A Bayesian Model Averaging Approach to the Quantification of Overlapping Peptides in an MALDI-TOF Mass Spectrum

Abstract

In a high-resolution MALDI-TOF mass spectrum, a peptide produces multiple peaks, corresponding to the isotopic variants of the molecules. An overlap occurs when two peptides appear in the vicinity of the mass coordinate, resulting in the difficulty of quantifying the relative abundance and the exact masses of these peptides. To address the problem, two factors need to be considered: (1) the variability pertaining to the abundances of the isotopic variants (2) extra information content needed to supplement the information contained in data. We propose a Bayesian model for the incorporation of prior information. Such information exists, for example, for the distribution of the masses of peptides and the abundances of the isotopic variants. The model we develop allows for the correct estimation of the parameters of interest. The validity of the modeling approach is verified by a real-life case study from a controlled mass spectrometry experiment and by a simulation study.

Site‐specific deamidation of glutamine: a new marker of bone collagen deterioration

“…We can extract the degree of deamidation from ZooMS spectra with no modification to our existing protocols, thereby not adding to the cost of analysis. Similar approaches indeed exist for MALDI‐TOF‐MS . In the study by Leo et al it was argued that collagen contained too many hydroxylation variants to be of significant use, but our method allows us to use deamidation as an indication for single bone samples for a low cost and in a fraction of the time.…”

Automatic annotation and visualization tool for mass spectrometry based glycomics

“…The strategy in MS‐Deconv was to transform the overlapping issue to a heaviest path problem in the envelope graph . In addition, a Bayesian statistics model has been established to estimate the parameters related to the monoisotopic masses and abundance ratio of overlapping peptides for MALDI‐TOF MS . PIRMD can recognize monoisotopic peptides in chimera spectra from high‐resolution mass spectra .…”

“…All separated eGIAV i ( i ∈ N optimal ) for the corresponding compositions were generated and further identified by mGIA algorithm. This is different from the previously reported deconvolution algorithm, which detects monoisotopic peaks of peptides in the overlapped region . The advantage of the proposed algorithm is that the confidence in each monoisotopic peak can be obtained by using mGIA in combination with the separated eGIAV.…”

“…Some isotopic peaks may be shared by different isotopic clusters, which result in missing or wrong annotation. Thus many deconvolution algorithms have been proposed to overcome the problem of overlapping isotopic peaks . As examples, the THRASH method and LASSO‐based variable selection models have been developed for deconvoluting peptides with overlapping isotope patterns of multiple charge states that are often detected in electrospray ionization (ESI)‐MS.…”

Deconvolution in mass spectrometry based proteomics

“…For MALDI‐TOFMS, a Bayesian statistics model has been established to estimate the parameters related to the monoisotopic masses and abundance ratios of overlapping peptides . The mass‐dependent mean pattern was used to isotopically deconvolute MALDI spectra and post‐source decay (PSD) spectra from complex peptide samples.…”