groupComparisonQCPlots: Visualization for model-based quality control in fitting...

In Vitek-Lab/MSstats: Protein Significance Analysis in DDA, SRM and DIA for Label-free or Label-based Proteomics Experiments

Visualization for model-based quality control in fitting model

Description

To check the assumption of linear model for whole plot inference, groupComparisonQCPlots takes the results after fitting models from function (groupComparison) as input and automatically generate two types of figures in pdf files as output: (1) normal quantile-quantile plot (specify "QQPlot" in option type) for checking normally distributed errors.; (2) residual plot (specify "ResidualPlot" in option type).

Usage

groupComparisonQCPlots(
  data,
  type,
  axis.size = 10,
  dot.size = 3,
  width = 10,
  height = 10,
  which.Protein = "all",
  address = ""
)

Arguments

data	output from function groupComparison.
type	choice of visualization. "QQPlots" represents normal quantile-quantile plot for each protein after fitting models. "ResidualPlots" represents a plot of residuals versus fitted values for each protein in the dataset.
axis.size	size of axes labels. Default is 10.
dot.size	size of points in the graph for residual plots and QQ plots. Default is 3.
width	width of the saved file. Default is 10.
height	height of the saved file. Default is 10.
which.Protein	Protein list to draw plots. List can be names of Proteins or order numbers of Proteins from levels(testResultOneComparison$ComparisonResult$Protein). Default is "all", which generates all plots for each protein.
address	name that will serve as a prefix to the name of output file.

Details

Results based on statistical models for whole plot level inference are accurate as long as the assumptions of the model are met. The model assumes that the measurement errors are normally distributed with mean 0 and constant variance. The assumption of a constant variance can be checked by examining the residuals from the model.

• QQPlots : a normal quantile-quantile plot for each protein is generated in order to check whether the errors are well approximated by a normal distribution. If points fall approximately along a straight line, then the assumption is appropriate for that protein. Only large deviations from the line are problematic.

• ResidualPlots : The plots of residuals against predicted(fitted) values. If it shows a random scatter, then the assumption is appropriate.

Value

produce a pdf file

Examples

QuantData <- dataProcess(SRMRawData, use_log_file = FALSE)
head(QuantData$FeatureLevelData)
levels(QuantData$FeatureLevelData$GROUP)
comparison <- matrix(c(-1,0,0,0,0,0,1,0,0,0),nrow=1)
row.names(comparison) <- "T7-T1"
colnames(comparison) <- unique(QuantData$ProteinLevelData$GROUP)
# Tests for differentially abundant proteins with models:
# label-based SRM experiment with expanded scope of biological replication.
testResultOneComparison <- groupComparison(contrast.matrix=comparison, data=QuantData,
use_log_file = FALSE)
# normal quantile-quantile plots
groupComparisonQCPlots(data=testResultOneComparison, type="QQPlots", address="")
# residual plots
groupComparisonQCPlots(data=testResultOneComparison, type="ResidualPlots", address="")

Vitek-Lab/MSstats documentation built on Nov. 4, 2024, 3:08 p.m.