Keep genes detected in enough samples

A gene is detected in a sample when its raw count is above the mean plus two standard deviations of that sample's negative controls. Control and housekeeping probes are always kept, where housekeeping probes are the probes of CodeClass Housekeeping. Endogenous genes chosen as housekeeping genes are filtered like any other gene, and the dropped ones are removed from the housekeeping_genes setting, which becomes NULL when none is left. Then the object uses its Housekeeping probes and warns, or, with no Housekeeping probe either, turns housekeeping_norm off and warns. The samples' Detection_rate and Housekeeping_detected are not recomputed after filtering. A gene's detection rate is the share of the samples with a detection limit in which it is detected, so a sample with fewer than two negative probes does not count. Endogenous genes whose detection rate is missing, because every sample with a detection limit has a missing count for them, are dropped, even with min_rate = 0. With all-zero negative controls the detection limit is 0, so every non-zero count counts as detected. An object without endogenous genes comes back unchanged. When no sample has a detection limit, filter_detected() stops with an error.

Usage

filter_detected(x, min_rate = 0.5)

Arguments

x

A nacho object from load_rcc() or normalise().

min_rate

The smallest share of samples with a detection limit, between 0 and 1, in which an endogenous gene must be detected.

Value

A nacho object with fewer probes.

Examples

data(GSE74821)
dim(filter_detected(GSE74821, min_rate = 0.5))
#> [1] 72 48