Various methods have been recently employed to characterize the structure of biological networks. In particular, the concept of network motif and the related one of coloured motif have proven useful to model the notion of a functional/evolutionary building block. Here we are interested in comparing two metabolic networks, viewed as graphs whose nodes represent reactions and are coloured according to the class of enzymes that catalyze reactions, based on their composition in small coloured motifs. We build an ad-hoc distance between two graphs which takes into account the colour frequency in the graphs and the probabilities of connection. For this, we use the exact expectation and covariance of counts of coloured motifs incoloured Erdos-Renyi random graphs (Schbath et al., 2009). Used iteratively, our method allows providing the most discriminating coloured motifs. (joint work with Stephane Robin and Elie-LaurentBenaroya)Reference 1Schbath, S., Lacroix, V. and Sagot, M.-F. (2009). Assessing the exceptionality of coloured motifs in networks. EURASIP Journal on Bioinformatics and Systems Biology.