Test type:
Input file(s):
Module subset:
Load example data:
Actions:

Genes shown in bold are in the main data set

General

tmod is an R package for analysing blood transcriptional modules (BTMs). This web page is an interface to that package. For more information, read the package vignette.

The definitions of the BTMs are taken either from Li et al. or Chaussabel et al.

In general, the genes in the modules are defined by their HGNC symbols, for example “GBP5” or “IFNB1”. You can also use Refseq or Unigene identifiers. The mapping of these identifiers to HGNC symbols is done using the BioConductor package org.Hs.eg.db


Tests

There are three statistical tests pre-configured for tmod:

  • Mann-Whitney U-test, and
  • CERNO, and
  • hypergeometric test for enrichment

First, select the desired statistical test under “Test type”. Next, upload the file (for U-test) or files (for hypergeometric test) under “Input file(s)”.

Finally, click on “Run tmod”.

Mann-Whitney U test

Input: The U test works on a single sorted list of gene symbols.

  • Create an Excel file with the first containing gene names
  • Sort the file e.g. by p-value or principal component weight
  • Export as “CSV” (comma separated values)
  • Click on “Choose file” in the “Tests” tab.
  • Click on the “Run tmod” button

The results should now appear below.

Output: The results table contains the following columns:

  • Plot: click on the button in this column to see the evidence plot
  • ID: module ID
  • Title: module name
  • U: the Mann-Whitney U statistics
  • N1: sum of ranks of genes in the module
  • auc: Area under curve, a measure of the effect size
  • P.Value: p-value of the U test, raw
  • adj.P.Val: p-value for the U test, corrected for multiple testing

CERNO test

The CERNO test has been described by Yamaguchi et al. (2007). It is very similar to the Mann-Whitney test in that it uses a single sorted list of gene symbols, but the statistics is calculated differently and appears to perform better for some of the modules.

Input: The CERNO test works on a single sorted list of gene symbols. Follow the instructions for the U test.

Output: The results table contains the following columns:

  • Plot: click on the button in this column to see the evidence plot
  • ID: module ID
  • Title: module name
  • cerno: the CERNO statistics
  • N1: sum of ranks of genes in the module
  • cES: average log-rank of the genes in the module
  • P.Value: p-value of the U test, raw
  • adj.P.Val: p-value for the U test, corrected for multiple testing

Hypergeometric test

Input: The HG test works on two unsorted lists of gene symbols. First list, called “foreground”, contains the symbols of genes that are thought to be for example differentially expressed. The second list contains all gene symbols that were included in the analysis (e.g. spotted on a microarray).

  • Create two Excel files:
    • foreground file with genes that are differentially expressed
    • background file with all genes (“gene universe”)
  • Go to the “Test” tab of the tmod interface
  • Under “Test type” select “hypergeometric test”
  • Click on both “Choose file” buttons that appear and select the foreground and the background
  • Click on the “Run tmod” button

The results should now appear below.

Output: The results table contains the following columns:

  • Plot: click on the button in this column to see the evidence plot
  • ID: module ID
  • Title: module name
  • b: number of genes in the foreground set that belong to the given module
  • B: total number of genes that belong to the given module
  • n: size of foreground set
  • N: size of gene universe (foreground + background)
  • E: enrichment coefficient equal to (b/n)/(B/N)
  • P.Value: p-value of the hypergeometric test, raw
  • adj.P.Val: p-value for the hypergeometric test, corrected for multiple testing

The result table

You can sort the result table by clicking on the respective headers. Furthermore, you can limit the displayed items by entering text in the boxes below the result table.


Evidence plots

By clicking on the radio button of the left side of the result table, you can view an “evidence plot” – a plot which visualizes the enrichment found in the data. Depending on the type of tests used, the evidence plot is either a ROC curve (for the U test) or a barplot (for the hypergeometric test).

Below you see an evidence plot generated by tmod in the single list mode (U test). The ROC curve shows an enrichment of the module LI.M37.1 (called enriched in neutrophils (I)).

The rug (grey bar) at the bottom of the plot shows the distribution of module genes within the total gene list. Each black vertical line denotes one gene.

Evidence plot generated by tmod


Problems and release notes

This is the initial release of tmod interface. If you have problems with the web page, please contact January Weiner.


Citing tmod

Please cite our paper ([PDF]) on tmod:

Weiner 3rd J, Domaszewska T. (2016) tmod: an R package for general and multivariate enrichment analysis. PeerJ Preprints 4:e2420v1 https://doi.org/10.7287/peerj.preprints.2420v1

Despite the misleading “not peer reviewed” disclaimer, this paper is a part of the GBC 2016 conference series and as such underwent peer review.

Download tmod

Older versions:

Download Example Data

Dependencies

tmod depends on the packages XML, beeswarm, pca3d and tagcloud.

Help and Documentation

Presentations

  • Presentation for GCB2016
    • online version
    • full archive download – contains R code for generating the plots in the presentation
  • Presentation “Functional multivariate analysis with the tmod package”:

Installation

Follow the standard procedures for installing R packages. If in doubt, save the package in a directory, and at the R prompt, type

install.package("/path/to/tmod/file/tmod_0.6.tar.gz")

Gallery of illustrations generated with tmod

For more details on these plots read the tmod package vignette

[basic2-1.png]

basic2-1.png

Basic example of the showgene() function.


[eight-1.png]

eight-1.png

Tag cloud generated from a result of an enrichment analysis.


[five-1.png]

five-1.png

Example of an evidence plot – a ROC curve describing the enrichment of genes from a gene set along a sorted list of genes.


[nine-1.png]

nine-1.png

Functional multivariate analysis: a principal component plot annotated with tag clouds generated from enrichment analysis of the components.


[pcsum2-1.png]

pcsum2-1.png

Serial enrichment panel of a PCA. Each row corresponds to a module; columns correspond to PCs. Size of red blobs corresponds to AUC, intensity of color to the p-value.


[pcsum3-1.png]

pcsum3-1.png

Serial enrichment of PCA. Overview panel showing fractions of genes with negative and positive weights in each enriched module. Generated with tmodPanelPlot().


[pplot1-1.png]

pplot1-1.png

Panel plot of enrichment for two coefficients in a limma analysis. Generated with tmodLimmaTest and tmodPanelPlot functions.


[pplot3-1.png]

pplot3-1.png

Panel plot of enrichment in two coefficients in a limma analysis. Pie charts show fractions of up- and down-regulated genes.


[pplot4-1.png]

pplot4-1.png

Panel rug plot showing fractions of significantly up-regulated genes in a limma two coefficient analysis.