Annotating Fragmentation Patterns

Abstract

Mass spectrometry is one of the key technologies in metabolomics for the identification and quantification of molecules in small concentrations. For identification, these molecules are fragmented by, e.g., tandem mass spectrometry, and masses and abundances of the resulting fragments are measured. Recently, methods for de novo interpretation of tandem mass spectra and the automated inference of fragmentation patterns have been developed. If the correct structural formula is known, then peaks in the fragmentation pattern can be annotated by substructures of the underlying compound. To determine the structure of these fragments manually is tedious and time-consuming. Hence, there is a need for automated identification of the generated fragments.

In this work, we consider the problem of annotating fragmentation patterns. Our input are fragmentation trees, representing tandem mass spectra where each peak has been assigned a molecular formula, and fragmentation dependencies are known. Given a fixed structural formula and any fragment molecular formula, we search for all structural fragments that satisfy elemental multiplicities. Ultimately, we search for a fragmentation pattern annotation with minimum total cleavage costs. We discuss several algorithmic approaches for this problem, including a randomized and a tree decomposition-based algorithm. We find that even though the problem of identifying structural fragments is NP-hard, instances based on molecular structures can be efficiently solved with a classical branch-and-bound algorithm.