Visualisation of gene expression data - the GE-biplot, the Chip-plot and the Gene-plot

Medical and Social Determinants of Brain Health and Dementia in a Multicultural Community Cohort of Older Adults

BACKGROUND

Socioeconomic status (SES), race, ethnicity, and medical comorbidities may contribute to Alzheimer's disease and related disorders (ADRD) health disparities.

OBJECTIVE

Analyze effects of social and medical determinants on cognition in 374 multicultural older adults participating in a community-based dementia screening program.

METHODS

We used the Montreal Cognitive Assessment (MoCA) and AD8 as measures of cognition, and a 3-way race/ethnicity variable (White, African American, Hispanic) and SES (Hollingshead index) as predictors. Potential contributors to health disparities included: age, sex, education, total medical comorbidities, health self-ratings, and depression. We applied K-means cluster analyses to study medical and social dimension effects on cognitive outcomes.

RESULTS

African Americans and Hispanics had lower SES status and cognitive performance compared with similarly aged Whites. We defined three clusters based on age and SES. Cluster #1 and #3 differed by SES but not age, while cluster #2 was younger with midlevel SES. Cluster #1 experienced the worse health outcomes while cluster #3 had the best health outcomes. Within each cluster, White participants had higher SES and better health outcomes, African Americans had the worst physical performance, and Hispanics had the most depressive symptoms. In cross-cluster comparisons, higher SES led to better health outcomes for all participants.

CONCLUSION

SES may contribute to disparities in access to healthcare services, while race and ethnicity may contribute to disparities in the quality and extent of services received. Our study highlights the need to critically address potential interactions between race, ethnicity, and SES which may better explain disparities in ADRD health outcomes.

Mining Gene Expression Profile with Missing Values: An Integration of Kernel PCA and Robust Singular Values Decomposition

Background:

Gene expression profiling and transcriptomics provide valuable information about the role of genes that are differentially expressed between two or more samples. It is always important and challenging to analyse High-throughput DNA microarray data with a number of missing values under various experimental conditions. </P><P> Objectives: Graphical data visualizations of the expression of all genes in a particular cell provide holistic views of gene expression patterns, which improve our understanding of cellular systems under normal and pathological conditions. However, current visualization methods are sensitive to missing values, which are frequently observed in microarray-based gene expression profiling, potentially affecting the subsequent statistical analyses.

Methods:

We addressed in this study the problem of missing values with respect to different imputation methods using gene expression biplot (GE biplot), one of the most popular gene visualization techniques. The effects of missing values for mining differentially expressed genes in gene expression data were evaluated using four well-known imputation methods: Robust Singular Value Decomposition (Robust SVD), Column Average (CA), Column Median (CM), and K-nearest Neighbors (KNN). Frobenius norm and absolute distances were used to measure the accuracy of the methods.

Results:

Three numerical experiments were performed using simulated data (i) and publicly available colon cancer (ii) and leukemia data (iii) to analyze the performance of each method. The results showed that CM and KNN performed better than Robust SVD and CA for identifying the index gene profile in the biplot visualization in both the simulation study and the colon cancer and leukemia microarray datasets.

Conclusion:

The impact of missing values on the GE biplot was smaller when the data matrix was imputed by KNN than by CM. This study concluded that KNN performed satisfactorily in generating a GE biplot in the presence of missing values in microarray data.

LOHAS: loss-of-heterozygosity analysis suite

Detection of loss of heterozygosity (LOH) plays an important role in genetic, genomic and cancer research. We develop computational methods to estimate the proportion of homozygous SNP calls, identify samples with structural alterations and/or unusual genotypic patterns, cluster samples with close LOH structures and map the genomic segments bearing LOH by analyzing data of genome-wide SNP arrays or customized SNP arrays. In addition to cancer genetics/genomics, we also apply the methods to study long contiguous stretches of homozygosity (LCSH) in general populations. The LCSH analysis aids in the identification of samples with complex LCSH patterns indicative of nonrandom mating and/or meiotic recombination cold spots, separation of samples with different genetic backgrounds and sex, and mapping of regions of LCSH. Affymetrix Human Mapping 500K Set SNP data from an acute lymphoblastic leukemia study containing 304 cancer patients and 50 normal controls and from the HapMap Project containing 30 African trios, 30 Caucasian trios and 90 independent Asian samples were analyzed. We identified common gene regions of LOH, e.g., ETV6 and CDKN1B, and identified frequent regions of LCSH, e.g., the region that encompasses the centromeric gene desert region of chromosome 16. Unsupervised analysis separated cancer subtypes and ethnic subpopulations by patterns of LOH/LCSH. Simulation studies considering LOH width, effect size and heterozygous interference fraction were performed, and the results show that the proposed LOH association test has good test power and controls type 1 error well. The developed algorithms are packaged into LOHAS written in R and R GUI.

Kernel-PCA data integration with enhanced interpretability

Background

Nowadays, combining the different sources of information to improve the biological knowledge available is a challenge in bioinformatics. One of the most powerful methods for integrating heterogeneous data types are kernel-based methods. Kernel-based data integration approaches consist of two basic steps: firstly the right kernel is chosen for each data set; secondly the kernels from the different data sources are combined to give a complete representation of the available data for a given statistical task.

Results

We analyze the integration of data from several sources of information using kernel PCA, from the point of view of reducing dimensionality. Moreover, we improve the interpretability of kernel PCA by adding to the plot the representation of the input variables that belong to any dataset. In particular, for each input variable or linear combination of input variables, we can represent the direction of maximum growth locally, which allows us to identify those samples with higher/lower values of the variables analyzed.

Conclusions

The integration of different datasets and the simultaneous representation of samples and variables together give us a better understanding of biological knowledge.

A comparison of methods for classifying clinical samples based on proteomics data: a case study for statistical and machine learning approaches

The discovery of protein variation is an important strategy in disease diagnosis within the biological sciences. The current benchmark for elucidating information from multiple biological variables is the so called “omics” disciplines of the biological sciences. Such variability is uncovered by implementation of multivariable data mining techniques which come under two primary categories, machine learning strategies and statistical based approaches. Typically proteomic studies can produce hundreds or thousands of variables, p, per observation, n, depending on the analytical platform or method employed to generate the data. Many classification methods are limited by an n≪p constraint, and as such, require pre-treatment to reduce the dimensionality prior to classification. Recently machine learning techniques have gained popularity in the field for their ability to successfully classify unknown samples. One limitation of such methods is the lack of a functional model allowing meaningful interpretation of results in terms of the features used for classification. This is a problem that might be solved using a statistical model-based approach where not only is the importance of the individual protein explicit, they are combined into a readily interpretable classification rule without relying on a black box approach. Here we incorporate statistical dimension reduction techniques Partial Least Squares (PLS) and Principal Components Analysis (PCA) followed by both statistical and machine learning classification methods, and compared them to a popular machine learning technique, Support Vector Machines (SVM). Both PLS and SVM demonstrate strong utility for proteomic classification problems.

Mining gene expression profiles: an integrated implementation of kernel principal component analysis and singular value decomposition

The detection of genes that show similar profiles under different experimental conditions is often an initial step in inferring the biological significance of such genes. Visualization tools are used to identify genes with similar profiles in microarray studies. Given the large number of genes recorded in microarray experiments, gene expression data are generally displayed on a low dimensional plot, based on linear methods. However, microarray data show nonlinearity, due to high-order terms of interaction between genes, so alternative approaches, such as kernel methods, may be more appropriate. We introduce a technique that combines kernel principal component analysis (KPCA) and Biplot to visualize gene expression profiles. Our approach relies on the singular value decomposition of the input matrix and incorporates an additional step that involves KPCA. The main properties of our method are the extraction of nonlinear features and the preservation of the input variables (genes) in the output display. We apply this algorithm to colon tumor, leukemia and lymphoma datasets. Our approach reveals the underlying structure of the gene expression profiles and provides a more intuitive understanding of the gene and sample association.

A new analysis tool for individual-level allele frequency for genomic studies

Background

Allele frequency is one of the most important population indices and has been broadly applied to genetic/genomic studies. Estimation of allele frequency using genotypes is convenient but may lose data information and be sensitive to genotyping errors.

Results

This study utilizes a unified intensity-measuring approach to estimating individual-level allele frequencies for 1,104 and 1,270 samples genotyped with the single-nucleotide-polymorphism arrays of the Affymetrix Human Mapping 100K and 500K Sets, respectively. Allele frequencies of all samples are estimated and adjusted by coefficients of preferential amplification/hybridization (CPA), and large ethnicity-specific and cross-ethnicity databases of CPA and allele frequency are established. The results show that using the CPA significantly improves the accuracy of allele frequency estimates; moreover, this paramount factor is insensitive to the time of data acquisition, effect of laboratory site, type of gene chip, and phenotypic status. Based on accurate allele frequency estimates, analytic methods based on individual-level allele frequencies are developed and successfully applied to discover genomic patterns of allele frequencies, detect chromosomal abnormalities, classify sample groups, identify outlier samples, and estimate the purity of tumor samples. The methods are packaged into a new analysis tool, ALOHA (Allele-frequency/Loss-of-heterozygosity/Allele-imbalance).

Conclusions

This is the first time that these important genetic/genomic applications have been simultaneously conducted by the analyses of individual-level allele frequencies estimated by a unified intensity-measuring approach. We expect that additional practical applications for allele frequency analysis will be found. The developed databases and tools provide useful resources for human genome analysis via high-throughput single-nucleotide-polymorphism arrays. The ALOHA software was written in R and R GUI and can be downloaded at http://www.stat.sinica.edu.tw/hsinchou/genetics/aloha/ALOHA.htm.

Nonparametric estimation of LOH using Affymetrix SNP genotyping arrays for unpaired samples

Studies of loss of heterozygosity (LOH) play an important role in cancer research. In this paper, we developed a two-step procedure to examine LOH by comparing unpaired tumour and normal samples. In the first step we determined which chromosomes significantly differ between the two sets of samples by using nonparametric procedures. We then used the biplot data visualisation technique and homozygosity intensity estimates to determine the regions of these chromosomes that required further examination. We illustrated our method by examining 22 autosomes in samples of 95 normal controls and 14 acute lymphoblastic leukaemia patients. The genomewide scan of LOH with the Affymetrix Human Mapping 100K Set successfully identified the important tumour suppressor gene, CDKN2A, whose deletion was validated by quantitative polymerase chain reaction in multiple patients of this study.

H-Profile plots for the discovery and exploration of patterns in gene expression data with an application to time course data

BACKGROUND

An ever increasing number of techniques are being used to find genes with similar profiles from microarray studies. Visualization of gene expression profiles can aid this process, potentially contributing to the identification of co-regulated genes and gene function as well as network development.

RESULTS

We introduce the h-Profile plot to display gene expression profiles. Thumbnail versions of plots of gene expression profiles are plotted at coordinates such that profiles of similar shape are located in the same sector, with decreasing variance towards the origin. Negatively correlated profiles can easily be identified. A new method for selecting genes with fixed periodicity, but different phase and amplitude is described and used to demonstrate the use of the plots on cell cycle data.

CONCLUSION

Visualization tools for gene expression data are important and h-profile plots provide a timely contribution to the field. They allow the simultaneous visualization of many gene expression profiles and can be used for the identification of genes with similar or reversed profiles, the foundation step in many analyses.

Vascular microarray profiling in two models of hypertension identifies caveolin-1, Rgs2 and Rgs5 as antihypertensive targets

BACKGROUND

Hypertension is a complex disease with many contributory genetic and environmental factors. We aimed to identify common targets for therapy by gene expression profiling of a resistance artery taken from animals representing two different models of hypertension. We studied gene expression and morphology of a saphenous artery branch in normotensive WKY rats, spontaneously hypertensive rats (SHR) and adrenocorticotropic hormone (ACTH)-induced hypertensive rats.

RESULTS

Differential remodeling of arteries occurred in SHR and ACTH-treated rats, involving changes in both smooth muscle and endothelium. Increased expression of smooth muscle cell growth promoters and decreased expression of growth suppressors confirmed smooth muscle cell proliferation in SHR but not in ACTH. Differential gene expression between arteries from the two hypertensive models extended to the renin-angiotensin system, MAP kinase pathways, mitochondrial activity, lipid metabolism, extracellular matrix and calcium handling. In contrast, arteries from both hypertensive models exhibited significant increases in caveolin-1 expression and decreases in the regulators of G-protein signalling, Rgs2 and Rgs5. Increased protein expression of caveolin-1 and increased incidence of caveolae was found in both smooth muscle and endothelial cells of arteries from both hypertensive models.

CONCLUSION

We conclude that the majority of differences in gene expression found in the saphenous artery taken from rats with two different forms of hypertension reflect distinctive morphological and physiological alterations. However, changes in common to caveolin-1 expression and G protein signalling, through attenuation of Rgs2 and Rgs5, may contribute to hypertension through augmentation of vasoconstrictor pathways and provide potential targets for common drug development.

Primary metabolic pathways and signal transduction in sunflower (Helianthus annuus L.): comparison of transcriptional profiling in leaves and immature embryos using cDNA microarrays

The early stage of embryo development is a critical step in plant production. To identify genes with potential roles in the early sunflower seed development, a cDNA microarray approach was employed. We developed a thematic cDNA microarray containing clones representing high sequence similarities with known or predicted Arabidopsis genes implicated in different metabolic and signal transduction pathways. This 800-element cDNA array was used to compare the expression patterns in leaves and immature embryos (2 mm and 6 mm). Statistical analysis, using two-step ANOVA, revealed that 143 cDNA clones can be considered as differentially expressed. Of these, 62 clones were found to be up-regulated in leaves, 81 in embryos whereas only seven clones displayed increased level of mRNA in the 6 mm embryos when compared with 2 mm embryos. The differentially expressed clones are distributed among many metabolic and signal transduction pathways. For example, genes related to fatty acid metabolism and amino acid biosynthesis exhibited preferential expression patterns in immature embryos. Also, clones potentially encoding enzymes involved in the metabolism of ascorbate and aldarate, pyruvate, propanoate and inositol, and citrate cycle were found to be up-regulated in embryos. In contrast, cDNA clones putatively involved in energy metabolism were more abundant in leaves than embryos. Clones encoding potential signal transduction components including receptors, protein kinases, protein phosphatases, and transcription factors were also identified, with preferential expression profiles in immature embryos. The expression patterns derived from this study provide initial characterization of metabolic pathways and signalling transduction networks occurring in the early stage of sunflower seed development.

Optimizing microarray in experimental hypertension

BACKGROUND

Genetic noise between outbred animals can potentially be a major confounder in the use of microarray technology for gene expression profiling. The study of paired organs from the same animal offers an alternative approach (e.g., for studies of the kidney in experimental hypertension). The present study was undertaken to determine the level of genetic noise between outbred adult Sprague-Dawley (SD) rats, and to determine the effects of unilateral nephrectomy on changes in gene expression as a basis for the design of microarray studies in experimental hypertension.

METHODS

Male SD rats (approximately 130 g) were acclimatized before measurement of tail-cuff systolic blood pressure (SBP) for 6 control days and 4 days of saline treatment. Left kidney nephrectomy was performed, and the tissue snap-frozen in liquid nitrogen for subsequent RNA extraction. Two weeks later, SBP was measured over 4 control and 8 saline treatment days, and the remaining right kidney removed and frozen. Total RNA purification, preparation of cRNA, hybridization, and scanning of the Rat U34A Affymetrix arrays were performed, and data analyzed using MAS5 software Affymetrix Suite (v5), Bioconductor, as well as statistical methods motivated by relevant simulations.

RESULTS

Gene expression profiles in the left control kidney were extremely consistent across animals. The expression profiles of pairs of kidneys from the same animal were, however, more similar than those of kidneys from different animals. Nephrectomy had little effect on the gene expression profiles in the time frame examined.

CONCLUSION

Despite the outbred nature of the rats used in this study, they are useful for gene expression profiling comparisons. The use of paired organs from an individual animal ensures even further genetic identity, allowing determination of genes modified by the treatment of interest.

Use of Principal Component Analysis and theGE-Biplot for the Graphical Exploration of Gene Expression Data

This note is in response to Wouters et al. (2003, Biometrics 59, 1131-1139) who compared three methods for exploring gene expression data. Contrary to their summary that principal component analysis is not very informative, we show that it is possible to determine principal component analyses that are useful for exploratory analysis of microarray data. We also present another biplot representation, the GE-biplot (Gene Expression biplot), that is a useful method for exploring gene expression data with the major advantage of being able to aid interpretation of both the samples and the genes relative to each other.