Compute PI (Percent Isoform) Matrix from Isoform Counts

This function calculates per-cell isoform usage (PI) values from an isoform-level count matrix stored in a Seurat object. For each gene, isoform counts are divided by the total counts of all isoforms belonging to that gene within each cell.

Usage

computePIMatrix(seurat_obj, assay = "ISO", slot = "counts")

Arguments

seurat_obj: A Seurat object containing isoform-level counts in one assay.
assay: Character string specifying the assay that contains isoform counts. Default is "ISO".
slot: Character string indicating which slot of the assay to use (e.g., "counts" or "data"). Default is "counts".

Value

A matrix of PI values (isoforms - cells) with NA where total gene counts are zero.

Details

The function expects rownames in the form GeneName..TranscriptID. The part before the first double dot (..) is interpreted as the gene identifier. PI values are computed per isoform and per cell as: $$PI = count(isoform) / sum(counts of all isoforms from the same gene)$$

Cells with zero total counts for a given gene return NA PI values for its isoforms.

Examples

library(Seurat)
#> Loading required package: SeuratObject
#> Loading required package: sp
#> 
#> Attaching package: ‘SeuratObject’
#> The following objects are masked from ‘package:base’:
#> 
#>     intersect, t
iso_names <- c("G1..Tx1", "G1..Tx2", "G2..Tx1", "G2..Tx2")
counts <- matrix(c(10, 5, 3, 7, 0, 4, 2, 6),
                 nrow = 4,
                 dimnames = list(iso_names, paste0("Cell", 1:2)))
seurat_obj <- CreateSeuratObject(counts = counts, assay = "ISO")
#> Warning: Data is of class matrix. Coercing to dgCMatrix.

pi <- computePIMatrix(seurat_obj)
#> Warning: The `slot` argument of `GetAssayData()` is deprecated as of SeuratObject 5.0.0.
#> ℹ Please use the `layer` argument instead.
#> ℹ The deprecated feature was likely used in the scIso package.
#>   Please report the issue at <https://github.com/alihamraoui/scIso/issues>.
#> [computePIMatrix] Extracted assay 'ISO' (4 isoforms - 2 cells).
#> [computePIMatrix] PI matrix computed successfully (dim: 4 - 2).
head(pi)
#>             Cell1 Cell2
#> G1..Tx1 0.6666667  0.00
#> G1..Tx2 0.3333333  1.00
#> G2..Tx1 0.3000000  0.25
#> G2..Tx2 0.7000000  0.75