Prediction of Post-translational Modifications of Proteins from 2-DE/MS Data

Abstract

The living cell is a complex entity consisting of nucleic acids, proteins, and otherbiomolecules that form an interrelated and dynamic network. The unraveling of this network is of great interest for scientists of different disciplines. With the sequencing of the genome a step was made to the understanding of the fundamental elements of the cells the genes. In humans, approximately 20,000 to 25,000 genes exist which encode about more than one million proteins. This complexity at the protein level is a result of alternative splicing and co- and post-translational modifications producing several protein species per transcript. Modifications are essential to the regulation of cellular processes and account for the activation or deactivation of enzymes and whole signaling pathways. The entirety of all proteins present in a cell at a fixed point of time and under particular biological conditions is called proteome, and the analysis of it is proteomics. One particular area of interest in proteomics is the identification of proteins and their post-translational modifications. Peptide mass fingerprinting is an established method and has proved useful to identify proteins by their amino acid sequence using mass spectrometry and protein sequence databases. This method relies on the idea of comparing experimental (measured) mass peaks to theoretical (calculated) masses, the latter being generated from a protein in a sequence database. As the mass of a modified protein differs from the mass of its unmodified counterpart, this mass distance is to be considered when detecting protein modifications with peptide mass fingerprinting. In the work described here, a novel algorithm was developed and implemented that allows for the identification of protein modifications from data derived by peptide mass fingerprinting. The algorithm transformed the process of predicting protein modifications to an extended Money Changing Problem of finding suitable combinations of modifications that explain the observed peak mass distances. Unlike common computational approaches the algorithm presented here will not be restricted in the number of modifications to be considered. Furthermore, this algorithm is efficient by calculating for a given list of modifications the combinations of modifications only once, independent of the number of queries. Although there exist hardly any frequencies of protein modifications, which turns the validation of the results very difficult, this novel approach is a promising step towards the unraveling of protein complexity.