experiments, may be predicted to be functionally related. We have introduced a natural extension of these approaches, instead identifying logical relationships involving triplets of proteins. Triplets provide for various discrete kinds of logic relationships, leading to detailed inferences about biological associations. For instance, a protein C might be encoded within an organism if, and only if, two other proteins A and B are also both encoded within the organism, thus suggesting that gene C is functionally related to genes A and B. The method has been applied fruitfully to both phylogenetic and microarray expression data, and has been used to associate logical combinations of protein activity with disease state phenotypes, revealing previously unknown ternary relationships among proteins, and illustrating the inherent complexities that arise in biological data.     

“Topological Significance” Analysis of Gene Expression and Proteomic Profiles from Prostate Cancer Cells Reveals Key Mechanisms of Androgen Response

Background

The problem of prostate cancer progression to androgen independence has been extensively studied. Several studies systematically analyzed gene expression profiles in the context of biological networks and pathways, uncovering novel aspects of prostate cancer. Despite significant research efforts, the mechanisms underlying tumor progression are poorly understood. We applied a novel approach to reconstruct system-wide molecular events following stimulation of LNCaP prostate cancer cells with synthetic androgen and to identify potential mechanisms of androgen-independent progression of prostate cancer.

Methodology/Principal Findings

We have performed concurrent measurements of gene expression and protein levels following the treatment using microarrays and iTRAQ proteomics. Sets of up-regulated genes and proteins were analyzed using our novel concept of “topological significance”. This method combines high-throughput molecular data with the global network of protein interactions to identify nodes which occupy significant network positions with respect to differentially expressed genes or proteins. Our analysis identified the network of growth factor regulation of cell cycle as the main response module for androgen treatment in LNCap cells. We show that the majority of signaling nodes in this network occupy significant positions with respect to the observed gene expression and proteomic profiles elicited by androgen stimulus. Our results further indicate that growth factor signaling probably represents a “second phase” response, not directly dependent on the initial androgen stimulus.

Conclusions/Significance

We conclude that in prostate cancer cells the proliferative signals are likely to be transmitted from multiple growth factor receptors by a multitude of signaling pathways converging on several key regulators of cell proliferation such as c-Myc, Cyclin D and CREB1. Moreover, these pathways are not isolated but constitute an interconnected network module containing many alternative routes from inputs to outputs. If the whole network is involved, a precisely formulated combination therapy may be required to fight the tumor growth effectively.