Application of the S-score algorithm for analysis of oligonucleotide microarrays

In the past several years, oligonucleotide microarrays have emerged as a widely used tool for the simultaneous, non-biased measurement of expression levels for thousands of genes. Several challenges exist in successfully utilizing this biotechnology; principal among these is analysis of microarray data. An experiment to measure differential gene expression can consist of a dozen microarrays, each consisting of over a hundred thousand data points. Previously, we have described the use of a novel algorithm for analyzing oligonucleotide microarrays and assessing changes in gene expression. This algorithm describes changes in expression in terms of the statistical significance (S-score) of change, which combines signals detected by multiple probe pairs according to an error model characteristic of oligonucleotide arrays. Software is available that simplifies the use of the application of this algorithm so that it may be applied to improving the analysis of oligonucleotide microarray data. The application of this method to problems of the central nervous system is discussed.     

Application of the S-score algorithm for analysis of oligonucleotide microarrays

In the past several years, oligonucleotide microarrays have emerged as a widely used tool for the simultaneous, non-biased measurement of expression levels for thousands of genes. Several challenges exist in successfully utilizing this biotechnology; principal among these is analysis of microarray data. An experiment to measure differential gene expression can consist of a dozen microarrays, each consisting of over a hundred thousand data points. Previously, we have described the use of a novel algorithm for analyzing oligonucleotide microarrays and assessing changes in gene expression [J. Mol. Biol. 317 (2002) 225]. This algorithm describes changes in expression in terms of the statistical significance (S-score) of change, which combines signals detected by multiple probe pairs according to an error model characteristic of oligonucleotide arrays. Software is available that simplifies the use of the application of this algorithm so that it may be applied to improving the analysis of oligonucleotide microarray data. The application of this method to problems of the central nervous system is discussed.     

Comparison of gene expression measurements from cDNA and 60-mer oligonucleotide microarrays

As the data generated by microarray technology continue to amass, it is necessary to compare and combine gene expression data from different platforms. To evaluate the performance of cDNA and long oligonucleotide (60-mer) arrays, we generated gene expression profiles for two cancer cell lines and compared the data between the two platforms. All 6182 unique genes represented on both platforms were included in the analysis. A limited correlation (r = 0.4708) was obtained and the difference in measurement of low-expression genes was considered to contribute to the limited correlation. Further restriction of the data set to differentially expressed genes detected in cDNA microarrays (1205 genes) and oligonucleotide arrays (1325 genes) showed modest correlations of 0.7076 and 0.6441 between the two platforms. Quantitative real-time PCR measurements of a set of 10 genes showed better correlation with oligonucleotide arrays. Our results demonstrate that there is substantial variation in the data generated from cDNA and 60-mer oligonucleotide arrays. Although general agreement was observed in measurements of differentially expressed genes, we suggest that data from different platforms could not be directly amassed.     

SScore: an R package for detecting differential gene expression without gene expression summaries

SUMMARY: SScore is an R package that facilitates the comparison of gene expression between Affymetrix GeneChips using the S-score algorithm. The S-score algorithm uses probe level data directly to assess differences in gene expression, without requiring a preliminary separate step of probe set expression summary estimation. Therefore, the algorithm avoids introduction of error associated with the expression summary estimation process and has been demonstrated to improve the accuracy of identifying differentially expressed genes. The S-score produces accurate results even when few or no replicates are available. AVAILABILITY: The R package SScore is available from Bioconductor at http://www.bioconductor.org     

Application of microarray technology in primate behavioral neuroscience research

Gene expression profiling of brain tissue samples applied to DNA microarrays promises to provide novel insights into the neurobiological bases of primate behavior. The strength of the microarray technology lies in the ability to simultaneously measure the expression levels of all genes in defined brain regions that are known to mediate behavior. The application of microarrays presents, however, various limitations and challenges for primate neuroscience research. Low RNA abundance, modest changes in gene expression, heterogeneous distribution of mRNA among cell subpopulations, and individual differences in behavior all mandate great care in the collection, processing, and analysis of brain tissue. A unique problem for nonhuman primate research is the limited availability of species-specific arrays. Arrays designed for humans are often used, but expression level differences are inevitably confounded by gene sequence differences in all cross-species array applications. Tools to deal with this problem are currently being developed. Here we review these methodological issues, and provide examples from our experiences using human arrays to examine brain tissue samples from squirrel monkeys. Until species-specific microarrays become more widely available, great caution must be taken in the assessment and interpretation of microarray data from nonhuman primates. Nevertheless, the application of human microarrays in nonhuman primate neuroscience research recovers useful information from thousands of genes, and represents an important new strategy for understanding the molecular complexity of behavior and mental health.     

Identification of differentially expressed genes in high-density oligonucleotide arrays accounting for the quantification limits of the technology

In DNA microarray analysis, there is often interest in isolating a few genes that best discriminate between tissue types. This is especially important in cancer, where different clinicopathologic groups are known to vary in their outcomes and response to therapy. The identification of a small subset of gene expression patterns distinctive for tumor subtypes can help design treatment strategies and improve diagnosis. Toward this goal, we propose a methodology for the analysis of high-density oligonucleotide arrays. The gene expression measures are modeled as censored data to account for the quantification limits of the technology, and two gene selection criteria based on contrasts from an analysis of covariance (ANCOVA) model are presented. The model is formulated in a hierarchical Bayesian framework, which in addition to making the fit of the model straightforward and computationally efficient, allows us to borrow strength across genes. The elicitation of hierarchical priors, as well as issues related to parameter identifiability and posterior propriety, are discussed in detail. We examine the performance of our proposed method on simulated data, then present a detailed case study of an endometrial cancer dataset.     

Monitoring cellular responses to Listeria monocytogenes with oligonucleotide arrays

Listeria monocytogenes is a pathogenic intracellular microorganism whose infection induces pleiotropic biological changes associated with host cell gene expression regulation. Here we define the gene expression profiles of the human promyelocytic THP1 cell line before and after L. monocytogenes infection. Gene expression was measured on a large scale via oligonucleotide microarrays with probe sets corresponding to 6,800 human genes. We assessed and discussed the reproducibility of the hybridization signatures. In addition to oligonucleotide arrays, we also performed the large scale gene expression measurement with two high-density membranes, assaying for 588 and 18,376 human genes, respectively. This work allowed the reproducible identification of 74 up-regulated RNAs and 23 down-regulated RNAs as a consequence of L. monocytogenes infection of THP1. The reliability of these data was reinforced by performing independent infections. Some of these detected RNAs were consistent with previous results, while some newly identified RNAs encode gene products that may play key roles in L. monocytogenes infection. These findings will undoubtedly enhance the understanding of L. monocytogenes molecular physiology and may help identify new therapeutic targets.     

Empirical validation of the S-Score algorithm in the analysis of gene expression data

Background

Current methods of analyzing Affymetrix GeneChip^® microarray data require the estimation of probe set expression summaries, followed by application of statistical tests to determine which genes are differentially expressed. The S-Score algorithm described by Zhang and colleagues is an alternative method that allows tests of hypotheses directly from probe level data. It is based on an error model in which the detected signal is proportional to the probe pair signal for highly expressed genes, but approaches a background level (rather than 0) for genes with low levels of expression. This model is used to calculate relative change in probe pair intensities that converts probe signals into multiple measurements with equalized errors, which are summed over a probe set to form the S-Score. Assuming no expression differences between chips, the S-Score follows a standard normal distribution, allowing direct tests of hypotheses to be made. Using spike-in and dilution datasets, we validated the S-Score method against comparisons of gene expression utilizing the more recently developed methods RMA, dChip, and MAS5.

Results

The S-score showed excellent sensitivity and specificity in detecting low-level gene expression changes. Rank ordering of S-Score values more accurately reflected known fold-change values compared to other algorithms.

Conclusion

The S-score method, utilizing probe level data directly, offers significant advantages over comparisons using only probe set expression summaries.     

A simple method to improve probe set estimates from oligonucleotide arrays

A popular commercially available oligonucleotide microarray technology employs sets of 25 base pair oligonucleotide probes for measurement of gene expression levels. A mathematical algorithm is required to compute an estimate of gene expression from the multiple probes. Previously proposed methods for summarizing gene expression data have either been substantially ad hoc or have relied on model assumptions that may be easily violated. Here we present a new algorithm for calculating gene expression from probe sets. Our approach is functionally related to leave-one-out cross-validation, a non-parametric statistical technique that is often applied in limited data situations. We illustrate this approach using data from our study seeking a molecular fingerprint of STAT3 regulated genes for early detection of human cancer.     

HugeIndex: a database with visualization tools for high-density oligonucleotide array data from normal human tissues

High-density oligonucleotide arrays are a powerful tool for uncovering changes in global gene expression in various disease states. To this end, it is essential to first characterize the variations of gene expression in normal physiological processes. We established the Human Gene Expression (HuGE) Index database (www.HugeIndex.org) to serve as a public repository for gene expression data on normal human tissues using high-density oligonucleotide arrays. This resource currently contains the results of 59 gene expression experiments on 19 human tissues. We provide interactive tools for researchers to query and visualize our data over the Internet. To facilitate data analysis, we cross-reference each gene on the array with its annotation in the LocusLink database at NCBI.     

Prokaryotic RNA preparation methods useful for high density array analysis: comparison of two approaches

High density oligonucleotide arrays have been used extensively for expression studies of eukaryotic organisms. We have designed a prokaryotic high density oligonucleotide array using the complete Escherichia coli genome sequence to monitor expression levels of all genes and intergenic regions in the genome. Because previously described methods for preparing labeled target nucleic acids are not useful for prokaryotic cell analysis using such arrays, a mRNA enrichment and direct labeling protocol was developed together with a cDNA synthesis protocol. The reproducibility of each labeling method was determined using high density oligonucleotide probe arrays as a read-out methodology and the expression results from direct labeling were compared to the expression results from the cDNA synthesis. About 50% of all annotated E.coli open reading frames are observed to be transcribed, as measured by both protocols, when the cells were grown in rich LB medium. Each labeling method individually showed a high degree of concordance in replica experiments (95 and 99%, respectively), but when each sample preparation method was compared to the other, ~32% of the genes observed to be expressed were discordant. However, both labeling methods can detect the same relative gene expression changes when RNA from IPTG-induced cells was labeled and compared to RNA from uninduced E.coli cells.     

Quantitative expression profile of distinct functional regions in the adult mouse brain

The adult mammalian brain is composed of distinct regions with specialized roles including regulation of circadian clocks, feeding, sleep/awake, and seasonal rhythms. To find quantitative differences of expression among such various brain regions, we conducted the BrainStars (B*) project, in which we profiled the genome-wide expression of ～50 small brain regions, including sensory centers, and centers for motion, time, memory, fear, and feeding. To avoid confounds from temporal differences in gene expression, we sampled each region every 4 hours for 24 hours, and pooled the samples for DNA-microarray assays. Therefore, we focused on spatial differences in gene expression. We used informatics to identify candidate genes with expression changes showing high or low expression in specific regions. We also identified candidate genes with stable expression across brain regions that can be used as new internal control genes, and ligand-receptor interactions of neurohormones and neurotransmitters. Through these analyses, we found 8,159 multi-state genes, 2,212 regional marker gene candidates for 44 small brain regions, 915 internal control gene candidates, and 23,864 inferred ligand-receptor interactions. We also found that these sets include well-known genes as well as novel candidate genes that might be related to specific functions in brain regions. We used our findings to develop an integrated database (http://brainstars.org/) for exploring genome-wide expression in the adult mouse brain, and have made this database openly accessible. These new resources will help accelerate the functional analysis of the mammalian brain and the elucidation of its regulatory network systems.     

Genome-wide screening of alpha-tocopherol sensitive genes in heart tissue from alpha-tocopherol transfer protein null mice (ATTP(-/-))

Alpha-tocopherol transfer protein (ATTP) null mice (ATTP(-/-)) have a systemic deficiency of alpha-tocopherol (AT). The heart AT levels of ATTP(-/-) are <10% of those in ATTP(+/+) mice. The genomic responses of heart to AT deficiency were determined in 3 months old male ATTP(-/-) mice and compared with their ATTP(+/+) littermate controls using Affymetrix 430A 2.0 high density oligonucleotide arrays. Differential analysis of approximately 13000 genes identified repression of genes related to immune system and activation of genes related to lipid metabolism and inflammation with no significant change in the expression of classical antioxidant genes (catalase, superoxide dismutase, glutathione peroxidase) in ATTP(-/-) as compared to ATTP(+/+) mice. The present data identifies novel classes of AT sensitive genes in heart tissue.     

Expression profiling by oligonucleotide microarrays spotted on coated polymer slides

We have developed a ready-to-spot polymer microarray slide, which is coated with a uniform layer of reactive electrophilic groups using anthraquinone-mediated photo-coupling chemistry. The slide coating reduces the hydrophobicity of the native polymer significantly, thereby enabling robust and efficient one-step coupling of spotted 5' amino-linked oligonucleotides onto the polymer slide. The utility of the coated polymer slide in gene expression profiling was assessed by fabrication of spotted oligonucleotide microarrays using a collection of 5' amino-linked 70-mer oligonucleotide probes representing 96 yeast genes from Operon. Two-colour hybridizations with labelled cDNA target pools derived from standard grown and heat-shocked wild type yeast cells could reproducibly measure heat shock induced expression of seven different heat shock protein (HSP) genes. Moreover, the observed fold changes were comparable to those reported previously using spotted cDNA arrays and high-density 25-mer oligonucleotide arrays from Affymetrix. The low hybridization signals obtained from the DeltaSSA4 mutant cDNA target, together with the high signal detected in two-colour hybridizations with heat-shocked wild type yeast relative to the DeltaSSA4 mutant strain implies that unspecific binding of cDNA target to the SSA4-specific 70-mer oligonucleotide probes is negligible. Combined, our results indicate that the coated polymer microarray slide represents a robust and cost-effective array platform for pre-spotted oligonucleotide arrays.     

Match-Only Integral Distribution (MOID) Algorithm for high-density oligonucleotide array analysis

Background

High-density oligonucleotide arrays have become a valuable tool for high-throughput gene expression profiling. Increasing the array information density and improving the analysis algorithms are two important computational research topics.

Results

A new algorithm, Match-Only Integral Distribution (MOID), was developed to analyze high-density oligonucleotide arrays. Using known data from both spiking experiments and no-change experiments performed with Affymetrix GeneChip^® arrays, MOID and the Affymetrix algorithm implemented in Microarray Suite 4.0 (MAS4) were compared. While MOID gave similar performance to MAS4 in the spiking experiments, better performance was observed in the no-change experiments. MOID also provides a set of alternative statistical analysis tools to MAS4. There are two main features that distinguish MOID from MAS4. First, MOID uses continuous P values for the likelihood of gene presence, while MAS4 resorts to discrete absolute calls. Secondly, MOID uses heuristic confidence intervals for both gene expression levels and fold change values, while MAS4 categorizes the significance of gene expression level changes into discrete fold change calls.

Conclusions

The results show that by using MOID, Affymetrix GeneChip^® arrays may need as little as ten probes per gene without compromising analysis accuracy.