resplot: Simple Graphical Summaries for Gene Set Enrichment Analysis...

In GSEAlm: Linear Model Toolset for Gene Set Enrichment Analysis

Description

Diagnostic plots for GSEA. 'resplot' and 'restrip' group residuals (or expression levels) from a specific gene-set by sample. 'mnDiffPlot' shows mean expression differences for a dichotomous phenotype, by gene, for a specific gene set.

Usage

resplot(GSname = "All", resmat, incidence = dumminc(resmat), fac,
     atomic = "Gene", core.text = "Residuals by Sample",
     yname = "Standardized Residual", xname = "Sample ID",
     ID = colnames(resmat), lims = 0, gnames = levels(factor(fac)),
     prefix = "", horiz = FALSE, colour=5,pch='+',...)

restrip(GSname = "All", resmat, incidence = dumminc(resmat), fac,
   atomic = "Gene", core.text = "Residuals by Sample",
  yname = "Standardized Residual", xname = "Sample ID", ID = colnames(resmat),
  gnames = levels(factor(fac)), prefix = "", colour=c(2:4,6), resort=TRUE,
  horiz = FALSE, resort.fun=num.positive, pch='+', ...)

mnDiffPlot(GSname = "All", exprmat, incidence = dumminc(exprmat), fac,
  atomic = "Gene", core.text = paste("Mean Expression Difference by",atomic),
  yname="Log Expression Ratio", xname="Log Expression",
  gnames = levels(factor(fac)), prefix = "", fitline=FALSE, varsize=FALSE,
  reverse=FALSE, ...)

Arguments

GSname	Gene-set Name. See "Details".
resmat,exprmat	Numerical matrix with the values to be plotted. See "Details".
incidence	Gene-set 0/1 membership matrix
fac	The phenotypical variable to plot by. Must be discrete. For 'mnDiffPlot', must be dichotomous.
atomic	string identifying the meaning of rows in the data matrix. Defaults to "Gene".
core.text,gnames,prefix,xname,yname	strings controlling the text of main and axis captions
ID	Group names associated with the data matrix columns
lims	plotting limits for the response axis
horiz	logical: whether the boxplots or strips should be horizontal (defaults to FALSE)
colour	color of boxplot filling ('resplot') or symbols ('restrip')
pch	the plotting symbol
resort	('restrip' only) whether to sort groups for better visibility
resort.fun	('restrip' only) what function to sort groups by. Ignored unless 'resort==TRUE'. See stripchart documentation for more details
fitline	('mnDiffPlot' only) logical: whether a loess fit should be plotted
varsize	('mnDiffPlot' only) logical: whether symbol sizes should be proportional to (t-test style) standard errors
reverse	('mnDiffPlot' only)logical: whether the factor's order should be reversed so that the second level is on the x-axis rather than the first one
...	Additional graphical parameters passed on to the generic plotting functions.

Details

These functions provide simple graphical summaries for processed gene-expression data, or other similar datasets for which matrix form is useful. They are tailored predominantly for GSEA, but are useful in general as well.

'resplot' calls boxplot and 'restrip' calls stripchart; both summarize *all* data points from those rows in 'resmat' which are members in the gene-set specified by 'GSname'. The summary is by column. For each level of 'fac' there will be a separate pane.

'mnDiffPlot' calls plot; it plots the mean differences, by row, between columns belonging to the two groups specified by 'fac', as a function of the mean values for the first group alone. Each row translates to a single point on the graph. Again, the summary is only for rows indicated by 'GSname'.

For gene-set selective plots to properly work, the incidence matrix needs to have non-empty row names, and 'GSname' must match one of them.

If both 'GSname' and 'incidence' are left blank, automatic utilities are called which help generate a summary of the entire matrix, by column.

All functions plot a reference line signalling zero. 'mnDiffPlot' also optionally plots a loess fit for expression differences (if 'fitline=TRUE').

Note

One can use 'resplot'/'restrip' to plot raw expression values rather than residuals; it all depends on what's in the data matrix.

Author(s)

Assaf Oron

See Also

boxplot,plot,stripchart,par,GOmnplot

Examples

data(sample.ExpressionSet)
lm1 = lmPerGene(sample.ExpressionSet,~sex)
r1 = getResidPerGene(lm1)
### now a boxplot of all residuals by sample
resplot(resmat=exprs(r1),fac=sample.ExpressionSet$sex)
### This plot is not very informative because of some gross outliers;
### try this instead
resplot(resmat=exprs(r1),fac=sample.ExpressionSet$sex,lims=c(-5,5))

### stripchart for first 10 genes
restrip(resmat=exprs(r1)[1:10,],fac=sample.ExpressionSet$type,prefix="Not")

### note the wild trajectory of the loess fit:
mnDiffPlot(exprmat=exprs(sample.ExpressionSet),fac=sample.ExpressionSet$type,xname="Raw Expression",yname="Expression Difference",fitline=TRUE)

GSEAlm documentation built on Nov. 8, 2020, 5:13 p.m.