correct_drift: Correct drift using cubic spline

In antonvsdata/notame: Workflow for non-targeted LC-MS metabolic profiling

Correct drift using cubic spline

Description

A wrapper function for applying cubic spline drift correction and saving before and after plots

Usage

correct_drift(
  object,
  log_transform = TRUE,
  spar = NULL,
  spar_lower = 0.5,
  spar_upper = 1.5,
  check_quality = FALSE,
  condition = "RSD_r < 0 & D_ratio_r < 0",
  plotting = FALSE,
  file = NULL,
  width = 16,
  height = 8,
  color = "QC",
  shape = NULL,
  color_scale = getOption("notame.color_scale_dis"),
  shape_scale = scale_shape_manual(values = c(15, 16))
)

Arguments

object	a MetaboSet object
log_transform	logical, should drift correction be done on log-transformed values? See Details
spar	smoothing parameter
spar_lower, spar_upper	lower and upper limits for the smoothing parameter
check_quality	logical, whether quality should be monitored.
condition	a character specifying the condition used to decide whether drift correction works adequately, see Details
plotting	logical, whether plots should be drawn
file	path to the PDF file where the plots should be saved
width, height	width and height of the plots in inches
color	character, name of the column used for coloring the points
shape	character, name of the column used for shape
color_scale, shape_scale	the color and shape scales as returned by a ggplot function

Details

If log_transform = TRUE, the correction will be done on log-transformed values. The correction formula depends on whether the correction is run on original values or log-transformed values. In log-space: corrected = original + mean of QCs - prediction by cubic spline. In original space: corrected = original * prediction for first QC / prediction for current point. We recommend doing the correction in the log-space since the log-transfomred data better follows the assumptions of cubic spline regression. The drift correction in the original space also sometimes results in negative values, and results in rejection of the drift corrrection procedure. If spar is set to NULL (the default), the smoothing parameter will be separately chosen for each feature from the range [spar_lower, spar_upper] using cross validation. If check_quality = TRUE, the condition parameter should be a character giving a condition compatible with dplyr::filter. The condition is applied on the changes in the quality metrics RSD, RSD_r, D_ratio and D_ratio_r. For example, the default is "RSD_r < 0 and D_ratio_r < 0", meaning that both RSD_r and D_ratio_r need to decrease in the drift correction, otherwise the drift corrected feature is discarded and the original is retained. If shape is set to NULL (the default), the column used for color is also used for shape

Value

MetaboSet object as the one supplied, with drift corrected features

See Also

dc_cubic_spline, smooth.spline for details about the regression, inspect_dc for analysing the drift correction results, save_dc_plots for plotting the drift correction process for each feature

Examples

corrected <- correct_drift(merged_sample)

antonvsdata/notame documentation built on Sept. 14, 2024, 11:09 p.m.