A combinatorial approach to the peptide feature matching problem for label-free quantification

Motivation: Label-free quantification is an important approach to identify biomarkers, as it measures the quantity change of peptides across different biological samples. One of the fundamental steps for label-free quantification is to match the peptide features that are detected in two datasets to each other. Although ad hoc software tools exist for the feature matching, the definition of a combinatorial model for this problem is still not available. Results: A combinatorial model is proposed in this article. Each peptide feature contains a mass value and a retention time value, which are used to calculate a matching weight between a pair of features. The feature matching is to find the maximum-weighted matching between the two sets of features, after applying a to-be-computed time alignment function to all the retention time values of one set of the features. This is similar to the maximum matching problem in a bipartite graph. But we show that the requirement of time alignment makes the problem NP-hard. Practical algorithms are also provided. Experiments on real data show that the algorithm compares favorably with other existing methods.