Peak Alignment Sample Clauses

Peak Alignment. After a set of suitable spectral peaks have been selected from multiple data sets, they need to be aligned before they can be compared against each other or combined to form an average peak profile. The variations in peak location between data sets generated by the same MS protocol have been well demonstrated in several studies (Torgrip et al., (2003)[117], Xxxxxx (2004)[42]). According to Xx et al., (2006)[122], this variation exists even among the selected peak profiles of technical replicates, i.e., replicates of the same sample run under similar conditions. The reasons for this variability are not well understood and are quite often not of primary interest, at least as far as the functionality of the peak alignment algorithms are concerned. Current peak alignment algorithms range from fairly simple to extremely com- plicated. Xxxxxxx et al., (2003) assumes that peak variation is less than the typical distance between peaks and uses a closest point matching method for peak align- ment. Xx et al., (2006)[122] uses the same concept of peak distance to address the alignment of multiple peak sets. Xxxxxxx et al., (2003)[27] align spectral peaks by pooling the list of detected peaks that differed in location by three clock ticks in the retention time axis or by 0.05% of the mass. Xxxxx et al., (2003a)[120] extrapolate each peak to its local neighborhood with a peak width equal to 0.4% of the m/z value of the middle point based on the observation that the shift of peaks is approx- imately 0.1% to 0.2% of the corresponding m/z value. In a separate study, Xxxxx et al., (2003b)[121] first calculates the number of peaks in all samples allowing certain shifts, and selected m/z values with the largest number of peaks. Next, this set of peaks is removed from all spectra and the procedure is iterated until all peaks are exhausted from all samples. Xxxxxxxxxx et al., (2004)[115] propose the use of com- plete linkage hierarchical clustering to cluster peaks. All peaks that fall into the same cluster are considered to represent the same peak. Xxxxxxxx and Xxxxx (2006)[99] uses wavelets to represent the MS data in a multiscale frequency domain and use a coarse-to-fine decomposition method to first align peaks at a dominant scale and then refine the alignment of other peaks at a finer scale. Xxxxxx (2004)[42] propose a para- metric warping model with polynomial or spline functions to align chromatograms by adding calibration sequences into chromatograms. Peak alignment ap...