Identify the best rank cut-off for significant CpGs.
rankDist.RdAutomated rank cut-off estimator for input CpGs.
rankDist(ranks, draw_intersects, noise_perc, mode, return.plot)Arguments
- ranks
getPCRanks output data frame.
- draw_intersects
T/F whether to draw intersect lines if return.plot=T
- noise_perc
Automatic=0.5, numeric between 0 and 1. Fraction of ranks to use to model the background noise. Not recommended to play with this value. Increasing/decreasing returns a looser/stricter threshold, respectively.
- mode
"intersect" or "strict", determine cut-off with "intersect" or "strict" method. "Strict" is recommended for sets with lower variability
- return.plot
T/F, whether to return a plot or a numeric
Value
If return.plot=T, a grob plotting the estimated cutoff on a plot of absolute eigenvector score vs. absolute rank order is returned. Otherwise, a numeric of the estimated cut-off is returned.
Examples
ranks <- getPCRanks(eigen, IDs = c("trt", "ctl"), PC = 1)
rankDist(ranks, mode="intersect")
#> Estimated rank cut-off for significant CpGs is 980.
