Figure 7

Robustness assessment of the fold change estimate. a) To assess the robustness of the model's fit to the observed spectra, one data point (Direct:S_D0-S_D8, Inverted:S_I0-S_I8) at a time has been removed and the resulting fold change recalculated. The graph shows for each experimentally observed spectra pair (x-axis) how the estimated fold change varies when one of the spectra peaks is missing. Overall only small degrees of variation upon data removal were observed, thereby demonstrating the redundancy of the spectra and robustness of the fold change estimate. The highlighted case where the fold change varies considerably is due to very low labeling efficiency (LabEff). A fold change saturation limit of +/- 5 was imposed for proper visualization. b) To challenge the method's robustness even further multiple data points have been removed. This has been done by removing peaks in pairs starting from the on-average least intense (R:S₈ means removing S_D8 and S_I8) and moving up (R:S₈S₇ means removing S_D7, S_I7, S_D8 and S_I8) etc. Overall it is seen that peptides with a high fold change tend to be more susceptible to missing values. The method appears to handle up to four missing values well, but in most of the observed cases even six missing values result in reasonable fold change values.