PAGE: parametric analysis of gene set enrichment.
Kim SY, Volsky DJ. · 2005
Read PAGE: parametric analysis of gene set enrichment. by Kim SY, Volsky DJ. (2005) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc1183189.txt.
Introduction
Kim and Volsky presented a 2005 method named PAGE for gene data. Labs often test thousands of genes at once on small chips. Single gene changes can look weak or noisy. Grouping genes into known sets gives a clearer signal of change.
PAGE uses standard math on those gene groups instead of slow shuffling tests. In trials it found more truly changed groups than the older method. It also ran much faster. Kim tested it on muscle data from young and old donors. Shared aging themes stood out more clearly at group level.
Catalog record
| CREATOR | Kim SY, Volsky DJ. |
|---|---|
| YEAR | 2005 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
PAGE: Parametric Analysis of Gene Set Enrichment By Seon-Young Kim, David J Volsky Published in: BMC Bioinformatics DOI: 10.1186/1471-2105-6-144 ABSTRACT Abstract Background Gene set enrichment analysis (GSEA) is a microarray data analysis method that uses predefined gene sets and ranks of genes to identify significant biological changes in microarray data sets. GSEA is especially useful when gene expression changes in a given microarray data set is minimal or moderate. Results We developed a modified gene set enrichment analysis method based on a parametric statistical analysis model. Compared with GSEA, the parametric analysis of gene set enrichment (PAGE) detected a larger number of significantly altered gene sets and their p-values were lower than the corresponding p-values calculated by GSEA. Because PAGE uses normal distribution for statistical inference, it requires less computation than GSEA, which needs repeated computation of the permutated data set. PAGE was able to detect significantly changed gene sets from microarray data irrespective of different Affymetrix probe level analysis methods or different microarray platforms. Comparison of two aged muscle microarray data sets at gene set level using PAGE revealed common biological themes better than comparison at individual gene level. Conclusion PAGE was statistically more sensitive and required much less computational effort than GSEA, it could identify significantly changed biological themes from microarray data irrespective of analysis methods or microarray platforms, and it was useful in comparison of multiple microarray data sets. We offer PAGE as a useful microarray analysis method. BACKGROUND High-throughput technologies such as DNA microarrays and proteomics are revolutionizing biology and medicine. Global gene expression profiling using microarrays monitors changes in expression of thousands of genes simultaneously. At the data acquisition level, gene expression profiles from a given system should be reproducible and yield statistically significant changes in gene expression [1]. The large amounts of data acquired must then be reduced or "translated" to a smaller set of genes representing meaningful biological differences between control and test systems and validated in an experimental or clinical setting [2]. Since inception of the microarray technology, significant technological and analytical improvements have been introduced to meet these challenges, from experimental design [1], probe-level analysis of oligonucleotide chips [3,4], data normalization [5], statistical analysis [6], clustering techniques [7-9], to various data mining tools [10-12]. A large number of studies used microarrays successfully to discern changes in gene expression patterns either in well defined cellular populations responding to a specific stimulus in vitro [13,14] or in complex clinical settings such as cancer and neurological diseases [15-17]. While individual microarray studies can be highly informative, it is generally difficult to compare independently obtained data sets addressing the same biological problem [18], regardless of whether the same or different microarray platform was used [19-23]. The poor congruence of cross-study comparisons was attributed to incorrectly annotated probes, non-sequence overlapping but Unigene-matching probes, variation in experimental conditions, and actual biological variations among different clinical or experimental materials used [19-23]. Several new bioinformatics programs have attempted to circumvent the variability between different published data sets at the data acquisition level by comparing gene expression results for coordinate changes in biological themes [12], for similarity of significance values for each gene obtained through meta-analysis methods [15], and for reproducible gene expression patterns revealed by integrative correlation statistics [24]. Each of the methods revealed some congruency among the data sets analyzed [12,15,24] indicating that results from various transcriptional profiling studies can eventually be integrated for better general definition of normal and disease-related processes. Another challenge of microarray data analysis is that the majority of genes in any genome-wide transcriptional profile data set are excluded from consideration because they show only subtle changes in expression. This problem was recently addressed in a gene expression profiling study of human diabetic muscles, in which no single gene (out of over 20,000) showed significant difference in expression between control and patients groups [25]. Assuming that gene expression changes can be detected at the level of co-regulated gene sets rather than individual genes, the authors devised a new analytical tool, GSEA, that tested predefined gene sets for association with disease phenotypes [25]. GSEA successfully detected oxidative phosphorylation as a biological theme that coordinately changed in diabetic muscles [25]. Although subsequent analysis suggested that the statistical tools used in GSEA may be biased toward assigning higher enrichment scores to gene sets of large size [26], the program significantly expands the potential for discovery of important process or disease related genes in a given microarray data set. In the present work we describe a modified gene set enrichment analysis strategy that improves analysis of minimally changed gene expression profiles. PAGE employs fold change between experimental groups or other parametric data to calculate Z scores of predefined gene sets and use normal distribution to infer statistical significance of gene sets. We show here that PAGE has several advantages over GSEA and is useful in comparison of multiple microarray data sets. RESULTS Statistical model and selection of minimal gene set size According to the Central Limit Theorem in statistics, the distribution of the average of randomly sampled n observations tends to follow normal distribution as the sampling size n becomes larger, even when the parent distribution from which the average is calculated is not normal. The distribution of the average of randomly sampled observations has the same mean as the parent distribution and its variance is equal to the variance of the parent divided by the sampling size [27]. In other words, when the mean and variance of the parent distribution (whether it is normally distributed or not) are μ and σ2 the average of n observations from the parent distribution will follow a normal distribution of mean μ and variance σ2/n when the sampling size n is large enough. In PAGE, the parent distribution is a distribution of any numerical values (also termed parameters here) that describe differential expression of genes among samples in a microarray data set. Usually, the values are a fold change for an individual gene between two experimental groups or they can be a correlation coefficient between clinical indices and individual gene expression values in a microarray data set. In most cases, the distribution of a parameter, i.e., a fold change values for all genes in a gene set between two experimental groups, is not normally distributed. However, as the Central Limit Theorem states, when we sample n observations from the parent distribution of a parameter, the average of the sampled observations tends to follow the normal distribution as our sampling size n becomes larger. Here, we define sampled observations as expression values for randomly chosen individual genes within pre-defined gene sets, which may be any randomly chosen groups of genes, groups of genes representing close family members with similar functions, genes in the same biological pathway, and so on. If we define a gene set of sufficiently large size, we can use the normal distribution to test the statistical significance of that gene set. To determine the minimal gene set size m, we first examined the distribution pattern of several microarray data sets. We used fold change between two experimental groups as a parameter and observed the distribution of fold change values in a microarray data set. As an example, we show