Comments on 'MMFPh: A Maximal Motif Finder for Phosphoproteomics Datasets'

The discovery of phosphorylation motifs helps to understand the underlying regulation mechanism and to facilitate the prediction of unknown phosphorylation sites. Several methods have been proposed to detect phosphorylation motifs from phosphoproteomics datasets. A recent article (Wang et al. 2012) published in Bioinformatics developed a new method MMFPh (Maximal Motif Finder for Phosphoproteomics Datasets), which aims at identifying all statistically significant motifs and returns the maximal ones (those not subsumed by motifs with more fixed amino acids). The empirical comparison with Motif-X (Schwartz and Gygi, 2005) and Motif-All (He et al. 2011) showed that MMFPh is able to find more important motifs than Motif-X and return less false positives than Motif-All. Both Motif-All and MMFPh claimed that they can identify all statistically significant motifs. However, the lack of a precise problem definition and the ‘completeness’ definition may bring some confusions for users to select the appropriate method for their tasks. In this letter, we like to clarify the difference of these two methods and show that MMFPh cannot find all (maximal) statistically significant motifs even with respect to their own completeness definition.