Selection bias in gene extraction on the basis of microarray gene-expression data

In the context of cancer diagnosis and treatment, we consider the problem of constructing an accurate prediction rule on the basis of a relatively small number of tumor tissue samples of known type containing the expression data on very many (possibly thousands) genes. Recently, results have been presented in the literature suggesting that it is possible to construct a prediction rule from only a few genes such that it has a negligible prediction error rate. However, in these results the test error or the leave-one-out cross-validated error is calculated without allowance for the selection bias. There is no allowance because the rule is either tested on tissue samples that were used in the first instance to select the genes being used in the rule or because the cross-validation of the rule is not external to the selection process; that is, gene selection is not performed in training the rule at each stage of the cross-validation process. We describe how in practice the selection bias can be assessed and corrected for by either performing a cross-validation or applying the bootstrap external to the selection process. We recommend using 10-fold rather than leave-one-out cross-validation, and concerning the bootstrap, we suggest using the so-called. 632+ bootstrap error estimate designed to handle overfitted prediction rules. Using two published data sets, we demonstrate that when correction is made for the selection bias, the cross-validated error is no longer zero for a subset of only a few genes.

Coordinated plant defense responses in Arabidopsis revealed by microarray analysis

Disease resistance is associated with a plant defense response that involves an integrated set of signal transduction pathways. Changes in the expression patterns of 2.375 selected genes were examined simultaneously by cDNA microarray analysis in Arabidopsis thaliana after inoculation with an incompatible fungal pathogen Alternaria brassicicola or treatment with the defense-related signaling molecules salicylic acid (SA), methyl jasmonate (MJ), or ethylene, Substantial changes (up- and down-regulation) in the steady-state abundance of 705 mRNAs were observed in response to one or more of the treatments, including known and putative defense-related genes and 106 genes with no previously described function or homology, In leaf tissue inoculated with A. brassicicola, the abundance of 168 mRNAs was increased more than 2.5-fold, whereas that of 39 mRNAs was reduced. Similarly, the abundance of 192, 221, and 55 mRNAs was highly (>2.5-fold) increased after treatment with SA, MJ, and ethylene, respectively. Data analysis revealed a surprising level of coordinated defense responses, including 169 mRNAs regulated by multiple treatments/defense pathways. The largest number of genes coinduced (one of four induced genes) and corepressed was found after treatments with SA and MJ. In addition, 50% of the genes induced by ethylene treatment were also induced by MJ treatment. These results indicated the existence of a substantial network of regulatory interactions and coordination occurring during plant defense among the different defense signaling pathways, notably between the salicylate and jasmonate pathways that were previously thought to act in an antagonistic fashion.

The Role of Prostate Specific Membrane Antigen and Pepsinogen C Tissue Expression as an Adjunctive Method to Prostate Cancer Diagnosis

Purpose: The diagnosis of prostate cancer in men with persistently increased prostate specific antigen after a negative prostate biopsy has become a great challenge for urologists and pathologists. We analyzed the diagnostic value of 6 genes in the tissue of patients with prostate cancer. Materials and Methods: The study was comprised of 50 patients with localized disease who underwent radical prostatectomy. Gene selection was based on a previous microarray analysis. Among 4,147 genes with different expressions between 2 pools of patients 6 genes (PSMA, TMEFF2, GREB1, TH1L, IgH3 and PGC) were selected. These genes were tested for diagnostic value using the quantitative reverse transcription polymerase chain reaction method. Initially malignant tissue samples from 33 patients were analyzed and in the second part of the study we analyzed benign tissue samples from the other 17 patients with prostate cancer. The control group was comprised of tissue samples of patients with benign prostatic hyperplasia. Results: Analysis of malignant prostatic tissue demonstrated that prostate specific membrane antigen was over expressed (mean 9 times) and pepsinogen C was under expressed (mean 1.3 X 10(-4) times) in all cases compared to benign prostatic hyperplasia. The other 4 tested genes showed a variable expression pattern not allowing for differentiation between benign and malignant cases. When we tested these results in the benign prostate tissues from patients with cancer, pepsinogen C maintained the expression pattern. In terms of prostate specific membrane antigen, despite over expression in most cases (mean 12 times), 2 cases (12%) presented with under expression. Conclusions: Pepsinogen C tissue expression may constitute a powerful adjunctive method to prostate biopsy in the diagnosis of prostate cancer cases.

Erection induced by Tx2-6 toxin of Phoneutria nigriventer spider: Expression profile of genes in the nitric oxide pathway of penile tissue of mice

The peptides Tx2-5 and Tx2-6, isolated from the whole venom of ""armed-spider"" Phoneutria nigniventer venom, are directly linked with the induction of persistent and painful erection in the penis of mammals. The erection induced by Tx2-6 has been associated with the activation of nitric oxide synthases. There is a scarcity of studies focusing on the outcome of Tx2-6 at the molecular level, by this reason we evaluated the gene profile activity of this toxin at the nitric oxide (NO) pathway. After microarray analyses on cavernous tissue of mice inoculated with Tx2-6 we found that only 10.4% (10/96) of these genes were differentially expressed, showing a limited effect of the toxin on the NO pathway. We found the genes sparc, ednrb, junb, cdkn1a, bcl2, ccl5, abcc1 over-expressed and the genes sod1, s100a10 and fth1 under-expressed after inoculation of Tx2-6. The differential expressions of sparc and ednrb genes were further confirmed using real-time PCR. Interestingly, ednrb activates the L-arginine/NO/cGMP pathway that is involved in the relaxation of the cavernous body. Therefore the priapism induced by Tx2-6 is a consequence of a highly specific interference of this neurotoxin with the NO pathway. (C) 2009 Elsevier Ltd. All rights reserved.

A mixture model-based approach to the clustering of microarray expression data

Motivation: This paper introduces the software EMMIX-GENE that has been developed for the specific purpose of a model-based approach to the clustering of microarray expression data, in particular, of tissue samples on a very large number of genes. The latter is a nonstandard problem in parametric cluster analysis because the dimension of the feature space (the number of genes) is typically much greater than the number of tissues. A feasible approach is provided by first selecting a subset of the genes relevant for the clustering of the tissue samples by fitting mixtures of t distributions to rank the genes in order of increasing size of the likelihood ratio statistic for the test of one versus two components in the mixture model. The imposition of a threshold on the likelihood ratio statistic used in conjunction with a threshold on the size of a cluster allows the selection of a relevant set of genes. However, even this reduced set of genes will usually be too large for a normal mixture model to be fitted directly to the tissues, and so the use of mixtures of factor analyzers is exploited to reduce effectively the dimension of the feature space of genes. Results: The usefulness of the EMMIX-GENE approach for the clustering of tissue samples is demonstrated on two well-known data sets on colon and leukaemia tissues. For both data sets, relevant subsets of the genes are able to be selected that reveal interesting clusterings of the tissues that are either consistent with the external classification of the tissues or with background and biological knowledge of these sets.

Protein-Binding Microarray Analysis of Tumor Suppressor AP2α Target Gene Specificity.

Cheap and massively parallel methods to assess the DNA-binding specificity of transcription factors are actively sought, given their prominent regulatory role in cellular processes and diseases. Here we evaluated the use of protein-binding microarrays (PBM) to probe the association of the tumor suppressor AP2α with 6000 human genomic DNA regulatory sequences. We show that the PBM provides accurate relative binding affinities when compared to quantitative surface plasmon resonance assays. A PBM-based study of human healthy and breast tumor tissue extracts allowed the identification of previously unknown AP2α target genes and it revealed genes whose direct or indirect interactions with AP2α are affected in the diseased tissues. AP2α binding and regulation was confirmed experimentally in human carcinoma cells for novel target genes involved in tumor progression and resistance to chemotherapeutics, providing a molecular interpretation of AP2α role in cancer chemoresistance. Overall, we conclude that this approach provides quantitative and accurate assays of the specificity and activity of tumor suppressor and oncogenic proteins in clinical samples, interfacing genomic and proteomic assays.

“Unmixing” Tissue Gene Expression Signatures from Tumor Biopsies

Emergent molecular measurement methods, such as DNA microarray, qRTPCR, andmany others, offer tremendous promise for the personalized treatment of cancer. Thesetechnologies measure the amount of specific proteins, RNA, DNA or other moleculartargets from tumor specimens with the goal of “fingerprinting” individual cancers. Tumorspecimens are heterogeneous; an individual specimen typically contains unknownamounts of multiple tissues types. Thus, the measured molecular concentrations resultfrom an unknown mixture of tissue types, and must be normalized to account for thecomposition of the mixture.For example, a breast tumor biopsy may contain normal, dysplastic and cancerousepithelial cells, as well as stromal components (fatty and connective tissue) and bloodand lymphatic vessels. Our diagnostic interest focuses solely on the dysplastic andcancerous epithelial cells. The remaining tissue components serve to “contaminate”the signal of interest. The proportion of each of the tissue components changes asa function of patient characteristics (e.g., age), and varies spatially across the tumorregion. Because each of the tissue components produces a different molecular signature,and the amount of each tissue type is specimen dependent, we must estimate the tissuecomposition of the specimen, and adjust the molecular signal for this composition.Using the idea of a chemical mass balance, we consider the total measured concentrationsto be a weighted sum of the individual tissue signatures, where weightsare determined by the relative amounts of the different tissue types. We develop acompositional source apportionment model to estimate the relative amounts of tissuecomponents in a tumor specimen. We then use these estimates to infer the tissuespecificconcentrations of key molecular targets for sub-typing individual tumors. Weanticipate these specific measurements will greatly improve our ability to discriminatebetween different classes of tumors, and allow more precise matching of each patient tothe appropriate treatment

Histological and histochemical evaluation of human oral mucosa constructs developed by tissue engineering

Reconstruction of large oral mucosa defects is often challenging, since the shortage of healthy oral mucosa to replace the excised tissues is very common. In this context, tissue engineering techniques may provide a source of autologous tissues available for transplant in these patients. In this work, we developed a new model of artificial oral mucosa generated by tissue engineering using a fibrin-agarose scaffold. For that purpose, we generated primary cultures of human oral mucosa fibroblasts and keratinocytes from small biopsies of normal oral mucosa using enzymatic treatments. Then we determined the viability of the cultured cells by electron probe quantitative X-ray microanalysis, and we demonstrated that most of the cells in the primary cultures were alive and had high K/Na ratios. Once cell viability was determined, we used the cultured fibroblasts and keratinocytes to develop an artificial oral mucosa construct by using a fibrin-agarose extracellular matrix and a sequential culture technique using porous culture inserts. Histological analysis of the artificial tissues showed high similarities with normal oral mucosa controls. The epithelium of the oral substitutes had several layers, with desmosomes and apical microvilli and microplicae. Both the controls and the oral mucosa substitutes showed high suprabasal expression of cytokeratin 13 and low expression of cytokeratin 10. All these results suggest that our model of oral mucosa using fibrin-agarose scaffolds show several similarities with native human oral mucosa.

Test of four colon cancer risk-scores in formalin fixed paraffin embedded microarray gene expression data.

BACKGROUND: Prognosis prediction for resected primary colon cancer is based on the T-stage Node Metastasis (TNM) staging system. We investigated if four well-documented gene expression risk scores can improve patient stratification. METHODS: Microarray-based versions of risk-scores were applied to a large independent cohort of 688 stage II/III tumors from the PETACC-3 trial. Prognostic value for relapse-free survival (RFS), survival after relapse (SAR), and overall survival (OS) was assessed by regression analysis. To assess improvement over a reference, prognostic model was assessed with the area under curve (AUC) of receiver operating characteristic (ROC) curves. All statistical tests were two-sided, except the AUC increase. RESULTS: All four risk scores (RSs) showed a statistically significant association (single-test, P < .0167) with OS or RFS in univariate models, but with HRs below 1.38 per interquartile range. Three scores were predictors of shorter RFS, one of shorter SAR. Each RS could only marginally improve an RFS or OS model with the known factors T-stage, N-stage, and microsatellite instability (MSI) status (AUC gains < 0.025 units). The pairwise interscore discordance was never high (maximal Spearman correlation = 0.563) A combined score showed a trend to higher prognostic value and higher AUC increase for OS (HR = 1.74, 95% confidence interval [CI] = 1.44 to 2.10, P < .001, AUC from 0.6918 to 0.7321) and RFS (HR = 1.56, 95% CI = 1.33 to 1.84, P < .001, AUC from 0.6723 to 0.6945) than any single score. CONCLUSIONS: The four tested gene expression-based risk scores provide prognostic information but contribute only marginally to improving models based on established risk factors. A combination of the risk scores might provide more robust information. Predictors of RFS and SAR might need to be different.

Integrative molecular bioinformatics study of human adrenocortical tumors : microRNA, tissue-specific target prediction, and pathway analysis.

MicroRNAs (miRs) are involved in the pathogenesis of several neoplasms; however, there are no data on their expression patterns and possible roles in adrenocortical tumors. Our objective was to study adrenocortical tumors by an integrative bioinformatics analysis involving miR and transcriptomics profiling, pathway analysis, and a novel, tissue-specific miR target prediction approach. Thirty-six tissue samples including normal adrenocortical tissues, benign adenomas, and adrenocortical carcinomas (ACC) were studied by simultaneous miR and mRNA profiling. A novel data-processing software was used to identify all predicted miR-mRNA interactions retrieved from PicTar, TargetScan, and miRBase. Tissue-specific target prediction was achieved by filtering out mRNAs with undetectable expression and searching for mRNA targets with inverse expression alterations as their regulatory miRs. Target sets and significant microarray data were subjected to Ingenuity Pathway Analysis. Six miRs with significantly different expression were found. miR-184 and miR-503 showed significantly higher, whereas miR-511 and miR-214 showed significantly lower expression in ACCs than in other groups. Expression of miR-210 was significantly lower in cortisol-secreting adenomas than in ACCs. By calculating the difference between dCT(miR-511) and dCT(miR-503) (delta cycle threshold), ACCs could be distinguished from benign adenomas with high sensitivity and specificity. Pathway analysis revealed the possible involvement of G2/M checkpoint damage in ACC pathogenesis. To our knowledge, this is the first report describing miR expression patterns and pathway analysis in sporadic adrenocortical tumors. miR biomarkers may be helpful for the diagnosis of adrenocortical malignancy. This tissue-specific target prediction approach may be used in other tumors too.

Tissue-based immune monitoring I: tumor core needle biopsies allow in-depth interrogation of the tumor microenvironment.

We sought to assess the feasibility and reproducibility of performing tissue-based immune characterization of the tumor microenvironment using CT-compatible needle biopsy material. Three independent biopsies were obtained intraoperatively from one metastatic epithelial ovarian cancer lesion of 7 consecutive patients undergoing surgical cytoreduction using a 16-gauge core biopsy needle. Core specimens were snap-frozen and subjected to immunohistochemistry (IHC) against human CD3, CD4, CD8, and FoxP3. A portion of the cores was used to isolate RNA for 1) real-time quantitative (q)PCR for CD3, CD4, CD8, FoxP3, IL-10 and TGF-beta, 2) multiplexed PCR-based T cell receptor (TCR) CDR3 Vβ region spectratyping, and 3) gene expression profiling. Pearson's correlations were examined for immunohistochemistry and PCR gene expression, as well as for gene expression array data obtained from different tumor biopsies. Needle biopsy yielded sufficient tissue for all assays in all patients. IHC was highly reproducible and informative. Significant correlations were seen between the frequency of CD3+, CD8+ and FoxP3+ T cells by IHC with CD3ε, CD8A, and FoxP3 gene expression, respectively, by qPCR (r=0.61, 0.86, and 0.89; all p< 0.05). CDR3 spectratyping was feasible and highly reproducible in each tumor, and indicated a restricted repertoire for specific TCR Vβ chains in tumor-infiltrating T cells. Microarray gene expression revealed strong correlation between different biopsies collected from the same tumor. Our results demonstrate a feasible and reproducible method of immune monitoring using CT-compatible needle biopsies from tumor tissue, thereby paving the way for sophisticated translational studies during tumor biological therapy.

A conceptual and practical overview of cDNA microarray technology: implications for basic and clinical sciences

cDNA microarray is an innovative technology that facilitates the analysis of the expression of thousands of genes simultaneously. The utilization of this methodology, which is rapidly evolving, requires a combination of expertise from the biological, mathematical and statistical sciences. In this review, we attempt to provide an overview of the principles of cDNA microarray technology, the practical concerns of the analytical processing of the data obtained, the correlation of this methodology with other data analysis methods such as immunohistochemistry in tissue microarrays, and the cDNA microarray application in distinct areas of the basic and clinical sciences.

Impact of (pro)renin receptor deficiency in adipose tissue using a genetically engineered mouse model

La stimulation du récepteur de la rénine/prorénine [(P) RR], un membre récemment découvert du système rénine-angiotensine (SRA), augmente l'activité du SRA et des voies de signalisation angiotensine II-indépendante. Pour étudier l'impact potentiel du (P)RR dans le développement de l`obésité, nous avons émis l'hypothèse que les souris déficientes en (P)RR uniquement dans le tissus adipeux (KO) auront une diminution du poids corporel en ciblant le métabolisme du tissu adipeux, l'activité locomoteur et/ou la prise alimentaire. Ainsi, des souris KO ont été générées en utilisant la technologie Cre/Lox. Le gain de poids et la prise alimentaire ont été évalués hebdomadairement dans les mâles et femelles KO et de type sauvage (WT) pendant 4 semaines alors qu’ils étaient maintenu sur une diète normal. De plus, un groupe de femelles a été placé pour 6 semaines sur une diète riche en gras et en glucides (HF/HC). La composition corporelle et l'activité ambulatoire ont été évaluées par l’EchoMRI et à l’aide de cages Physioscan, respectivement. Les tissus adipeux ont été prélevés et pesés. De plus, les gras péri-gonadaux ont été utilisés pour le microarray. Finalement, le niveaux d'expression d'ARNm du (P)RR ont été évalués. Comme le gène du (P)RR est situé sur le chromosome X, les mâles étaient des KOs complets et les femelles étaient des KOs partielles. Les souris KO avaient un poids corporel significativement plus petit par rapport à WT, les différences étant plus prononcées chez les mâles. De plus, les femelles KOs étaient résistantes à l'obésité lorsqu'elles ont été placées sur la diète HF/HC et donc elles avaient significativement moins de masse grasse par rapport aux WTs. L’analyse histologique des gras péri-gonadaux des KOs nous ont dévoilés qu’il avait une réduction du nombre d'adipocytes mais de plus grande taille. Bien qu'il n'y ait eu aucun changement dans la consommation alimentaire, une augmentation de près de 3 fois de l'activité ambulatoire a été détectée chez les mâles. De plus, nous avons observé que leurs tibias étaient de longueur réduite ce qui suggère fortement l'affection de leur développement. Les gras péri-gonadaux des souris KO avaient une expression réduite de l`ABLIM2 (Actin binding LIM protein family, member 2) qui est associé avec le diabète de type II chez l'humain. Ainsi, les données recueillies suggèrent fortement que le (P)RR est impliquée dans la régulation du poids corporelle.

“Unmixing” Tissue Gene Expression Signatures from Tumor Biopsies

Emergent molecular measurement methods, such as DNA microarray, qRTPCR, and many others, offer tremendous promise for the personalized treatment of cancer. These technologies measure the amount of specific proteins, RNA, DNA or other molecular targets from tumor specimens with the goal of “fingerprinting” individual cancers. Tumor specimens are heterogeneous; an individual specimen typically contains unknown amounts of multiple tissues types. Thus, the measured molecular concentrations result from an unknown mixture of tissue types, and must be normalized to account for the composition of the mixture. For example, a breast tumor biopsy may contain normal, dysplastic and cancerous epithelial cells, as well as stromal components (fatty and connective tissue) and blood and lymphatic vessels. Our diagnostic interest focuses solely on the dysplastic and cancerous epithelial cells. The remaining tissue components serve to “contaminate” the signal of interest. The proportion of each of the tissue components changes as a function of patient characteristics (e.g., age), and varies spatially across the tumor region. Because each of the tissue components produces a different molecular signature, and the amount of each tissue type is specimen dependent, we must estimate the tissue composition of the specimen, and adjust the molecular signal for this composition. Using the idea of a chemical mass balance, we consider the total measured concentrations to be a weighted sum of the individual tissue signatures, where weights are determined by the relative amounts of the different tissue types. We develop a compositional source apportionment model to estimate the relative amounts of tissue components in a tumor specimen. We then use these estimates to infer the tissuespecific concentrations of key molecular targets for sub-typing individual tumors. We anticipate these specific measurements will greatly improve our ability to discriminate between different classes of tumors, and allow more precise matching of each patient to the appropriate treatment

Early Changes in Gene Expression Induced by Tobacco Smoke: Evidence for the Importance of Estrogen within Lung Tissue

Lung cancer is the leading cause of cancer deaths in the United States, surpassing breast cancer as the primary cause of cancer-related mortality in women. The goal of the present study was to identify early molecular changes in the lung induced by exposure to tobacco smoke and thus identify potential targets for chemoprevention. Female A/J mice were exposed to either tobacco smoke or HEPA-filtered air via a whole-body exposure chamber (6 h/d, 5 d/wk for 3, 8, and 20 weeks). Gene expression profiles of lung tissue from control and smoke-exposed animals were established using a 15K cDNA microarray. Cytochrome P450 1b1, a phase I enzyme involved in both the metabolism of xenobiotics and the 4-hydroxylation of 17 beta-estradiol (E(2)), was modulated to the greatest extent following smoke exposure. A panel of 10 genes were found to be differentially expressed in control and smoke-exposed lung tissues at 3, 8, and 20 weeks (P < 0.001). The interaction network of these differentially expressed genes revealed new pathways modulated by short-term smoke exposure, including estrogen metabolism. In addition, E(2) was detected within murine lung tissue by gas chromatography-coupled mass spectrometry and immunohistochemistry. Identification of the early molecular events that contribute to lung tumor formation is anticipated to lead to the development of promising targeted chemopreventive therapies. In conclusion, the presence of E2 within lung tissue when combined with the modulation of cytochrome P450 1b1 and other estrogen metabolism genes by tobacco smoke provides novel insight into a possible role for estrogens in lung cancer. Cancer Prev Res; 3(6); 707-17. (C) 2010 AACR.