Skip to contents

Sums the values of each gene type (from a GTF file) per sample and divides by the sample total, for example to see how much of the TPM goes to rRNA, mitochondrial genes or lncRNA. Genes that the annotation does not know are counted as "unannotated", and genes whose name has several types as "multiple".

Usage

rsem_biotype_profile(
  x,
  annotation,
  types = NULL,
  ids = attr(x, "gene_ids", exact = TRUE)
)

Arguments

x

The gene by sample table (TPM or counts), as returned by rsem_read().

annotation

Path to a GTF file (plain or .gz; GENCODE, Ensembl or NCBI, with gene_type or gene_biotype), or a data frame with the columns gene_id, gene_type and optionally gene_name. It should be the annotation the RSEM reference was built from.

types

Gene types to report (values of gene_type, such as "rRNA" or "protein_coding"). Default: every type found in x.

ids

The attribute "gene_ids" of the table rsem_read() returned (default: that attribute of x, when present). Rows are then matched to the annotation through their original gene IDs, which avoids losing genes whose OrgDb symbol differs from the GTF gene name. Without it, rows are matched by Ensembl ID or by gene name.

Value

A data frame with one row per sample and type: sample_id, type, genes (genes of that type in x), value (sum) and share (value / sample total).

Details

A type that the annotation does not contain cannot be measured: if a requested type has no gene in the annotation, its share is NA, not 0, and a message says so. For example, an annotation without rRNA genes says nothing about rRNA contamination. Reads from genes outside the annotation were never quantified by RSEM and are not in any share.

Examples

ann <- data.frame(gene_id = c("ENSG01", "ENSG02", "ENSG03"),
                  gene_name = c("A", "B", "C"),
                  gene_type = c("protein_coding", "rRNA", "lncRNA"))
x <- data.frame(S1 = c(10, 5, 1), S2 = c(20, 0, 2), row.names = c("A", "B", "C"))
rsem_biotype_profile(x, ann)
#> rsem_biotype_profile: 3 genes x 2 samples; 0 of the genes not in the annotation
#>   sample_id           type genes value      share
#> 1        S1         lncRNA     1     1 0.06250000
#> 2        S2         lncRNA     1     2 0.09090909
#> 3        S1 protein_coding     1    10 0.62500000
#> 4        S2 protein_coding     1    20 0.90909091
#> 5        S1           rRNA     1     5 0.31250000
#> 6        S2           rRNA     1     0 0.00000000