knitr::opts_chunk$set(dev = 'pdf')

This report presents an overview of the DEP analysis.

Summary

Original file contains r nrow(data) proteins groups, of which r nrow(dep) proteins were reproducibly quantified.

In total r ncol(data) samples were detected.

r colnames(data)

r nrow(dep[rowData(dep)$significant, ]) proteins differ significantly between samples.

Parameters used:

\pagebreak

Protein numbers

Protein identifications per sample.

pg_width = ncol(filt) / 3
if (pg_width > 10) { pg_width = 10 }
plot_numbers(filt)

Proteins coverage in all samples.

plot_coverage(filt)

\pagebreak

Normalization

The data was normalized using 'vsn'.

norm_height = ncol(filt) / 2
plot_normalization(filt, norm)

\pagebreak

Missing values

To asses the type of missing data (random or not), a heatmap of missing values is plotted. If data is randomly missing, use the "knn" option for imputation. If the missing data is biased to certain samples (e.g. controls) which are expected to be depleted of certain proteins, use the "QRILC", "MinProb" or "man" options for imputation.

plot_missval(norm)

\pagebreak

Quantitative information of proteins with and without missing values.

plot_detect(norm)

\pagebreak

Intensity distributions before and after imputation.

plot_imputation(norm, dep)

\pagebreak

Output plots of the data

PCA plot

num = 400
if (nrow(dep) < 400) num = nrow(dep)
plot_pca(dep, n = num)

Correlation matrix

plot_cor(dep)

Heatmaps

len = nrow(dep[rowData(dep)$significant, ]) / 7
if (len < 2) { len = 2 }
df <- assay(dep[rowData(dep)$significant, ]) - rowMeans(assay(dep[rowData(dep)$significant, ]))
col_lim <- quantile(df, probs= 0.95) - quantile(df, probs= 0.05)
if (len < col_lim) { col_lim = 2 }
plot_heatmap(dep, "contrast", k = 6, col_limit = (2 * col_lim))
width = ncol(filt) / 3
if (width < 5) { width = 5 }
if (width > 10) { width = 10 }
plot_heatmap(dep, "centered", k = 6, col_limit = col_lim)


squirrelandr/DEP documentation built on May 7, 2019, 9:31 a.m.