design_filter: Filter low expressed transcripts by taking into account the...

In nodrogluap/gumshoe: Supplemental Statistical Tests and Utilities for Kallisto/Sleuth RNASeq Analysis

Filter low expressed transcripts by taking into account the regression model.

Description

Filter low expressed transcripts by taking into account the regression model.

Usage

design_filter(meta, model, row, min_reads = 5, min_prop = 0.47)

Arguments

meta	the same data frame that would be passed as the first argument to sleuth_prep
model	the regression model that would be passed as the second argument to sleuth_prep
row	the count data for a given transcript, as fed by sleuth_prep to the filter_fun method
min_reads	minimum number of reads required to consider a transcript present in a sample (default value meant to match the behaviour of the default filter in Sleuth called basic_filter)
min_prop	minimum proportion of samples *in a given experimental factor combination* required to consider a transcript present in the experiment (default value meant to match the behaviour of the default filter in Sleuth called basic_filter although it looks for the proportion across ALL sample)

Value

true or false, indicating whether or not the transcript in question passes the filter criteria

Examples

# Given a sample metadata table with the columns \code{sample}, \code{path}, \code{treatment}, and \code{sex}
# create a model with the two factors plus their interaction. Use a design filter so that for example a transcript that only shows up in treated males (1/4 of the dataset) will not be filtered out
# (as it would by the default filter which would require half of ALL sample to express the transcript).
model <- ~treatment*sex
so <- sleuth_prep(sample_meta, model, filter_fun=function(x){design_filter(sample_meta, model, x)})

nodrogluap/gumshoe documentation built on Feb. 28, 2023, 6:15 p.m.