Genome-wide estimation of gender differences in the gene expression of human livers: statistical design and analysis

Cluster analysis on high dimensional RNA-seq data with applications to cancer research - An evaluation study

BACKGROUND

Clustering of gene expression data is widely used to identify novel subtypes of cancer. Plenty of clustering approaches have been proposed, but there is a lack of knowledge regarding their relative merits and how data characteristics influence the performance. We evaluate how cluster analysis choices affect the performance by studying four publicly available human cancer data sets: breast, brain, kidney and stomach cancer. In particular, we focus on how the sample size, distribution of subtypes and sample heterogeneity affect the performance.

RESULTS

In general, increasing the sample size had limited effect on the clustering performance, e.g. for the breast cancer data similar performance was obtained for n = 40 as for n = 330. The relative distribution of the subtypes had a noticeable effect on the ability to identify the disease subtypes and data with disproportionate cluster sizes turned out to be difficult to cluster. Both the choice of clustering method and selection method affected the ability to identify the subtypes, but the relative performance varied between data sets, making it difficult to rank the approaches. For some data sets, the performance was substantially higher when the clustering was based on data from only one sex compared to data from a mixed population. This suggests that homogeneous data are easier to cluster than heterogeneous data and that clustering males and females individually may be beneficial and increase the chance to detect novel subtypes. It was also observed that the performance often differed substantially between females and males.

CONCLUSIONS

The number of samples seems to have a limited effect on the performance while the heterogeneity, at least with respect to sex, is important for the performance. Hence, by analyzing the genders separately, the possible loss caused by having fewer samples could be outweighed by the benefit of a more homogeneous data.

Identification and characterization of sexual dimorphism‑linked gene expression profile in hepatocellular carcinoma

Hepatocellular carcinoma (HCC) is sexually disparate in humans, with a significantly increased prevalence in males. The molecular mechanisms by which the inhibition or development of liver cancer are facilitated require further investigation with regard to sex factors affecting disease progression. In the present study, functional signatures of differentially expressed genes (DEGs) were screened in female and male tumors via bioinformatics analysis. The following gene chip expression profiles were downloaded from the Gene Expression Omnibus: GSE19665, GSE23342 and GSE9843. They comprised cancerous and non-cancerous tissue from patients with HCC and included critical sex features. Further evaluation of selected DEGs in the two sexual groups was performed via hierarchical clustering analysis. Venn diagram and functional protein-protein interaction (PPI) network analyses were performed. Survival analysis of patients with differences in gene expression levels was subsequently performed using the Kaplan-Meier Plotter database. Certain identified DEGs were common in female and male tumor samples, whereas others exhibited a sexually-biased expression profile. Gene Ontology revealed that the cell cycle module ‘biological process’ was enriched in tumors derived from both sexes, whereas the metabolic pathways and drug metabolism modules were only significantly enriched in cancer tissues from male subjects. A number of hub DEGs in the cell cycle and p53 signaling pathways were involved in significant protein-protein interaction (PPI) modules, including CDK1 and CCNB1. These DEGs were upregulated in tumors derived from female subjects compared with those derived from male subjects, and could be used as markers of poor prognosis in male patients. Other genes, such as CYP3A4 and SERPINA4, were identified in metabolic pathways, and were downregulated in male compared with female subjects. These genes were associated with a decreased survival rate. The data demonstrated that sex differences in physiology may regulate the levels of gene expression and/or activity, including gene function associated with oncogenesis and the outcomes of liver cancer. Additional surveys are required to explore in detail the molecular mechanisms underlying the differences in gene expression between the two sexes during the development of liver cancer.

Transcriptome profiling of peripheral blood cells identifies potential biomarkers for doxorubicin cardiotoxicity in a rat model

Aims

Doxorubicin (DOX), a widely used anticancer agent, can cause an unpredictable cardiac toxicity which remains a major limitation in cancer chemotherapy. There is a need for noninvasive, sensitive and specific biomarkers which will allow identifying patients at risk for DOX-induced cardiotoxicity to prevent permanent cardiac damage. The aim of this study was to investigate whether the expression of specific genes in the peripheral blood can be used as surrogate marker(s) for DOX-induced cardiotoxicity.

Methods/Results

Rats were treated with a single dose of DOX similar to one single dose that is often administered in humans. The cardiac and peripheral blood mononuclear cells (PBMCs) genome-wide expression profiling were examined using Illumina microarrays. The results showed 4,409 differentially regulated genes (DRG) in the hearts and 4,120 DRG in PBMC. Of these 2411 genes were similarly DRG (SDRG) in both the heart and PBMC. Pathway analysis of the three datasets of DRG using Gene Ontology (GO) enrichment analysis and Ingenuity Pathways Analysis (IPA) showed that most of the genes in these datasets fell into pathways related to oxidative stress response and protein ubiquination. IPA search for potential eligible biomarkers for cardiovascular disease within the SDRG list revealed 188 molecules.

Conclusions

We report the first in-depth comparison of DOX-induced global gene expression profiles of hearts and PBMCs. The high similarity between the gene expression profiles of the heart and PBMC induced by DOX indicates that the PBMC transcriptome may serve as a surrogate marker of DOX-induced cardiotoxicity. Future directions of this research will include analysis of PBMC expression profiles of cancer patients treated with DOX-based chemotherapy to identify the cardiotoxicity risk, predict DOX-treatment response and ultimately to allow individualized anti-cancer therapy.

The facial evolution: looking backward and moving forward

Three-dimensional (3D) facial analysis is ideal for high-resolution, nonionizing, noninvasive objective, high-throughput phenotypic, and phenomic studies. It is a natural complement to (epi)genetic technologies to facilitate advances in the understanding of rare and common diseases. The face is uniquely reflective of the primordial tissues, and there is evidence supporting the application of 3D facial analysis to the investigation of variation and disease including studies showing that the face can reflect systemic health, provides diagnostic clues to disorders, and that facial variation reflects biological pathways. In addition, facial variation has been related to evolutionary factors. The purpose of this review is to look backward to suggest that knowledge of human evolution supports, and may instruct, the application and interpretation of studies of facial morphology for documentation of human variation and investigation of its relationships with health and disease. Furthermore, in the context of advances of deep phenotyping and data integration, to look forward to suggest approaches to scalable implementation of facial analysis, and to suggest avenues for future research and clinical application of this technology.

Comparative hepatic proteome analysis between lean and obese rats fed a high-fat diet reveals the existence of gender differences

Gender differences in obesity stem from metabolic and hormonal differences between sexes and contribute to differences between women and men in health risks attributable to obesity. We hypothesized that liver may be an ideal target for the evaluation of gender differences in obesity development in response to a high-fat diet (HFD). Therefore, to test this hypothesis, we performed a global proteome analysis in the liver of lean and obese rats of both genders who were fed an HFD through 2-DE combined with MALDI-TOF-MS. When rats were exposed to HFD, male rats gained more body weight with increased values of plasma biochemical parameters than female rats. Image analysis and further statistical analysis of a 2-DE protein map allowed for the detection and identification of 34 proteins that were significantly modulated in a gender-dependent manner. We found 19 proteins showing identical gender-different regulation in both normal diet (ND) and HFD. Five proteins also showed clear gender differences in both ND and HFD; however, their regulation modes in HFD were opposite to those in ND. Of particular interest, 10 proteins showed gender differences only in either ND or HFD rats. Present proteomic insight into gender-dimorphic protein modulation in liver would aid in the improvement of gender awareness in the health-care system and in implementation of evidence-based gender-specific clinical recommendations.

Transcriptional profiling of human liver identifies sex-biased genes associated with polygenic dyslipidemia and coronary artery disease

Sex-differences in human liver gene expression were characterized on a genome-wide scale using a large liver sample collection, allowing for detection of small expression differences with high statistical power. 1,249 sex-biased genes were identified, 70% showing higher expression in females. Chromosomal bias was apparent, with female-biased genes enriched on chrX and male-biased genes enriched on chrY and chr19, where 11 male-biased zinc-finger KRAB-repressor domain genes are distributed in six clusters. Top biological functions and diseases significantly enriched in sex-biased genes include transcription, chromatin organization and modification, sexual reproduction, lipid metabolism and cardiovascular disease. Notably, sex-biased genes are enriched at loci associated with polygenic dyslipidemia and coronary artery disease in genome-wide association studies. Moreover, of the 8 sex-biased genes at these loci, 4 have been directly linked to monogenic disorders of lipid metabolism and show an expression profile in females (elevated expression of ABCA1, APOA5 and LDLR; reduced expression of LIPC) that is consistent with the lower female risk of coronary artery disease. Female-biased expression was also observed for CYP7A1, which is activated by drugs used to treat hypercholesterolemia. Several sex-biased drug-metabolizing enzyme genes were identified, including members of the CYP, UGT, GPX and ALDH families. Half of 879 mouse orthologs, including many genes of lipid metabolism and homeostasis, show growth hormone-regulated sex-biased expression in mouse liver, suggesting growth hormone might play a similar regulatory role in human liver. Finally, the evolutionary rate of protein coding regions for human-mouse orthologs, revealed by dN/dS ratio, is significantly higher for genes showing the same sex-bias in both species than for non-sex-biased genes. These findings establish that human hepatic sex differences are widespread and affect diverse cell metabolic processes, and may help explain sex differences in lipid profiles associated with sex differential risk of coronary artery disease.

Designing Toxicogenomics Studies that use DNA Array Technology

Background

Bioassays are routinely used to evaluate the toxicity of test agents. Experimental designs for bioassays are largely encompassed by fixed effects linear models. In toxicogenomics studies where DNA arrays measure mRNA levels, the tissue samples are typically generated in a bioassay. These measurements introduce additional sources of variation, which must be properly managed to obtain valid tests of treatment effects.

Results

An analysis of covariance model is developed which combines a fixed-effects linear model for the bioassay with important variance components associated with DNA array measurements. These models can accommodate the dominant characteristics of measurements from DNA arrays, and they account for technical variation associated with normalization, spots, dyes, and batches as well as the biological variation associated with the bioassay. An example illustrates how the model is used to identify valid designs and to compare competing designs.

Conclusions

Many toxicogenomics studies are bioassays which measure gene expression using DNA arrays. These studies can be designed and analyzed using standard methods with a few modifications to account for characteristics of array measurements, such as multiple endpoints and normalization. As much as possible, technical variation associated with probes, dyes, and batches are managed by blocking treatments within these sources of variation. An example shows how some practical constraints can be accommodated by this modelling and how it allows one to objectively compare competing designs.

Designing toxicogenomics studies that use DNA array technology

BACKGROUND

Bioassays are routinely used to evaluate the toxicity of test agents. Experimental designs for bioassays are largely encompassed by fixed effects linear models. In toxicogenomics studies where DNA arrays measure mRNA levels, the tissue samples are typically generated in a bioassay. These measurements introduce additional sources of variation, which must be properly managed to obtain valid tests of treatment effects.

RESULTS

An analysis of covariance model is developed which combines a fixed-effects linear model for the bioassay with important variance components associated with DNA array measurements. These models can accommodate the dominant characteristics of measurements from DNA arrays, and they account for technical variation associated with normalization, spots, dyes, and batches as well as the biological variation associated with the bioassay. An example illustrates how the model is used to identify valid designs and to compare competing designs.

CONCLUSIONS

Many toxicogenomics studies are bioassays which measure gene expression using DNA arrays. These studies can be designed and analyzed using standard methods with a few modifications to account for characteristics of array measurements, such as multiple endpoints and normalization. As much as possible, technical variation associated with probes, dyes, and batches are managed by blocking treatments within these sources of variation. An example shows how some practical constraints can be accommodated by this modelling and how it allows one to objectively compare competing designs.

Gender-specific effects of cytokine gene polymorphisms on childhood vaccine responses

Cytokine gene polymorphisms affect vaccine responses and gender-specific effects are known for many phenotypes. Therefore, this study investigated gender-specific effects of cytokine gene polymorphisms on vaccine responses. In 263 2-year-old subjects selected for parental history of atopy, boys with IL-4 C-589T and IL-4Ralpha I50V genotypes associated with atopy had increased Diptheria Toxoid (DiphTox) and Tetanus Toxoid (TetTox) responses compared with the remaining alleles (IL-4 C-589T: DipTox p=0.01, TetTox p=0.04; IL-4Ralpha.I50V: DipTox p=0.04, TetTox p=0.08). Contrastingly, girls with IL-10 -592C genotypes associated with atopy had lower levels of DiphTox (p=0.03) and TetTox (p=0.02) responses compared with the remaining allele. Additionally, interaction effects were found for IL-4 C-589T (p=0.01) and IL-4Ralpha I50V (p=0.04) polymorphisms. In conclusion, these findings support the interaction of primary genetic and modifying factors on vaccine responses and the importance of atopic genetics to these responses.

The role of genomics and proteomics: technologies in studying non-alcoholic fatty liver disease

Non-alcoholic fatty liver disease (NAFLD) and non-alcoholic steatohepatitis (NASH) are examples of complex diseases accompanied by changes in the expression of thousands of genes and a plethora of proteins encoded by these genes. Before the era of high-throughput analysis, typical translational research initiatives, aimed at defining the molecular targets for complex diseases, were performed on gene-by-gene basis. Innovative technologies, such as expression microarrays, mass spectromety, and reverse proteomics, now allow investigators to reveal complex patterns of the expression of biologically active molecules. For this reason, high-throughput approaches may be well suited for studies designed to untangle the molecular basis of the chronic liver diseases such as NAFLD.

A threshold neurotoxic amphetamine exposure inhibits parietal cortex expression of synaptic plasticity-related genes

Compulsive drug abuse has been conceptualized as a behavioral state where behavioral stimuli override normal decision making. Clinical studies of methamphetamine users have detailed decision making changes and imaging studies have found altered metabolism and activation in the parietal cortex. To examine the molecular effects of amphetamine (AMPH) on the parietal cortex, gene expression responses to amphetamine challenge (7.5 mg/kg) were examined in the parietal cortex of rats pretreated for nine days with either saline, non-neurotoxic amphetamine, or neurotoxic AMPH dosing regimens. The neurotoxic AMPH exposure [three doses of 7.5 mg/kg/day AMPH (6 h between doses), for nine days] produced histological signs of neurotoxicity in the parietal cortex while a non-neurotoxic dosing regimen (2.0 mg/kg/day x 3) did not. Neurotoxic AMPH pretreatment resulted in significantly diminished AMPH challenge-induced mRNA increases of activity-regulated cytoskeletal protein (ARC), nerve growth-factor inducible protein A (NGFI-A), and nerve growth-factor inducible protein B (NGFI-B) in the parietal cortex while neither saline pretreatment nor non-neurotoxic AMPH pretreatment did. This effect was specific to these genes as tissue plasminogen activator (t-PA), neuropeptide Y (NPY) and c-jun expression in response to AMPH challenge was unaltered or enhanced by amphetamine pretreatments. In the striatum, there were no differences between saline, neurotoxic AMPH, and non-neurotoxic AMPH pretreatments on ARC, NGFI-A or NGFI-B expression elicited by the AMPH challenge. These data indicate that the responsiveness of synaptic plasticity-related genes is sensitive to disruption specifically in the parietal cortex by threshold neurotoxic AMPH exposures.

A method for computing the overall statistical significance of a treatment effect among a group of genes

BACKGROUND

In studies that use DNA arrays to assess changes in gene expression, our goal is to evaluate the statistical significance of treatments on sets of genes. Genes can be grouped by a molecular function, a biological process, or a cellular component, e.g., gene ontology (GO) terms. The meaning of an affected GO group is often clearer than interpretations arising from a list of the statistically significant genes.

RESULTS

Computer simulations demonstrated that correlations among genes invalidate many statistical methods that are commonly used to assign significance to GO terms. Ignoring these correlations overstates the statistical significance. Meta-analysis methods for combining p-values were modified to adjust for correlation. One of these methods is elaborated in the context of a comparison between two treatments. The form of the correlation adjustment depends upon the alternative hypothesis.

CONCLUSION

Reliable corrections for the effect of correlations among genes on the significance level of a GO term can be constructed for an alternative hypothesis where all transcripts in the GO term increase (decrease) in response to treatment. For general alternatives, which allow some transcripts to increase and others to decrease, the bias of naïve significance calculations can be greatly decreased although not eliminated.

Proceedings of the Second Annual Conference of the MidSouth Computational Biology and Bioinformatics Society

The MCBIOS 2004 conference brought together regional researchers and students in biology, computer science and bioinformatics on October 7th-9th 2004 to present their latest work. This editorial describes the conference itself and introduces the twelve peer-reviewed manuscripts accepted for publication in the Proceedings of the MCBIOS 2004 Conference. These manuscripts included new methods for analysis of high-throughput gene expression experiments, EST clustering, analysis of mass spectrometry data and genomic analysis