MALDI-TOF Baseline Drift Removal Using Stochastic Bernstein Approximation
- PDF / 1,215,582 Bytes
- 9 Pages / 600.03 x 792 pts Page_size
- 38 Downloads / 141 Views
MALDI-TOF Baseline Drift Removal Using Stochastic Bernstein Approximation Joseph Kolibal1 and Daniel Howard2 1 Department
of Mathematics, College of Science & Technology, The University of Southern Mississippi, Hattiesburg, MS 39406-0001, USA 2 QinetiQ PLC, Malvern, Worcestershire WR14 3PS, United Kingdom Received 7 July 2005; Revised 21 August 2005; Accepted 1 December 2005 Stochastic Bernstein (SB) approximation can tackle the problem of baseline drift correction of instrumentation data. This is demonstrated for spectral data: matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF) data. Two SB schemes for removing the baseline drift are presented: iterative and direct. Following an explanation of the origin of the MALDI-TOF baseline drift that sheds light on the inherent difficulty of its removal by chemical means, SB baseline drift removal is illustrated for both proteomics and genomics MALDI-TOF data sets. SB is an elegant signal processing method to obtain a numerically straightforward baseline shift removal method as it includes a free parameter σ(x) that can be optimized for different baseline drift removal applications. Therefore, research that determines putative biomarkers from the spectral data might benefit from a sensitivity analysis to the underlying spectral measurement that is made possible by varying the SB free parameter. This can be manually tuned (for constant σ) or tuned with evolutionary computation (for σ(x)). Copyright © 2006 Hindawi Publishing Corporation. All rights reserved.
1.
INTRODUCTION
Each measurement analysis tool for determining the presence and concentration of biomolecules has its particular signal processing challenge. Consider some of these challenges for two of the most powerful tools: microarray analysis and spectral analysis. For example, the proximity of dots in a microarray can cause a degree of correlation between neighboring dots that must be removed with signal processing. With spectral analysis, typical signal processing challenges are (a) baseline drift correction; (b) denoising by smoothing and averaging of signals; (c) peak alignment; and (d) peak identification. This paper tackles baseline drift correction with algorithms that are based on a recent method of signal processing, stochastic Bernstein (SB) approximation [1]. Although baseline drift correction is illustrated with respect to matrixassisted laser desorption/ionization time-of-flight (MALDITOF) [2] data, our approach has much wider application. Other types of spectral data suffer from baseline drift and, potentially, this technique can also assist with a variety of instrumentation (not necessarily in the bioinformatics domain) that suffers from baseline drift (e.g., [3]). Consider MALDI-TOF and baseline drift. For instrumental reasons that are not easy to control, multiple MALDI-TOF measurements on the same biological sample
can result in curves at different heights. The drifted baselines must be corrected before comparing peak intensities. Section 2 discusses concepts that
Data Loading...
