pcmpApp: Define a shiny app for comparing projections
In vjcitn/pcmp: projection comparison app

Description Usage Arguments Value Note Examples

Define a shiny app for comparing projections

1 2	pcmpApp(sce, assayind = 1, cellIdTag = ".cellid", selectionPrefix = "sel_", numDiscGenes = 20)

`sce`	SingleCellExperiment instance with reducedDims populated with various candidates for projection from assay data –
`assayind`	numeric(1) 'i' argument to assay(sce, i), defaults to 1, for extracting expression values
`cellIdTag`	character(1) defaulting to ".cellid", added to colData(sce) for tracking selections
`selectionPrefix`	character(1) defaulting to "sel_", prefix for the name of a selection indicator variable added to colData(sce)
`numDiscGenes`	numeric(1) used to limit number of differentially expressed genes tabulated per selection

will return an updated SingleCellExperiment

In sce300xx, names(reducedDims) are 'PPCA', 'UMAP*', 'TSNE*' for projections generated by irlba::irlba, umap::umap, and Rtsne::Rtsne respectively. Prior to pcmp version 0.4.0, the projections were computed and prepared differently. As Rtsne no longer allows dims > 3, we have separately computed and stored t-SNE projections into 2- and 3- dimensional spaces, labeled TSNE2 and TSNE3. (In previous versions of pcmp, the 4-d t-SNE was computed and marginal planar projections were provided.) Note that tabulation of DE genes can be very slow if the input SingleCellExperiment has more than 10k rows.

if (interactive()) {
 message("be sure to use example(pcmpApp, ask=FALSE)")
 lit = pcmp::sce300xx
 sds = rowSds(assay(lit))
 top1k = order(sds,decreasing=TRUE)[1:1000]
 lit = lit[top1k,]
 chk = pcmpApp(lit) # fires up browser, if ask != FALSE you must
 # assent to each update 
 try(chk)  # may error if no selections
 try(head(metadata(chk)$limmaTabs,3))  # likewise
 try(table(metadata(chk)$limmaTabs$featid))  # likewise
}