Nonparametric Pre-processing Methods and Inference Tools for Analyzing Time-of-flight Mass Spectrometry Data

Anestis Antoniadis∗ , Sophie Lambert-Lacroix and Frédérique Letué, Laboratoire IMAG-LMC, University Joseph Fourier, BP 53, 38041 Grenoble Cedex 9, France and Jérémie Bigot University Paul Sabatier, Toulouse, France. Abstract The objective of this paper is to contribute to the methodology available for extracting and analyzing signal content from protein mass spectrometry data. Data from MALDI-TOF or SELDI-TOF spectra require considerable signal pre-processing such as noise removal and baseline level error correction. After removing the noise by an invariant wavelet transform, we develop a background correction method based on penalized spline quantile regression and apply it to MALDI-TOF (matrix assisted laser deabsorbtion time-of-flight) spectra obtained from serum samples. The results show that the wavelet transform technique combined with nonparametric quantile regression can handle all kinds of background and low signal-to-background ratio spectra; it requires no prior knowledge about the spectra composition, no selection of suitable background correction points, and nomathematical assumption of the background distribution. We further present a multi-scale based novel spectra alignment methodology useful in a functional analysis of variance method for identifying proteins that are differentially expressed between different type tissues. Our approaches are compared with several existing approaches in the recent literature and are tested on simulated and some real data. The results indicate that the proposed schemes enable accurate diagnosis based on the over-expression of a small number of identified proteins with high sensitivity.