An efficient and robust statistical modeling approach to discover differentially expressed genes using genomic expression profiles

Tumor-specific genetic expression profile of metastatic oral squamous cell carcinoma

BACKGROUND

Metastasis is the most important predictor of survival in patients with oral squamous cell carcinoma (OSCC). We tested the hypothesis that there is a genetic expression profile associated with OSCC metastasis.

METHODS

We obtained samples from 6 OSCC node-positive primary tumors and their matched metastatic lymph nodes, and 5 OSCC node-negative primary tumors. Using laser capture microdissection, we isolated OSCC cells from metastatic lymph nodes and compared them with those from matched primary tumors and unmatched node-negative primary tumors using Affymetrix Human Genome Focus arrays.

RESULTS

Comparison of tumor cells from the lymph nodes with those from the unmatched, node-negative primary tumors revealed differential expression of 160 genes. Hierarchical clustering and principal component analysis using this 160-gene set showed that the node-negative samples were distinguishable from both, node-positive primary tumors and tumors in the lymph nodes. Many of the expression changes found in the metastatic cells from the lymph nodes were also found in the node-positive primary tumors. Immunohistochemical analysis for transglutaminase-3 and keratin 16 confirmed the differential genetic expression for these genes.

CONCLUSION

These preliminary results suggest that there may be a metastatic gene expression profile present in node-positive primary OSCC.

A review of feature selection techniques in bioinformatics

Feature selection techniques have become an apparent need in many bioinformatics applications. In addition to the large pool of techniques that have already been developed in the machine learning and data mining fields, specific applications in bioinformatics have led to a wealth of newly proposed techniques. In this article, we make the interested reader aware of the possibilities of feature selection, providing a basic taxonomy of feature selection techniques, and discussing their use, variety and potential in a number of both common as well as upcoming bioinformatics applications.

A comprehensive evaluation of SAM, the SAM R-package and a simple modification to improve its performance

Background

The Significance Analysis of Microarrays (SAM) is a popular method for detecting significantly expressed genes and controlling the false discovery rate (FDR). Recently, it has been reported in the literature that the FDR is not well controlled by SAM. Due to the vast application of SAM in microarray data analysis, it is of great importance to have an extensive evaluation of SAM and its associated R-package (sam2.20).

Results

Our study has identified several discrepancies between SAM and sam2.20. One major difference is that SAM and sam2.20 use different methods for estimating FDR. Such discrepancies may cause confusion among the researchers who are using SAM or are developing the SAM-like methods. We have also shown that SAM provides no meaningful estimates of FDR and this problem has been corrected in sam2.20 by using a different formula for estimating FDR. However, we have found that, even with the improvement sam2.20 has made over SAM, sam2.20 may still produce erroneous and even conflicting results under certain situations. Using an example, we show that the problem of sam2.20 is caused by its use of asymmetric cutoffs which are due to the large variability of null scores at both ends of the order statistics. An obvious approach without the complication of the order statistics is the conventional symmetric cutoff method. For this reason, we have carried out extensive simulations to compare the performance of sam2.20 and the symmetric cutoff method. Finally, a simple modification is proposed to improve the FDR estimation of sam2.20 and the symmetric cutoff method.

Conclusion

Our study shows that the most serious drawback of SAM is its poor estimation of FDR. Although this drawback has been corrected in sam2.20, the control of FDR by sam2.20 is still not satisfactory. The comparison between sam2.20 and the symmetric cutoff method reveals that the relative performance of sam2.20 to the symmetric cutff method depends on the ratio of induced to repressed genes in a microarray data, and is also affected by the ratio of DE to EE genes and the distributions of induced and repressed genes. Numerical simulations show that the symmetric cutoff method has the biggest advantage over sam2.20 when there are equal number of induced and repressed genes (i.e., the ratio of induced to repressed genes is 1). As the ratio of induced to repressed genes moves away from 1, the advantage of the symmetric cutoff method to sam2.20 is gradually diminishing until eventually sam2.20 becomes significantly better than the symmetric cutoff method when the differentially expressed (DE) genes are either all induced or all repressed genes. Simulation results also show that our proposed simple modification provides improved control of FDR for both sam2.20 and the symmetric cutoff method.

Methods for Estimating and Mitigating Errors in Spotted, Dual-color DNA Microarrays

The conceptual simplicity of DNA microarray technology often belies the complex nature of the measurement errors inherent in the methodology. As the technology has developed, the importance of understanding the sources of uncertainty in the measurements and developing ways to control their influence on the conclusions drawn has become apparent. In this review, strategies for modeling measurement errors and minimizing their effect on the outcome of experiments using a variety of techniques are discussed in the context of spotted, dual-color microarrays. First, methods designed to reduce the influence of random variability through data filtering, replication, and experimental design are introduced. This is followed by a review of data analysis methods that partition the variance into random effects and one or more systematic effects, specifically two-sample significance testing and analysis of variance (ANOVA) methods. Finally, the current state of measurement error models for spotted microarrays and their role in variance stabilizing transformations are discussed.

CLEAR-test: combining inference for differential expression and variability in microarray data analysis

A common goal of microarray experiments is to detect genes that are differentially expressed under distinct experimental conditions. Several statistical tests have been proposed to determine whether the observed changes in gene expression are significant. The t-test assigns a score to each gene on the basis of changes in its expression relative to its estimated variability, in such a way that genes with a higher score (in absolute values) are more likely to be significant. Most variants of the t-test use the complete set of genes to influence the variance estimate for each single gene. However, no inference is made in terms of the variability itself. Here, we highlight the problem of low observed variances in the t-test, when genes with relatively small changes are declared differentially expressed. Alternatively, the z-test could be used although, unlike the t-test, it can declare differentially expressed genes with high observed variances. To overcome this, we propose to combine the z-test, which focuses on large changes, with a chi(2) test to evaluate variability. We call this procedure CLEAR-test and we provide a combined p-value that offers a compromise between both aspects. Analysis of three publicly available microarray datasets reveals the greater performance of the CLEAR-test relative to the t-test and alternative methods. Finally, empirical and simulated data analyses demonstrate the greater reproducibility and statistical power of the CLEAR-test and z-test with respect to current alternative methods. In addition, the CLEAR-test improves the z-test by capturing reproducible genes with high variability.

Quantification of N-(deoxyguanosin-8-yl)-4-aminobiphenyl adducts in human lymphoblastoid TK6 cells dosed with N-hydroxy-4-acetylaminobiphenyl and their relationship to mutation, toxicity, and gene expression profiling

Gene expression profiles that are anchored to phenotypic endpoints may lead to the identification of signatures that predict mutagenicity or carcinogenicity. The study presented here describes the analysis of DNA adducts in the human TK6 lymphoblastoid cell line after exposure to N-hydroxy-4-aminobiphenyl, a mutagenic metabolite of 4-aminobiphenyl. A validated nano-LC microelectrospray mass spectrometry assay is reported for the detection and quantification of N-(deoxyguanosin-8-yl)-4-aminobiphenyl (dG-C8-ABP), the principal DNA adduct of 4-aminobiphenyl. Limits of quantification, based on a signal-to-noise ratio of 10:1, are determined to correspond to approximately 27 fg of dG-C8-ABP injected on-column. The assay has been used to measure the steady-state levels of the adduct in the human TK6 lymphoblastoid cell line as a function of dose (0.5, 1.0, and 10.0 microM) and time (2, 6, and 27 h) after exposure to N-hydroxy-4-aminobiphenyl. The levels of dG-C8-ABP adducts in the cells, ranging from 18 to 500 adducts in 10(9) nucleotides, were then correlated to cell toxicity, induced mutation at the TK (thymidine kinase) and HPRT loci, and gene expression profiling through microarray analysis. Cell cultures were evaluated for toxicity by growth curve extrapolation, mutation assays were performed on the HPRT and TK loci, and gene expression profiles were generated by analyses using microarray technology. In the mutation assay analysis, as the toxicant concentration increased, there was an increase in mutation fraction, indicating a direct correlation to metabolite dosing level and mutations occurring at these two loci. Statistical analysis of the gene expression data determined that a total of 2250 genes exhibited statistically significant changes in expression after treatment with N-OH-AABP (P < 0.05). Among the genes identified, 2245 were up-regulated, whereas 5 genes that had functions in cell survival and cell growth and, hence, could be indicators of toxicity, were down-regulated relative to controls. The results demonstrate the value of anchoring gene expression patterns to phenotypic markers, such as DNA adduct levels, toxicity, and mutagenicity.

Empirical comparison of tests for differential expression on time-series microarray experiments

Methods for identifying differentially expressed genes were compared on time-series microarray data simulated from artificial gene networks. Select methods were further analyzed on existing immune response data of Boldrick et al. (2002, Proc. Natl. Acad. Sci. USA 99, 972-977). Based on the simulations, we recommend the ANOVA variants of Cui and Churchill. Efron and Tibshirani's empirical Bayes Wilcoxon rank sum test is recommended when the background cannot be effectively corrected. Our proposed GSVD-based differential expression method was shown to detect subtle changes. ANOVA combined with GSVD was consistent on background-normalized simulation data. GSVD with empirical Bayes was consistent without background correction. Based on the Boldrick et al. data, ANOVA is best suited to detect changes in temporal data, while GSVD and empirical Bayes effectively detect individual spikes or overall shifts, respectively. For methods tested on simulation data, lowess after background correction improved results. On simulation data without background correction, lowess decreased performance compared to median centering.

Identification of Differentially Expressed Genes with Multivariate Outlier Analysis

DNA microarray offers a powerful and effective technology to monitor the changes in the gene expression levels for thousands of genes simultaneously. It is being widely applied to explore the quantitative alternation in gene regulation in response to a variety of aspects including diseases and exposure of toxicant. A common task in analyzing microarray data is to identify the differentially expressed genes under two different experimental conditions. Because of the large number of genes and small number of arrays, and higher signal-noise ratio in microarray data, many traditional approaches seem improper. In this paper, a multivariate mixture model is applied to model the expression level of replicated arrays, considering the differentially expressed genes as the outliers of the expression data. In order to detect the outliers of the multivariate mixture model, an effective and robust statistical method is first applied to microarray analysis. This method is based on the analysis of kurtosis coefficient (KC) of the projected multivariate data arising from a mixture model so as to identify the outliers. We utilize the multivariate KC algorithm to our microarray experiment with the control and toxic treatment. After the processing of data, the differential genes are successfully identified from 1824 genes on the UCLA M07 microarray chip. We also use the RT-PCR method and two robust statistical methods, minimum covariance determinant (MCD) and minimum volume ellipsoid (MVE), to verify the expression level of outlier genes identified by KC algorithm. We conclude that the robust multivariate tool is practical and effective for the detection of differentially expressed genes.

Estimating the False Discovery Rate Using Mixed Normal Distribution for Identifying Differentially Expressed Genes in Microarray Data Analysis

The recent development of DNA microarray technology allows us to measure simultaneously the expression levels of thousands of genes and to identify truly correlated genes with anticancer drug response (differentially expressed genes) from many candidate genes. Significance Analysis of Microarray (SAM) is often used to estimate the false discovery rate (FDR), which is an index for optimizing the identifiability of differentially expressed genes, while the accuracy of the estimated FDR by SAM is not necessarily confirmed. We propose a new method for estimating the FDR assuming a mixed normal distribution on the test statistic and examine the performance of the proposed method and SAM using simulated data. The simulation results indicate that the accuracy of the estimated FDR by the proposed method and SAM, varied depending on the experimental conditions. We applied both methods to actual data comprised of expression levels of 12,625 genes of 10 responders and 14 non-responders to docetaxel for breast cancer. The proposed method identified 280 differentially expressed genes correlated with docetaxel response using a cut-off value for achieving FDR <0.01 to prevent false-positive genes, although 92 genes were previously thought to be correlated with docetaxel response ones.

Amplification of sense-stranded prokaryotic RNA

Microarray expression analysis has proven to be a valuable methodology. In eukaryotic systems where RNA is limiting, established protocols for amplification of mRNA, which rely on the poly(A) tails, are well established. In contrast, the difficulty in amplifying prokaryotic mRNA has limited the application of microarrays to microbiology. Here we present a method for the Linear Amplification of Prokaryotic Transcripts (LAPT) that is efficient and unbiased. The overhang tailing activity of Moloney murine leukemia virus reverse transcriptase is used to add the T7 promoter to cDNAs during reverse transcription. The promoter addition is uncoupled from the initial priming event allowing the promoter to be attached to the 5' end of the RNA transcript. This enables the amplification of sense-stranded RNA that is representative of the complexity and distribution of the original transcript pool. In microarray assays amplified prokaryotic RNA (10 ng total RNA starting material) showed good Spearman correlations to an unamplified control sample. Using genome-directed primers to bias addition of a T7-promoter to bacterial transcripts allowed amplification of prokaryotic transcripts in the presence of mammalian RNA (at a eukaryotic/prokaryotic RNA ratio of 500 to 1). This technology should facilitate the study of prokaryotic transcriptomes in situations, such as in vivo studies or mixed microbial populations, where the prokaryotic RNA amount is limited and/or the nontarget/target RNA ratios is high.

Analysis of the Leishmania donovani transcriptome reveals an ordered progression of transient and permanent changes in gene expression during differentiation

Leishmania donovani is an intracellular protozoan parasite that causes kala-azar in humans. During infection the extracellular insect forms (promastigotes) undergo rapid differentiation to intracellular amastigotes that proliferates in phagolysosomes of mammalian macrophages. We used microarray-based expression profiling to investigate the time-course of changes in RNA abundance during promastigote-to-amastigote differentiation in a host-free system that mimics this process. These studies revealed that several hundred genes underwent an ordered progression of transient or permanent up- and down-regulation during differentiation. Genes that were permanently up-regulated in amastigotes were enriched for transporters and surface proteins, but under-represented in genes involved in protein and other metabolism. Most of these changes occurred late in the differentiation process, when morphological differentiation was essentially complete. Down-regulated genes were over-represented in those involved in cell motility, growth and/or maintenance, and these changes generally occurred earlier in the process. Genes that were transiently up- or down-regulated during differentiation included those encoding heat shock proteins, ubiquitin hydrolases, RNA binding proteins, protein kinases, a protein phosphatase, and a histone deacetylase. These results suggest that changes in mRNA abundance may be important in signal transduction, as well as protein and mRNA turnover, during differentiation. In addition to these mRNA changes, other transcripts including one or more rRNAs and snoRNAs, and non-coding RNAs from several telomeres, also showed substantial changes in abundance during the differentiation process. This paper provides the first genome-scale quantitative analysis of gene expression during the transition from promastigotes to amastigotes and demonstrates the utility of the host-free differentiation system.

A nonparametric likelihood ratio test to identify differentially expressed genes from microarray data

Microarray experiments contribute significantly to the progress in disease treatment by enabling a precise and early diagnosis. One of the major objectives of microarray experiments is to identify differentially expressed genes under various conditions. The statistical methods currently used to analyse microarray data are inadequate, mainly due to the lack of understanding of the distribution of microarray data. We present a nonparametric likelihood ratio (NPLR) test to identify differentially expressed genes using microarray data. The NPLR test is highly robust against extreme values and does not assume the distribution of the parent population. Simulation studies show that the NPLR test is more powerful than some of the commonly used methods, such as the two-sample t-test, the Mann-Whitney U-test and significance analysis of microarrays (SAM). When applied to microarray data, we found that the NPLR test identifies more differentially expressed genes than its competitors. The asymptotic distribution of the NPLR test statistic and the p-value function is presented. The application of the NPLR method is shown, using both synthetic and real-life data. The biological significance of some of the genes detected only by the NPLR method is discussed.

A bootstrap test for the analysis of microarray experiments with a very small number of replications

In microarray studies it is common that the number of replications (i.e. the sample size) is small and that the distribution of expression values differs from normality. In this situation, permutation and bootstrap tests may be appropriate for the identification of differentially expressed genes. However, unlike bootstrap tests, permutation tests are not suitable for very small sample sizes, such as three per group. A variety of different bootstrap tests exists. For example, it is possible to adjust the data to have a common mean before the bootstrap samples are drawn. For small significance levels, which can occur when a large number of genes is investigated, the original bootstrap test, as well as a bootstrap test suggested for the Behrens-Fisher problem, have no power in cases of very small sample sizes. In contrast, the modified test based on adjusted data is powerful. Using a Monte Carlo simulation study, we demonstrate that the difference in power can be huge. In addition, the different tests are illustrated using microarray data.

Incorporating genome-scale tools for studying energy homeostasis

Mammals have evolved complex regulatory systems that enable them to maintain energy homeostasis despite constant environmental challenges that limit the availability of energy inputs and their composition. Biological control relies upon intricate systems composed of multiple organs and specialized cell types that regulate energy up-take, storage, and expenditure. Because these systems simultaneously perform diverse functions and are highly integrated, they are extremely difficult to understand in terms of their individual component contributions to energy homeostasis. In order to provide improved treatments and clinical options, it is important to identify the principle genetic and molecular components, as well as the systemic features of regulation. To begin, many of these features can be discovered by integrating experimental technologies with advanced methods of analysis. This review focuses on the analysis of transcriptional data derived from microarrays and how it can complement other experimental techniques to study energy homeostasis.

Systems interface biology

The field of systems biology has attracted the attention of biologists, engineers, mathematicians, physicists, chemists and others in an endeavour to create systems-level understanding of complex biological networks. In particular, systems engineering methods are finding unique opportunities in characterizing the rich behaviour exhibited by biological systems. In the same manner, these new classes of biological problems are motivating novel developments in theoretical systems approaches. Hence, the interface between systems and biology is of mutual benefit to both disciplines.

Unsupervised clustering in mRNA expression profiles

The development of microarray technologies gives scientists the ability to examine, discover and monitor the mRNA transcript levels of thousands of genes in a single experiment. Nonetheless, the tremendous amount of data that can be obtained from microarray studies presents a challenge for data analysis. The most commonly used computational approach for analyzing microarray data is cluster analysis, since the number of genes is usually very high compared to the number of samples. In this paper, we investigate the application of the recently proposed k-windows clustering algorithm on gene expression microarray data. This algorithm apart from identifying the clusters present in a data set also calculates their number and thus requires no special knowledge about the data. To improve the quality of the clustering, we employ various dimension reduction techniques and propose a hybrid one. The results obtained by the application of the algorithm exhibit high classification success.

Prognostic DNA methylation biomarkers in ovarian cancer

PURPOSE

Aberrant DNA methylation, now recognized as a contributing factor to neoplasia, often shows definitive gene/sequence preferences unique to specific cancer types. Correspondingly, distinct combinations of methylated loci can function as biomarkers for numerous clinical correlates of ovarian and other cancers.

EXPERIMENTAL DESIGN

We used a microarray approach to identify methylated loci prognostic for reduced progression-free survival (PFS) in advanced ovarian cancer patients. Two data set classification algorithms, Significance Analysis of Microarray and Prediction Analysis of Microarray, successfully identified 220 candidate PFS-discriminatory methylated loci. Of those, 112 were found capable of predicting PFS with 95% accuracy, by Prediction Analysis of Microarray, using an independent set of 40 advanced ovarian tumors (from 20 short-PFS and 20 long-PFS patients, respectively). Additionally, we showed the use of these predictive loci using two bioinformatics machine-learning algorithms, Support Vector Machine and Multilayer Perceptron.

CONCLUSION

In this report, we show that highly prognostic DNA methylation biomarkers can be successfully identified and characterized, using previously unused, rigorous classifying algorithms. Such ovarian cancer biomarkers represent a promising approach for the assessment and management of this devastating disease.

Exploratory data analysis of DNA microarrays by multivariate curve resolution

In this work, the application of a multivariate curve resolution procedure based on alternating least squares optimization (MCR-ALS) for the analysis of data from DNA microarrays is proposed. For this purpose, simulated and publicly available experimental data sets have been analyzed. Application of MCR-ALS, a method that operates without the use of any training set, has enabled the resolution of the relevant information about different cancer lines classification using a set of few components; each of these defined by a sample and a pure gene expression profile. From resolved sample profiles, a classification of samples according to their origin is proposed. From the resolved pure gene expression profiles, a set of over- or underexpressed genes that could be related to the development of cancer diseases has been selected. Advantages of the MCR-ALS procedure in relation to other previously proposed procedures such as principal component analysis are discussed.

Improving reliability and performance of DNA microarrays

A great many platforms and versions of the microarray technology, with different characteristics and applications, have been developed. This review will describe some key issues in reliability and performance with the two most commonly used platforms for gene expression analysis, in situ-synthesized oligonucleotide microarrays or GeneChips and spotted microarrays. Some recent advances and new applications within the field will be mentioned briefly.

Individualized markers optimize class prediction of microarray data

Background

Identification of molecular markers for the classification of microarray data is a challenging task. Despite the evident dissimilarity in various characteristics of biological samples belonging to the same category, most of the marker – selection and classification methods do not consider this variability. In general, feature selection methods aim at identifying a common set of genes whose combined expression profiles can accurately predict the category of all samples. Here, we argue that this simplified approach is often unable to capture the complexity of a disease phenotype and we propose an alternative method that takes into account the individuality of each patient-sample.

Results

Instead of using the same features for the classification of all samples, the proposed technique starts by creating a pool of informative gene-features. For each sample, the method selects a subset of these features whose expression profiles are most likely to accurately predict the sample's category. Different subsets are utilized for different samples and the outcomes are combined in a hierarchical framework for the classification of all samples. Moreover, this approach can innately identify subgroups of samples within a given class which share common feature sets thus highlighting the effect of individuality on gene expression.

Conclusion

In addition to high classification accuracy, the proposed method offers a more individualized approach for the identification of biological markers, which may help in better understanding the molecular background of a disease and emphasize the need for more flexible medical interventions.

ADGO: analysis of differentially expressed gene sets using composite GO annotation

MOTIVATION

Genes are typically expressed in modular manners in biological processes. Recent studies reflect such features in analyzing gene expression patterns by directly scoring gene sets. Gene annotations have been used to define the gene sets, which have served to reveal specific biological themes from expression data. However, current annotations have limited analytical power, because they are classified by single categories providing only unary information for the gene sets.

RESULTS

Here we propose a method for discovering composite biological themes from expression data. We intersected two annotated gene sets from different categories of Gene Ontology (GO). We then scored the expression changes of all the single and intersected sets. In this way, we were able to uncover, for example, a gene set with the molecular function F and the cellular component C that showed significant expression change, while the changes in individual gene sets were not significant. We provided an exemplary analysis for HIV-1 immune response. In addition, we tested the method on 20 public datasets where we found many 'filtered' composite terms the number of which reached approximately 34% (a strong criterion, 5% significance) of the number of significant unary terms on average. By using composite annotation, we can derive new and improved information about disease and biological processes from expression data.

AVAILABILITY

We provide a web application (ADGO: http://array.kobic.re.kr/ADGO) for the analysis of differentially expressed gene sets with composite GO annotations. The user can analyze Affymetrix and dual channel array (spotted cDNA and spotted oligo microarray) data for four species: human, mouse, rat and yeast.

CONTACT

chu@kribb.re.kr

SUPPLEMENTARY INFORMATION

http://array.kobic.re.kr/ADGO.

Comparison of various statistical methods for identifying differential gene expression in replicated microarray data

DNA microarray is a new tool in biotechnology, which allows the simultaneous monitoring of thousands of gene expression in cells. The goal of differential gene expression analysis is to identify those genes whose expression levels change significantly by the experimental conditions. Although various statistical methods have been suggested to confirm differential gene expression, only a few studies compared the performance of the statistical tests. In our study, we extensively compared three types of parametric methods such as T-test, B-statistic and Bayes T-test and three types of non-parametric methods such as samroc, significance analysis of microarray and a modified mixture model using both the simulated datasets and the three real microarray experiments.

A two-sample Bayesian t-test for microarray data

BACKGROUND

Determining whether a gene is differentially expressed in two different samples remains an important statistical problem. Prior work in this area has featured the use of t-tests with pooled estimates of the sample variance based on similarly expressed genes. These methods do not display consistent behavior across the entire range of pooling and can be biased when the prior hyperparameters are specified heuristically.

RESULTS

A two-sample Bayesian t-test is proposed for use in determining whether a gene is differentially expressed in two different samples. The test method is an extension of earlier work that made use of point estimates for the variance. The method proposed here explicitly calculates in analytic form the marginal distribution for the difference in the mean expression of two samples, obviating the need for point estimates of the variance without recourse to posterior simulation. The prior distribution involves a single hyperparameter that can be calculated in a statistically rigorous manner, making clear the connection between the prior degrees of freedom and prior variance.

CONCLUSION

The test is easy to understand and implement and application to both real and simulated data shows that the method has equal or greater power compared to the previous method and demonstrates consistent Type I error rates. The test is generally applicable outside the microarray field to any situation where prior information about the variance is available and is not limited to cases where estimates of the variance are based on many similar observations.

Analysis of cellular responses to aflatoxin B(1) in yeast expressing human cytochrome P450 1A2 using cDNA microarrays

Aflatoxin B1 (AFB(1)) is a potent human hepatotoxin and hepatocarcinogen produced by the mold Aspergillus flavus. In human, AFB(1) is bioactivated by cytochrome P450 (CYP450) enzymes, primarily CYP1A2, to the genotoxic epoxide that forms N(7)-guanine DNA adducts. To characterize the transcriptional responses to genotoxic insults from AFB(1), a strain of Saccharomyces cerevisiae engineered to express human CYP1A2 was exposed to doses of AFB(1) that resulted in minimal lethality, but substantial genotoxicity. Flow cytometric analysis demonstrated a dose and time dependent S phase delay under the same treatment conditions, indicating a checkpoint response to DNA damage. Replicate cDNA microarray analyses of AFB(1) treated cells showed that about 200 genes were significantly affected by the exposure. The genes activated by AFB(1)-treatment included RAD51, DUN1 and other members of the DNA damage response signature reported in a previous study with methylmethane sulfonate and ionizing radiation [A.P. Gasch, M. Huang, S. Metzner, D. Botstein, S.J. Elledge, P.O. Brown, Genomic expression responses to DNA-damaging agents and the regulatory role of the yeast ATR homolog Mec1p, Mol. Biol. Cell 12 (2001) 2987-3003]. However, unlike previous studies using highly cytotoxic doses, environmental stress response genes [A.P. Gasch, P.T. Spellman, C.M. Kao, O. Carmel-Harel, M.B. Eisen, G. Storz, D. Botstein, P.O. Brown, Genomic expression programs in the response of yeast cells to environmental changes, Mol. Biol. Cell 11 (2000) 4241-4257] were largely unaffected by our dosing regimen. About half of the transcripts affected are also known to be cell cycle regulated. The most strongly repressed transcripts were those encoding the histone genes and a group of genes that are cell cycle regulated and peak in M phase and early G1. These include most of the known daughter-specific genes. The rapid and coordinated repression of histones and M/G1-specific transcripts cannot be explained by cell cycle arrest, and suggested that there are additional signaling pathways that directly repress these genes in cells under genotoxic stress.

Rank-based methods as a non-parametric alternative of the T-statistic for the analysis of biological microarray data

We have recently introduced a rank-based test statistic, RankProducts (RP), as a new non-parametric method for detecting differentially expressed genes in microarray experiments. It has been shown to generate surprisingly good results with biological datasets. The basis for this performance and the limits of the method are, however, little understood. Here we explore the performance of such rank-based approaches under a variety of conditions using simulated microarray data, and compare it with classical Wilcoxon rank sums and t-statistics, which form the basis of most alternative differential gene expression detection techniques. We show that for realistic simulated microarray datasets, RP is more powerful and accurate for sorting genes by differential expression than t-statistics or Wilcoxon rank sums - in particular for replicate numbers below 10, which are most commonly used in biological experiments. Its relative performance is particularly strong when the data are contaminated by non-normal random noise or when the samples are very inhomogenous, e.g. because they come from different time points or contain a mixture of affected and unaffected cells. However, RP assumes equal measurement variance for all genes and tends to give overly optimistic p-values when this assumption is violated. It is therefore essential that proper variance stabilizing normalization is performed on the data before calculating the RP values. Where this is impossible, another rank-based variant of RP (average ranks) provides a useful alternative with very similar overall performance. The Perl scripts implementing the simulation and evaluation are available upon request. Implementations of the RP method are available for download from the authors website (http://www.brc.dcs.gla.ac.uk/glama).

[15] Extrapolating Traditional DNA Microarray Statistics to Tiling and Protein Microarray Technologies

A credit to microarray technology is its broad application. Two experiments--the tiling microarray experiment and the protein microarray experiment--are exemplars of the versatility of the microarrays. With the technology's expanding list of uses, the corresponding bioinformatics must evolve in step. There currently exists a rich literature developing statistical techniques for analyzing traditional gene-centric DNA microarrays, so the first challenge in analyzing the advanced technologies is to identify which of the existing statistical protocols are relevant and where and when revised methods are needed. A second challenge is making these often very technical ideas accessible to the broader microarray community. The aim of this chapter is to present some of the most widely used statistical techniques for normalizing and scoring traditional microarray data and indicate their potential utility for analyzing the newer protein and tiling microarray experiments. In so doing, we will assume little or no prior training in statistics of the reader. Areas covered include background correction, intensity normalization, spatial normalization, and the testing of statistical significance.

Analysis of the female gametophyte transcriptome of Arabidopsis by comparative expression profiling

The extensive data on the transcription of the plant genome are derived primarily from the sporophytic generation. There currently is little information on genes that are expressed during female gametophyte development in angiosperms, and it is not known whether the female gametophyte transcriptome contains a major set of genes that are not expressed in the sporophyte or whether it is primarily a subset of the sporophytic transcriptome. Because the embryo sac is embedded within the maternal ovule tissue, we have utilized the Arabidopsis (Arabidopsis thaliana) mutant sporocyteless that produces ovules without embryo sacs, together with the ATH1 Arabidopsis whole-genome oligonucleotide array, to identify genes that are preferentially or specifically expressed in female gametophyte development. From analysis of the datasets, 225 genes are identified as female gametophyte genes, likely a lower limit as stringent criteria were used for the analysis, eliminating many low expressed genes. Nearly 45% of the identified genes were not previously detected by sporophytic expression profiling, suggesting that the embryo sac transcriptome may contain a significant fraction of transcripts restricted to the gametophyte. Validation of six candidate genes was performed using promoterbeta-glucuronidase fusions, and all of these showed embryo sac-specific expression in the ovule. The unfiltered expression data from this study can be used to evaluate the possibility of female gametophytic expression for any gene in the ATH1 array, and contribute to identification of the functions of the component of the Arabidopsis genome not represented in studies of sporophytic expression and function.

Discovering causes and cures for cancer from gene expression analysis

Tumorigenesis is governed by a series of complex genetic and epigenetic changes. Both mechanisms can result in either the silencing or aberrant expression of messages in a cell. Gene expression profiling techniques such as the serial analysis of gene expression (SAGE) or microarray analysis can provide global overviews of these changes, as well identify key genes and pathways involved in this process. This review outlines the current roles of these techniques in cancer research, and how they may contribute to finding not only mechanisms of this disease, but potential targets for therapy.

Gene selection using support vector machines with non-convex penalty

MOTIVATION

With the development of DNA microarray technology, scientists can now measure the expression levels of thousands of genes simultaneously in one single experiment. One current difficulty in interpreting microarray data comes from their innate nature of 'high-dimensional low sample size'. Therefore, robust and accurate gene selection methods are required to identify differentially expressed group of genes across different samples, e.g. between cancerous and normal cells. Successful gene selection will help to classify different cancer types, lead to a better understanding of genetic signatures in cancers and improve treatment strategies. Although gene selection and cancer classification are two closely related problems, most existing approaches handle them separately by selecting genes prior to classification. We provide a unified procedure for simultaneous gene selection and cancer classification, achieving high accuracy in both aspects.

RESULTS

In this paper we develop a novel type of regularization in support vector machines (SVMs) to identify important genes for cancer classification. A special nonconvex penalty, called the smoothly clipped absolute deviation penalty, is imposed on the hinge loss function in the SVM. By systematically thresholding small estimates to zeros, the new procedure eliminates redundant genes automatically and yields a compact and accurate classifier. A successive quadratic algorithm is proposed to convert the non-differentiable and non-convex optimization problem into easily solved linear equation systems. The method is applied to two real datasets and has produced very promising results.

AVAILABILITY

MATLAB codes are available upon request from the authors.

Routine expression profiling of microarray gene signatures in acute leukaemia by real-time PCR of human bone marrow

Cancer subtype diagnosis using microarray signatures has the potential to transform pathological diagnosis but the routine measurement of genes signatures remains difficult. Reverse transcription polymerase chain reaction (RT-PCR) measurement of Indicator genes for acute myeloid leukaemia (AML) and acute lymphoblastic leukaemia (ALL) was used to determine gene signatures. Bone marrow (BM) mononuclear cells were sorted into total, CD34(+) and CD34(-) fractions, and mRNAs globally amplified from each fraction using polyA PCR. The expression profile of the 17 top-ranked genes distinguishing AML and ALL were measured by RT-PCR in five ALL, 26 AML, 12 AML remission, four chronic myeloid leukaemia (CML) and nine morphologically normal BM samples. All but two of the genes measured showed similar expression in AML and ALL to that reported previously. Specifically, c-MYB (P </= 0.04) was significantly increased in ALL in the total fraction, whilst HOXA9 (P </= 0.19) and cystatin c (P </= 0.01) were increased in AML in the CD34(+) and CD34(-) fractions, respectively. c-MYB, hSNF2, RBAP48, HKRT-1, LYN, CD33, Adipsin and HOXA9 were increased in AML compared with remission AML, indicating an ability to determine disease activity. The method used is simple, sensitive and robust, enabling routine clinical use, and it can also be extended to other tumours types with gene signatures.

Minimum redundancy feature selection from microarray gene expression data

How to selecting a small subset out of the thousands of genes in microarray data is important for accurate classification of phenotypes. Widely used methods typically rank genes according to their differential expressions among phenotypes and pick the top-ranked genes. We observe that feature sets so obtained have certain redundancy and study methods to minimize it. We propose a minimum redundancy - maximum relevance (MRMR) feature selection framework. Genes selected via MRMR provide a more balanced coverage of the space and capture broader characteristics of phenotypes. They lead to significantly improved class predictions in extensive experiments on 6 gene expression data sets: NCI, Lymphoma, Lung, Child Leukemia, Leukemia, and Colon. Improvements are observed consistently among 4 classification methods: Naive Bayes, Linear discriminant analysis, Logistic regression, and Support vector machines. SUPPLIMENTARY: The top 60 MRMR genes for each of the datasets are listed in http://crd.lbl.gov/~cding/MRMR/. More information related to MRMR methods can be found at http://www.hpeng.net/.

Gene selection from microarray data for cancer classification--a machine learning approach

A DNA microarray can track the expression levels of thousands of genes simultaneously. Previous research has demonstrated that this technology can be useful in the classification of cancers. Cancer microarray data normally contains a small number of samples which have a large number of gene expression levels as features. To select relevant genes involved in different types of cancer remains a challenge. In order to extract useful gene information from cancer microarray data and reduce dimensionality, feature selection algorithms were systematically investigated in this study. Using a correlation-based feature selector combined with machine learning algorithms such as decision trees, naïve Bayes and support vector machines, we show that classification performance at least as good as published results can be obtained on acute leukemia and diffuse large B-cell lymphoma microarray data sets. We also demonstrate that a combined use of different classification and feature selection approaches makes it possible to select relevant genes with high confidence. This is also the first paper which discusses both computational and biological evidence for the involvement of zyxin in leukaemogenesis.

Evaluation of nine strategies for analyzing a cDNA toxicology microarray data set

Microarray technology with two-color-based cDNA is commonly used for drug development, as well as for a much broader range of biomedical research. Among all the applications, two-group design is probably most commonly used for comparing, e.g., normal and abnormal tissue samples, tissues treated and untreated, or individuals responded and not responded to a drug. Despite the apparent simplicity, there are numerous methods for analyzing such data in a statistically rigorous manner. Here, we discuss nine different analytical strategies, each of which is derived under a set of "reasonable" assumptions. Some of them resemble methods developed for different contexts. In the absence of the truth, investigators should consider underlying assumptions before taking one or more of these strategies for analyzing data from a particular experiment. The issue here is what are the similarities and differences between these analytical strategies. We present these strategies in the context of an actual microarray experiment performed at the U.S. Food and Drug Administration.

Regulation of mouse hepatic genes in response to diet induced obesity, insulin resistance and fasting induced weight reduction

Background

Obesity is associated with insulin resistance that can often be improved by caloric restriction and weight reduction. Although many physiological changes accompanying insulin resistance and its treatment have been characterized, the genetic mechanisms linking obesity to insulin resistance are largely unknown. We used DNA microarrys and RT-PCR to investigate significant changes in hepatic gene transcription in insulin resistant, diet-induced obese (DIO)-C57/BL/6J mice and DIO-C57/BL/6J mice fasted for 48 hours, whose weights returned to baseline levels during these conditions.

Results

Transcriptional profiling of hepatic mRNA revealed over 1900 genes that were significantly perturbed between control, DIO, and fasting/weight reduced DIO mice. From this set, our bioinformatics analysis identified 41 genes that rigorously discriminate these groups of mice. These genes are associated with molecular pathways involved in signal transduction, and protein metabolism and secretion. Of particular interest are genes that participate in pathways responsible for modulating insulin sensitivity. DIO altered expression of genes in directions that would be anticipated to antagonize insulin sensitivity, while fasting/ weight reduction partially or completely normalized their levels. Among these discriminatory genes, Sh3kbp1 and RGS3, may have special significance. Sh3kbp1, an endogenous inhibitor of PI-3-kinase, was upregulated by high-fat feeding, but normalized to control levels by fasting/weight reduction. Because insulin signaling occurs partially through PI-3-kinase, increased expression of Sh3kbp1 by DIO mice may contribute to hepatic insulin resistance via inhibition of PI-3-kinase. RGS3, a suppressor of G-protein coupled receptor generation of cAMP, was repressed by high-fat feeding, but partially normalized by fasting/weight reduction. Decreased expression of RGS3 may augment levels of cAMP and thereby contribute to increased, cAMP-induced, hepatic glucose output via phosphoenolpyruvate carboxykinase (PCK1), whose mRNA levels were also elevated.

Conclusion

These findings demonstrate that hepatocytes respond to DIO and weight reduction by controlling gene transcription in a variety of important molecular pathways. Future studies that characterize the physiological significance of the identified genes in modulating energy homeostasis could provide a better understanding of the mechanisms linking DIO with insulin resistance.

Feature selection and classification for microarray data analysis: evolutionary methods for identifying predictive genes

BACKGROUND

In the clinical context, samples assayed by microarray are often classified by cell line or tumour type and it is of interest to discover a set of genes that can be used as class predictors. The leukemia dataset of Golub et al. 1 and the NCI60 dataset of Ross et al. 2 present multiclass classification problems where three tumour types and nine cell lines respectively must be identified. We apply an evolutionary algorithm to identify the near-optimal set of predictive genes that classify the data. We also examine the initial gene selection step whereby the most informative genes are selected from the genes assayed.

RESULTS

In the absence of feature selection, classification accuracy on the training data is typically good, but not replicated on the testing data. Gene selection using the RankGene software 3 is shown to significantly improve performance on the testing data. Further, we show that the choice of feature selection criteria can have a significant effect on accuracy. The evolutionary algorithm is shown to perform stably across the space of possible parameter settings - indicating the robustness of the approach. We assess performance using a low variance estimation technique, and present an analysis of the genes most often selected as predictors.

CONCLUSION

The computational methods we have developed perform robustly and accurately, and yield results in accord with clinical knowledge: A Z-score analysis of the genes most frequently selected identifies genes known to discriminate AML and Pre-T ALL leukemia. This study also confirms that significantly different sets of genes are found to be most discriminatory as the sample classes are refined.

A class of models for analyzing GeneChip gene expression analysis array data

Background

Various analytical methods exist that first quantify gene expression and then analyze differentially expressed genes from Affymetrix GeneChip^® gene expression analysis array data. These methods differ in the choice of probe measure (quantification of probe hybridization), summarizing multiple probe intensities into a gene expression value, and analysis of differential gene expression. Research papers that describe these methods focus on performance, and how their approaches differ from others. To better understand the common features and differences between various methods, and to evaluate their impact on the results of gene expression analysis, we describe a class of models, referred to as generalized probe models (GPMs), which encompass various currently available methods.

Results

Using an empirical dataset, we compared different formulations of GPMs, and GPMs with three other commonly used methods, i.e. MAS 5.0, dChip, and RMA. The comparison shows that, on a genome-wide scale , different methods yield similar results if the same probe measures are chosen.

Conclusion

In this paper we present a general framework, i.e. GPMs, which encompasses various methods. GPMs permit the use of a wide range of probe measures and facilitate appropriate comparison between commonly used methods. We demonstrate that the dissimilar results stem primarily from different choice of probe measures, rather than other factors.

ArrayVigil: a methodology for statistical comparison of gene signatures using segregated-one-tailed (SOT) Wilcoxon's signed-rank test

Due to versatile diagnostic and prognostic fidelity molecular signatures or fingerprints are anticipated as the most powerful tools for cancer management in the near future. Notwithstanding the experimental advancements in microarray technology, methods for analyzing either whole arrays or gene signatures have not been firmly established. Recently, an algorithm, ArraySolver has been reported by Khan for two-group comparison of microarray gene expression data using two-tailed Wilcoxon signed-rank test. Most of the molecular signatures are composed of two sets of genes (hybrid signatures) wherein up-regulation of one set and down-regulation of the other set collectively define the purpose of a gene signature. Since the direction of a selected gene's expression (positive or negative) with respect to a particular disease condition is known, application of one-tailed statistics could be a more relevant choice. A novel method, ArrayVigil, is described for comparing hybrid signatures using segregated-one-tailed (SOT) Wilcoxon signed-rank test and the results compared with integrated-two-tailed (ITT) procedures (SPSS and ArraySolver). ArrayVigil resulted in lower P values than those obtained from ITT statistics while comparing real data from four signatures.

Small, fuzzy and interpretable gene expression based classifiers

MOTIVATION

Interpretation of classification models derived from gene-expression data is usually not simple, yet it is an important aspect in the analytical process. We investigate the performance of small rule-based classifiers based on fuzzy logic in five datasets that are different in size, laboratory origin and biomedical domain.

RESULTS

The classifiers resulted in rules that can be readily examined by biomedical researchers. The fuzzy-logic-based classifiers compare favorably with logistic regression in all datasets.

AVAILABILITY

Prototype available upon request.

A novel Mixture Model Method for identification of differentially expressed genes from DNA microarray data

Background

The main goal in analyzing microarray data is to determine the genes that are differentially expressed across two types of tissue samples or samples obtained under two experimental conditions. Mixture model method (MMM hereafter) is a nonparametric statistical method often used for microarray processing applications, but is known to over-fit the data if the number of replicates is small. In addition, the results of the MMM may not be repeatable when dealing with a small number of replicates. In this paper, we propose a new version of MMM to ensure the repeatability of the results in different runs, and reduce the sensitivity of the results on the parameters.

Results

The proposed technique is applied to the two different data sets: Leukaemia data set and a data set that examines the effects of low phosphate diet on regular and Hyp mice. In each study, the proposed algorithm successfully selects genes closely related to the disease state that are verified by biological information.

Conclusion

The results indicate 100% repeatability in all runs, and exhibit very little sensitivity on the choice of parameters. In addition, the evaluation of the applied method on the Leukaemia data set shows 12% improvement compared to the MMM in detecting the biologically-identified 50 expressed genes by Thomas et al. The results witness to the successful performance of the proposed algorithm in quantitative pathogenesis of diseases and comparative evaluation of treatment methods.

A two-step strategy for detecting differential gene expression in cDNA microarray data

A mixed-model approach is proposed for identifying differential gene expression in cDNA microarray experiments. This approach is implemented by two interconnected steps. In the first step, we choose a subset of genes that are potentially expressed differentially among treatments with a loose criterion. In the second step, these potential genes are used for further analyses and data-mining with a stringent criterion, in which differentially expressed genes (DEGs) are confirmed and some quantities of interest (such as gene x treatment interaction) are estimated. By simulating datasets with DEGs, we compare our statistical method with a widely used method, the t-statistic, for single genes. Simulation results show that our approach produces a high power and a low false discovery rate for DEG identification. We also investigate the impacts of various source variations resulting from microarray experiments on the efficiency of DEG identification. Analysis of a published experiment studying unstable transcripts in Arabidopsis illustrates the utility of our method. Our method identifies more novel and biologically interesting unstable transcripts than those reported in the original literature.

Rank products: a simple, yet powerful, new method to detect differentially regulated genes in replicated microarray experiments

One of the main objectives in the analysis of microarray experiments is the identification of genes that are differentially expressed under two experimental conditions. This task is complicated by the noisiness of the data and the large number of genes that are examined simultaneously. Here, we present a novel technique for identifying differentially expressed genes that does not originate from a sophisticated statistical model but rather from an analysis of biological reasoning. The new technique, which is based on calculating rank products (RP) from replicate experiments, is fast and simple. At the same time, it provides a straightforward and statistically stringent way to determine the significance level for each gene and allows for the flexible control of the false-detection rate and familywise error rate in the multiple testing situation of a microarray experiment. We use the RP technique on three biological data sets and show that in each case it performs more reliably and consistently than the non-parametric t-test variant implemented in Tusher et al.'s significance analysis of microarrays (SAM). We also show that the RP results are reliable in highly noisy data. An analysis of the physiological function of the identified genes indicates that the RP approach is powerful for identifying biologically relevant expression changes. In addition, using RP can lead to a sharp reduction in the number of replicate experiments needed to obtain reproducible results.

Visualization of large-scale correlations in gene expressions

Large-scale expression data are today measured for several thousands of genes simultaneously. Furthermore, most genes are being categorized according to their properties. This development has been followed by an exploration of theoretical tools to integrate these diverse data types. A key problem is the large noise-level in the data. Here, we investigate ways to extract the remaining signals within these noisy data sets. We find large-scale correlations within data from Saccharomyces cerevisiae with respect to properties of the encoded proteins. These correlations are visualized in a way that is robust to the underlying noise in the measurement of the individual gene expressions. In particular, for S. cerevisiae we observe that the proteins corresponding to the 400 highest expressed genes typically are localized to the cytoplasm. These most expressed genes are not essential for cell survival.

Extreme value distribution based gene selection criteria for discriminant microarray data analysis using logistic regression

One important issue commonly encountered in the analysis of microarray data is to decide which and how many genes should be selected for further studies. For discriminant microarray data analyses based on statistical models, such as the logistic regression models, gene selection can be accomplished by a comparison of the maximum likelihood of the model given the real data, L(D|M), and the expected maximum likelihood of the model given an ensemble of surrogate data with randomly permuted label, L(D(0)|M). Typically, the computational burden for obtaining L(D(0)M) is immense, often exceeding the limits of available computing resources by orders of magnitude. Here, we propose an approach that circumvents such heavy computations by mapping the simulation problem to an extreme-value problem. We present the derivation of an asymptotic distribution of the extreme-value as well as its mean, median, and variance. Using this distribution, we propose two gene selection criteria, and we apply them to two microarray datasets and three classification tasks for illustration.

Tests for finding complex patterns of differential expression in cancers: towards individualized medicine

Background

Microarray studies in cancer compare expression levels between two or more sample groups on thousands of genes. Data analysis follows a population-level approach (e.g., comparison of sample means) to identify differentially expressed genes. This leads to the discovery of 'population-level' markers, i.e., genes with the expression patterns A > B and B > A. We introduce the PPST test that identifies genes where a significantly large subset of cases exhibit expression values beyond upper and lower thresholds observed in the control samples.

Results

Interestingly, the test identifies A > B and B < A pattern genes that are missed by population-level approaches, such as the t-test, and many genes that exhibit both significant overexpression and significant underexpression in statistically significantly large subsets of cancer patients (ABA pattern genes). These patterns tend to show distributions that are unique to individual genes, and are aptly visualized in a 'gene expression pattern grid'. The low degree of among-gene correlations in these genes suggests unique underlying genomic pathologies and high degree of unique tumor-specific differential expression. We compare the PPST and the ABA test to the parametric and non-parametric t-test by analyzing two independently published data sets from studies of progression in astrocytoma.

Conclusions

The PPST test resulted findings similar to the nonparametric t-test with higher self-consistency. These tests and the gene expression pattern grid may be useful for the identification of therapeutic targets and diagnostic or prognostic markers that are present only in subsets of cancer patients, and provide a more complete portrait of differential expression in cancer.

A Genomic Predictor of Oral Squamous Cell Carcinoma

OBJECTIVES/HYPOTHESIS

The objective was to identify a genomic profile that predicts the likelihood of oral squamous cell carcinoma compared with normal oral mucosa in unknown tissue samples.

STUDY DESIGN

Using a training set of tissue samples that were histologically classified as oral squamous cell carcinoma or normal mucosa, the authors used principal component analysis to develop a genomic predictor for oral squamous cell carcinoma. On a separate test set of unclassified samples, the authors used the predictor to classify the samples, then evaluated the performance of the predictor using histological diagnosis.

METHODS

The authors used a data set consisting of messenger RNA extracted from 29 oral squamous cell carcinoma and 19 normal oral mucosa tissue samples and hybridized to Affymetrix oligonucleotide microarrays containing probe sets for 7070 genes and expressed sequence tags. The samples were divided into a training set of 15 oral squamous cell carcinoma and 10 normal samples and a test set consisting of the remaining samples. Using principal component analysis on the training set, the authors found a composite gene expression vector (principal component vector), which they used to compute likelihood ratios for oral squamous cell carcinoma on the test set. By calculating the contribution of each gene to the principal component vector, the authors identified genes with the greatest predictive value.

RESULTS

Using the likelihood ratio, the authors correctly classified all 23 samples in the test set as either oral squamous cell carcinoma or normal. The authors found that many of the most predictive genes are known to be markers of squamous cell carcinoma or normal mucosa.

CONCLUSION

Principal component analysis can be used with genomic microarray data to correctly predict the presence of oral squamous cell carcinoma in unknown tissue samples.