ANOVA: Data set ANOVA

In DanteR: Proteomics data analysis tool

Description

Perform ANOVA on peptide crosstab data, possibly using additional row-wise factors.

Usage

ANOVA(dataset, split.by.row.metadata = FALSE, split.row.metadata.field = NULL, column.metadata.fields = NULL, do.interactions = FALSE, block.order.function.name = "median", nrow.block.min = 1, nrow.block.max = 5, useglm = "lm", use.weight = FALSE, weight.function = "NULL", weight.par = 0, formula.string = ".", do.residuals = FALSE, first.level = character(0), first.level.contrasts = character(0), progressbar = NULL, progresslabel = NULL, return.all.fits = FALSE, return.residuals = FALSE, ...)

Arguments

dataset	A data frame or array containing numerical data.
split.by.row.metadata	Do ANOVA row by row, or 2-factor ANOVA on entire blocks specified by row metadata?
split.row.metadata.field	Field corresponding to proteins
column.metadata.fields	Factors to use in ANOVA
do.interactions	Use interactions between factors
block.order.function.name	How to order blocks of peptides within a protein
nrow.block.min	Min number of rows per block (peptides per protein). Turn this to 2 to exclude one hit wonders.
nrow.block.max	Max number of rows per block (peptides per protein)
useglm	Type of model to fit the data to
use.weight	Weight the model or not?
weight.function	String containing the weighting function
weight.par	String containing the weighting parameter
formula.string	String containing the formula
do.residuals	Return a list of all residuals for the fit
first.level	Specify first level of factor - for one factor only
first.level.contrasts	Contrast scheme for first factor
progressbar	Internal argument for DanteR
progresslabel	Internal argument for DanteR
return.all.fits	Return a list of all fitted models
return.residuals	Return an array of residuals for the fit
...	Additional arguments

Details

There is a comprehensive ANOVA scheme included in DAnTE. This model can account for unbalanced data using marginal sums of squares in the case of 2-way ANOVA (or n-Way) and can account for random effects such as LC column effects etc through a REML based multi level model. User can also check interactions in a higher order (n-Way) ANOVA. For N peptides or proteins, depending on whether protein-level ANOVA is selected or not, the output will be a Nx2K column array containing estimate sizes and p-values for each factor comparison.

The ANOVA function is performed according to Oberg et al.

For non protein level ANOVA and one factor the result is identical to a 2-sided t-test performed on each row.

For protein level ANOVA and one factor, the following steps are performed for each protein:

1. Take the N most abundant peptides by their median or mean abundance (N is defaulted to 5) 2. Fit the following linear 2-factor ANOVA model to the data:

y_ij = alpha_i + beta_(Pr[i], Tr[j]) + delta_ij

Associated with each peptide intensity is a protein identity Pr, so the i'th peptide belongs to protein Pr[i]. The recorded log-intensity of the i'th peptide belonging to the Pr[i]'th protein in the j'th experiment is y_ij. This is expected to depend on a peptide-dependent ionization efficiency i; a treatment effect beta_(Pr[i], Tr[j]) depending on the peptide's originating protein Pr[i] and the experimental treatment group Tr[j]; i.i.d. random noise delta_ij. Thus all data is fit to the linear model.

The weight function optionally weights the noise delta_ij. For LTQ LC-MS data a moderate exponential weighting y_ij ~ exp(-0.25y) fits instrument noise well.

References

Oberg, A. L.; Mahoney, D. W.; Eckel-Passow, J. E.; Malone, C. J.; Wolfinger, R. D.; Hill, E. G.; Cooper, L. T.; Onuma, O. K.; Spiro, C.; Therneau, T. M.; Bergen, H. R., 3rd J Proteome Res 2008, 7, 225.

DanteR documentation built on May 2, 2019, 6:11 p.m.