How much each gene's effective length varies between samples
Source:R/rsem_length_profile.R
rsem_length_profile.RdSummarizes the gene-level effective lengths from
rsem_read(expr = "length"). RSEM's gene effective length is an average of
its transcripts' effective lengths weighted by their abundance, so it
changes between samples when transcript usage or the fragment length
distribution changes, and it is 0 where every transcript is shorter than
the mean fragment length. Large variation is therefore not by itself a sign
of an error or of poor quality.
Value
A data frame with one row per gene: gene, median, mean,
cv (standard deviation / mean, over all samples), zero_fraction
(share of samples with length 0) and max_min_ratio (largest / smallest
nonzero length; NA with fewer than two nonzero samples).
Details
Genes with length 0 in some samples are the ones that stop
tximport/DESeq2::DESeqDataSetFromTximport(); zero_fraction > 0 finds
them.
Examples
len <- data.frame(S1 = c(1000, 0, 500), S2 = c(1100, 200, 500),
row.names = c("A", "B", "C"))
rsem_length_profile(len)
#> rsem_length_profile: 3 genes x 2 samples; length 0 in some samples: 1 genes, in every sample: 0 genes
#> gene median mean cv zero_fraction max_min_ratio
#> 1 A 1050 1050 0.0673435 0.0 1.1
#> 2 B 100 100 1.4142136 0.5 NA
#> 3 C 500 500 0.0000000 0.0 1.0