preprocess: Prepares Data for IDR Analysis

In kkrismer/idr2d: Irreproducible Discovery Rate for Genomic Interactions Data

Prepares Data for IDR Analysis

Description

This method removes invalid values, establishes the correct ranking, and breaks ties prior to IDR analysis.

Inf and -Inf are replaced by max(x) * max_factor and min(x) / max_factor, respectively.

NA values in x are replaced by mean(x).

All values in x are transformed using the transformation specified in value_transformation.

Lastly, a small amount of noise is added to x to break ties. The magnitude of the noise is controlled by jitter_factor.

Usage

preprocess(
  x,
  value_transformation = c("identity", "additive_inverse", "multiplicative_inverse",
    "log", "log_additive_inverse"),
  max_factor = 1.5,
  jitter_factor = 1e-04
)

Arguments

x	numeric vector of values
value_transformation	the values in x have to be transformed in a way such that when ordered in descending order, more significant interactions end up on top of the list. If the values in x are p-values, "log_additive_inverse" is recommended. The following transformations are supported: "identity" no transformation is performed on x "additive_inverse" x. = -x "multiplicative_inverse" x. = 1 / x "log" x. = log(x). Note: zeros are replaced by .Machine$double.xmin "log_additive_inverse" x. = -log(x), recommended if x are p-values. Note: zeros are replaced by .Machine$double.xmin either "ascending" (more significant interactions have lower value in value column) or "descending" (more significant interactions have higher value in value column)
max_factor	numeric; controls the replacement values for Inf and -Inf. Inf are replaced by max(x) * max_factor and -Inf are replaced by min(x) / max_factor.
jitter_factor	numeric; controls the magnitude of the noise that is added to x. This is done to break ties in x. Set jitter_factor = NULL for no jitter.

Value

numeric vector; transformed and stripped values of x, ready for IDR analysis

Examples

rep1_df <- idr2d:::chiapet$rep1_df
rep1_df$fdr <- preprocess(rep1_df$fdr, "log_additive_inverse")

kkrismer/idr2d documentation built on Feb. 7, 2024, 2:23 p.m.