Critical assessment of proteome-wide label-free absolute abundance estimation strategies

There is a great interest in reliable ways to obtain absolute protein abundances at a proteome-wide scale. To this end, label-free LC-MS/MS quantification methods have been proposed where all identified proteins are assigned an estimated abundance. Several variants of this quantification approach have been presented, based on either the number of spectral counts per protein or MS1 peak intensities. Equipped with several datasets representing real biological environments, containing a high number of accurately quantified reference proteins, we evaluate five popular low-cost and easily implemented quantification methods (Absolute Protein Expression, Exponentially Modified Protein Abundance Index, Intensity-Based Absolute Quantification Index, Top3, and MeanInt). Our results demonstrate considerably improved abundance estimates upon implementing accurately quantified reference proteins; that is, using spiked in stable isotope labeled standard peptides or a standard protein mix, to generate a properly calibrated quantification model. We show that only the Top3 method is directly proportional to protein abundance over the full quantification range and is the preferred method in the absence of reference protein measurements. Additionally, we demonstrate that spectral count based quantification methods are associated with higher errors than MS1 peak intensity based methods. Furthermore, we investigate the impact of miscleaved, modified, and shared peptides as well as protein size and the number of employed reference proteins on quantification accuracy.