Statistical Methods in Preprocessing Proteomics Data Sample Clauses

Statistical Methods in Preprocessing Proteomics Data. The typical first step in proteomics data analysis is the removal or reduction of sys- tematic artifacts introduced by the instrumentation and experimental protocols, and by the random background fluctuations produced by chemical and electronic noise. Secondly, spectra from multiple MS runs need to be aligned with respect to reten- tion time of the compounds. These prerequisite data adjustment steps are known as preprocessing, and at a minimum, involve the following steps: spectrum calibration; base-line correction; smoothing; peak identification; and intensity normalization and Figure 2.1: Stable isotope labeling - (A) SILAC and (B) ICAT flow diagrams. peak alignment. Preprocessing generally starts with aligning individual spectra based on the max- imum observed intensity of an internal calibrant. Removal of high frequency back- ground noise is usually achieved with some form of local smoothing. For example, with MALDI data, Xx et al., (2003)[119] uses a local linear regression method to es- timate the background intensity values, and then subtract the fitted values from the local linear regression result. Xxxxxxxx et al., (2003)[13] consider a semi-monotonic baseline correction method in their analysis of SELDI data. Xxx et al., (2003)[76] compute the convex hull of the peak spectrum, and subtract the estimated convex hull from the original spectrum to get the baseline noise-corrected spectrum. In the following sections, we give a brief overview of peak selection and alignment methods, which are arguably the two most important preprocessing steps.