Suggest how many unwanted factors RUVg should remove

For each k from 0 to max_k, corrects the log2 positive-normalised counts with RUVg, using the housekeeping genes as controls, and reports the mean interquartile range of the relative log expression (RLE) and the share of variance on the first principal component. The suggested k is the smallest whose RLE spread is within 5 % of the smallest one: removing more factors risks removing biology.

Usage

suggest_ruv_k(x, max_k = 5)

Arguments

x

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

max_k

The largest k to try. It is lowered to one less than the number of housekeeping genes, and to two less than the number of samples, when those are smaller. A k above that cap fits the samples exactly and flattens the RLE spread, so it is never tried. With one or two samples only k = 0 is tried.

Value

A data frame with k, rle_iqr, pc1_variance and suggested. pc1_variance is NA when the corrected values have no variance, as with one sample.

Examples

data(GSE74821)
suggest_ruv_k(GSE74821)
#>   k  rle_iqr pc1_variance suggested
#> 1 0 1.333779    0.3216721     FALSE
#> 2 1 1.287619    0.3542053     FALSE
#> 3 2 1.215446    0.3354141     FALSE
#> 4 3 1.195121    0.3528246     FALSE
#> 5 4 1.154325    0.3223319      TRUE
#> 6 5 1.121317    0.3358693     FALSE