Fig. 9.

The log₁₀ frequency of MS/MS correlations by X!TANDEM from authentic endogenous tryptic peptides from 563,095 MS/MS spectra collected from EDTA plasma versus the frequency from an equal number of random MS/MS spectra for the estimation of type I error rate in the proteins identified by Chi Square. The corrected mean peptide-to-protein counts of the first 200 protein accessions from X!TANDEM are shown. Note the peptide-to-protein distribution of real spectra exceeds that of random spectra by ≥ order of magnitude at all values and the Chi Square test indicates that the probability that the experimental data is the same as the null random model is essentially zero (p < 0.0001) and is a formal estimate of the type I error with respect to a null random model for the whole experiment using the classical statistical approach. The result indicates that the X!TANDEM algorithm that fits MS/MS spectra to predicted peptide spectra shows an acceptable type I error rate over the whole experiment