On the preprocessing of mass spectrometry proteomics data

Abstract

Mass-Spectrometry (MS) based biological analysis is a powerful approach for discovering novel biomarkers or identifying patterns and associations in biological samples. Each value of a spectrum is composed of two measurements, m/Z (mass to charge ratio) and intensity. Even if data produced by mass spectrometers contains potentially huge amount of information, data are often affected by errors and noise due to sample preparation and instrument approximation. Preprocessing consists of (possibly) eliminating noise from spectra and identifying significant values (peaks). Preprocessing techniques need to be applied before performing analysis: cleaned spectra may then be analyzed by using data mining techniques or can be compared with known spectra in databases. This paper surveys different techniques for spectra preprocessing, working either on a single spectrum, or on an entire data set. We analyze preprocessing techniques aiming to correct intensity and m/Z values in order to: (i) reduce noise, (ii) reduce amount of data, and (iii) make spectra comparable. © Springer-Verlag Berlin Heidelberg 2006.