Peak pattern deconvolution for Protein Mass Spectrometry by non-negative ls/lad template matching

Share:

Description

The package provides functionality to extract isotopic peak patterns from raw mass spectra. This is done by fitting a large set of template basis functions to the raw spectrum using nonnegative least squares (ls) or nonnegative least absolute deviation (lad). Ideally, the nonnegativity constraint in combination with nonnegativity of the template basis functions effects that templates not matching the data are assigned an extremely low weight such that one can easily identify isotopic patterns present and not present in the spectrum. In practice, the templates only approximate the peak patterns, where the quality the approximation crucially depends on how well the shapes of the templates fit the isotopic patterns contained in a spectrum. For this reason, the package offers the flexible function fitModelParameters which tries to estimate model parameters, e.g. the width of a Gaussian bump, in a way tailored to the peak shapes in the data. As second peak model in addition to the standard Gaussian, the package offers full support for the Exponential Modified Gaussian.
The function getPeaklist predicts the set of isotopic peak patterns present in the spectrum in a fully automatic, yet customizable way. The main benefits of our approach are that

  1. Overlapping peak patterns can be resolved.

  2. The complete spectrum can be processed as a whole or in large sections by exploiting the sparse nature of the problem.

  3. The set of parameters in getPeaklist are easy to interpret and require only very basic knowledge of statistics.

  4. A theoretically well-founded post-processing procedure is used.

  5. The result can be analyzed visually in a detailed way using the function visualize.

Package: IPPD
Type: Package
Version: 1.3.1
Date: 2012-01-17
License: GPL (version 2 or later)

Author(s)

Martin Slawski ms@cs.uni-saarland.de,
Rene Hussong rene.hussong@uni.lu,
Andreas Hildebrandt andreas.hildebrandt@uni-mainz.de,
Matthias Hein hein@cs.uni-saarland.de

Maintainer: Martin Slawski ms@cs.uni-saarland.de.