Figure 4

The score matrix for dynamic programming peak alignment. The score matrices for the alignment of two peak lists obtained from the GC-MS segments shown in Figure 3. Three matrices are shown with peak similarity (color coding is according to the scale between 0 and 1). For each score matrix the two axis refer to two peak lists to be aligned, with 14 signal peaks in each list (stationary phase peaks plotted along the x-axis and mid-log phase peaks plotted on the y-axis). The top panel shows the score matrix based on the retention time similarities only (as given the second term in the Equation [2]). The middle panel shows the score matrix based on the similarities in mass spectra taken at the peak apex (as given by the first term in the Equation [2]). The bottom panel is the total peak similarity function, as given by the Equation [2] (this is element-by-element product of the retention time and mass spectra score matrices). The traceback resulting from the application of dynamic programming is also shown in bottom panel. From the traceback the best alignment of peaks was deduced, as shown in Figure 3(b).