Figure 3

Fold change quality assessment. To assess the quality of the estimated fold change the average error was calculated as a function of the imposed noise level (red) and the observable error to signal ratio (ε/S-ratio in blue) calculated as part of the model fitting. The sigmoid shape of the ε/S-ratio reflects that when the imposed noise fits the model (low ε/S-ratio) the fold change error does not decrease, and when the imposed noise doesn't fit (high ε/S-ratio) the fold change error remains level. The observation that the ε/S-ratio (blue) is flat at the ends of the sigmoid reflect the fact that by chance, for low ε/S-ratios the noise accidentally fits with the isotopic contour, thus increasing the fold change error. The flat end for high ε/S-ratios is just the inverse i.e. by chance the noise pulls in opposite directions and cancels out in the linear regression fit, thus lowering the fold change error while having a large ε/S-ratio.