Parse transcript counts and additional data from salmon
Usage
digestSalmon(
paths,
max_sets = 2L,
aux_dir = "aux_info",
name_fun = basename,
verbose = TRUE,
extra_assays = NULL,
max_boot = Inf,
...
)Arguments
- paths
Vector of file paths to directories containing salmon results
- max_sets
The maximum number of indexes permitted
- aux_dir
Subdirectory where bootstraps and meta_info.json are stored
- name_fun
Function applied to paths to provide colnames in the returned object. Set to NULL or c() to disable.
- verbose
Print progress messages
- extra_assays
Can take values in c("TPM", "effectiveLength", "length") to optionally request TPM, effectiveLength or length as assays. Including the length assay is intended for the use case of personalised transcriptomes where transcript lengths may no longer be uniform across samples. None will be returned by default
- max_boot
The maximum number of bootstraps to use. Setting this to zero will ignore all bootstraps and the scaledCounts assay will not be included in the returned object
- ...
Not used
Value
A SummarizedExperiment object containing assays for counts and scaledCounts. The scaledCounts assay contains counts divided by overdispersions. rowData in the returned object will also include transcript-lengths along with the overdispersion estimates used to return the scaled counts. TPM, effectiveLength and length can be returned as additional assays by specifying one or more of these in the extra_assays argument
Details
This function is based heavily on edgeR::catchSalmon() however, there are
some important differences:
A SummarizedExperiment object is returned
Differing numbers of transcripts are allowed between samples
The second point is intended for the scenario where some samples may have been aligned to a full reference, with remaining samples aligned to a partially masked reference (e.g. chrY). This will lead to differing numbers of transcripts within each salmon index, however, common estimates of overdispersions are required for scaling transcript-level counts. By default, the function will error if >2 different sets of transcripts are detected, however this can be modified using the max_sets argument.
This greater flexibility also requires more stringent checking and, as such, for smaller datasets, digestSalmon may be slower that the edgeR function.
The SummarizedExperiment object returned may also contain multiple assays, as described elsewhere on this page