Alignment of mass spectrometry data by clique finding and optimization

Mass spectrometry (MS) is becoming a popular approach for quantifying the protein composition of complex samples. A great challenge for comparative proteomic profiling is to match corresponding peptide features from different experiments to ensure that the same protein intensities are correctly identified. Multi-dimensional data acquisition from liquid-chromatography mass spectrometry (LC-MS) makes the alignment problem harder. We propose a general paradigm for aligning peptide features using a bounded error model. Our method is tolerant of imperfect measurements, missing peaks, and extraneous peaks. It can handle an arbitrary number of dimensions of separation, and is very fast in practice even for large data sets. Finally, its parameters are intuitive and we describe a heuristic for estimating them automatically. We demonstrate results on single- and multi-dimensional data.