Author Etienne Thevenot (W4M Core Development Team, MetaboHUB Paris, CEA)
References
Tool updates
See the NEWS section at the bottom of this page
Performs hierarchical clustering on both the samples (rows) and variables (columns) of the dataMatrixDisplays the dataMatrix with sorted rows and samples and the dendrograms (heatmap)In the output dataMatrix, sampleMetadata and variableMetadata files sample and variables are sorted according to the dendrogramsOptionally, indicates the groups of samples and/or variables obtained by cutting the dendrograms into a specific number of partitionsNote: 1) Computations rely on the 'hclust' function. The dissimilarity is 1 - cor (where cor is the Spearman correlation) and the 'ward.D' aggregating method is used.2) A "blue-orange-red" palette is generated with the function 'colorRampPalette'; By default, variables are standardized (mean-centered and unit-scaled) to enhance contrast on the figure; standardization can be turned off by using the full list of parameters; in any case, standardizing is performed after the computation of clusters, for display only3) When a specific number of sample and/or variable groups (i.e. > 1) are selected, the group numbers are indicated on the plot and in an additional 'heat_clust" column in the sampleMetadata and/or variableMetadata4) Example of computation times: for 126 variables: a few seconds; for 4324 variables: 30 min
File type | Format |
---|---|
1 : Data matrix | tabular |
2 : Sample metadata | tabular |
3 : Variable metadata | tabular |
See the W4M00001a_sacurine-subset-statistics shared history in the Shared Data/Published Histories menu (https://galaxy.workflow4metabolomics.org/history/list_published)
INTERNAL MODIFICATIONS
Minor internal modifications
NEW FEATURES
Default method for the correlation coefficient is now 'pearson', instead of 'spearman' previously (the latter can still be selected in the advanced parameters)
The 1-abs(correlation) dissimilarity is now available (in addition to the default '1-correlation') in case the sign of correlations between samples and between variables does not matter, as well as the euclidean, maximum, manhattan, canberra, binary, and minkowski dissimilarities
A new red-green color scale is available
INTERNAL MODIFICATIONS
Creating additional files for planemo and travis running and installation validation
INTERNAL MODIFICATIONS
Internal replacement of the as.hclust function which happened to produce error messages