Comparing the Diagnostic Classification Accuracy of iTRAQ, Peak-Area, Spectral-Counting, and emPAI Methods for Relative Quantification in Expression Proteomics

Diagnostic classification
accuracy is critical in expression proteomics
to ensure that as many true differences as possible are identified
with acceptable false-positive rates. We present a comparison of the
diagnostic accuracy of iTRAQ with three label-free methods, peak area,
spectral counting, and emPAI, for relative quantification using a
spiked proteome standard. We provide the first validation of emPAI
for intersample relative quantification and find clear differences
among the four quantification approaches that could be considered
when designing an experiment. Spectral counting was observed to perform
surprisingly well in all regards. Peak area performed best for smaller
fold differences and was shown to be capable of discerning a 1.1-fold
difference with acceptable specificity and sensitivity. The performance
of iTRAQ was dramatically worse than the label-free methods with low
abundance proteins. Using the iTRAQ data set for validation, we also
demonstrate a novel iTRAQ analysis regime that avoids the use of ratios
in significance testing and outperforms a common commercial alternative.