filter_outliers: Filter lowly abundant features

In PhilipBerg/PaiR: Imputation and Significance Analysis of Proteomics Data

Filter lowly abundant features

Description

Function for filtering lowly abundant features. By default, it uses all numerical columns. Missing values are always considered as outliers.

Usage

filter_outliers(data, target = NULL, percent = 1, k = 1.5, lower_limit = NULL)

Arguments

data	data to filter featuers from.
target	columns to base the filtering on, supports tidyselect-package.
percent	A feature gets filtered out if it is lowly abundant or missing in percent columns.
k	Parameter for the lower limit of Tukey's fence, any value bellow this will be considered an outlier.
lower_limit	a user defined lower limit at which a measurement is considered an outlier.

Value

data with outliers removed

Examples

# Since Tukey's fences are not ideal for raw proteomics data one could use
# the e.g., the tenth percentile as a indicator of lower abundance
filter_outliers(yeast, lower_limit = stats::quantile(yeast[-1], .1, na.rm = TRUE))

# We recommend normalizing the data before filtering outliers with Tukey's fences.
# This way we ensure that no peptides are considered outliers as an effect
# of a set of samples, one average, have lower quantification or that the
# lower fence is smaller then the smallest value in the dataset
yeast <- psrn(yeast, "identifier")
filter_outliers(yeast, -1, 1, 1.5)

PhilipBerg/PaiR documentation built on March 18, 2022, noon