Project description:Background Many microarray experiments search for genes with differential expression between a common “reference” group and multiple test groups, like in the case of time-course designs or of various treatments versus a control condition. In such cases, currently employed statistical approaches based on t-test or close derivatives have limited efficacy, mostly because estimation of noise is done on only two groups at time. Alternative approaches based on ANOVA correctly capture noise from all the groups, but then do not confront single test groups with the reference. We therefore conceived a statistical test for pairwise comparisons between the reference group and each test group that uses within-group variance calculated from all the groups. Results We implemented an R-Bioconductor package named Mulcom, with a statistical test derived from the Dunnett’s test, designed to compare multiple experimental groups against a common reference. In addition to the basic Dunnett’s t value, the package includes an optional minimal fold-change threshold, m. Thanks to automated, permutation-based estimation of False Discovery Rate (FDR), the package also permits fast optimization of the test, to obtain the maximum number of significant genes at a given FDR value. When applied on a time-course experiment profiled in parallel on two microarray platforms, and compared with currently used tests, Mulcom displayed higher concordance of significant genes in the two array platforms, and higher enrichment in functional annotation to categories related to the biology of the experiment. Conclusions The Mulcom package provides a fast and powerful tool for the identification of differentially expressed genes when several experimental conditions are compared with a common reference. We found that Mulcom leads to lists of differentially expressed genes that are particularly consistent across microarray platforms and enriched in significant classes of genes. In our opinion, the main reasons for these good performances are three: (i) within-group variability is estimated from all experimental groups even if only two of them are compared each time; (ii) the optional fold-change threshold m avoids false positives due to aberrantly low within-group variability; (iii) automated test optimization allows maximizing sensitivity without compromising specificity.

Project description:BackgroundWe present the algorithm PFClust (Parameter Free Clustering), which is able automatically to cluster data and identify a suitable number of clusters to group them into without requiring any parameters to be specified by the user. The algorithm partitions a dataset into a number of clusters that share some common attributes, such as their minimum expectation value and variance of intra-cluster similarity. A set of n objects can be clustered into any number of clusters from one to n, and there are many different hierarchical and partitional, agglomerative and divisive, clustering methodologies available that can be used to do this. Nonetheless, automatically determining the number of clusters present in a dataset constitutes a significant challenge for clustering algorithms. Identifying a putative optimum number of clusters to group the objects into involves computing and evaluating a range of clusterings with different numbers of clusters. However, there is no agreed or unique definition of optimum in this context. Thus, we test PFClust on datasets for which an external gold standard of 'correct' cluster definitions exists, noting that this division into clusters may be suboptimal according to other reasonable criteria. PFClust is heuristic in the sense that it cannot be described in terms of optimising any single simply-expressed metric over the space of possible clusterings.ResultsWe validate PFClust firstly with reference to a number of synthetic datasets consisting of 2D vectors, showing that its clustering performance is at least equal to that of six other leading methodologies - even though five of the other methods are told in advance how many clusters to use. We also demonstrate the ability of PFClust to classify the three dimensional structures of protein domains, using a set of folds taken from the structural bioinformatics database CATH.ConclusionsWe show that PFClust is able to cluster the test datasets a little better, on average, than any of the other algorithms, and furthermore is able to do this without the need to specify any external parameters. Results on the synthetic datasets demonstrate that PFClust generates meaningful clusters, while our algorithm also shows excellent agreement with the correct assignments for a dataset extracted from the CATH part-manually curated classification of protein domain structures.

Dataset Information

A parameter-free statistical test for neuronal responsiveness.

Publications

A parameter-free statistical test for neuronal responsiveness.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets