Peak detection and alignment
If the complete amount of data from a sample is included in a data analysis this means about 10 K data points per SELDI spectra, or datafiles in the approximate size of GB if we have 2-dimensional LC/MS data. Ideally, Peak detection will reduce the data to hold only the information from the analytical signals, and the noise will be removed.
When data from different samples are to be compared, it is important that datapoints on the same position between samples corresponds to the same analyte (e.g. protein). This holds for both “raw” data and peak detected signals. A manual inspection of data will correct for signal misalignments intuitively, but a computer based multivariate data analysis is highly dependent of perfectly aligned signals.
New papers with contributions in peak detection and alignment methods are continuously published. The method of choice will highly depend on the accuracy and precision of the actual study, i.e. the analytical method (including sample preparation), the analytical signal, and also the application (samples). This is the reason why a lot of methods need to be used in parallel and no “optimal” general method are available. Hence, development of peak detection and alignment methods are also developed at KBC.
