Figure 1.

Self-convolution of spectra in the training data set. A spectrum self-convolution, as described by Dancík et al.,²⁷ is the product of a spectrum and its reflection. Formally, eq 1 describes it as the product of intensity of a peak and its cognate. Eq 2 introduces a mass offset, O, applied to the cognate peak. In this figure, O is plotted along the x-axis. The y-axis represent the value of the convolution in intensity units. As the self-convolution in eq 2 is applied to many spectra (all spectra in the training set), frequently observed offsets stand out. x = 0 represents the matching of b and y ions. The peak at x=1 represents matching of an isotope to the b/y ladder: b + 1 and y, or b and y + 1. The peaks at −18, −17 and −98 correspond to the neutral losses of water and ammonia and phosphate. The peak for phosphate loss is not used in the final model. See Results and Discussion for possible explanation.