[IDEA] Metric: Correlation between samples

[cbielow]

e.g. using R^2 could hint at non-specific sample loss in some samples, i.e. overall reduced reproducibility and false positive biomarkers, due to different sample handling and materials (tubes, addition of compounds, adsorption, ...). This works well only for technical replicates, since different biol. conditions might obviously not be perfectly correlated.
Reference could be the sample whose median intensity is representative (rather than using the median of each peptide, since this destroys the connection between peptides)