Estimation of consistent identification of the E coli transcriptional classification.

We assess the correctness of our identified gene-pairs with the E coli activation (ac) and repression (re) relationships used by SynTReN to build the networks. This is equivalent to a check on specificity. We additionally wished to identify gene-pairs which were highly likely to occur, based on the ac definitions, but including transitive relations, that is - all the genes that are connected by an ac network path-length of 2. This 2-path network is not a full prediction of all observed relations in the data-file as it does not include the du and re pairs. We calculated the sum of the E coli definition adjacency matrices for ac (+1) and re (−1) for path-length 1, 2 and 3 and again compared this network with our identified pairs. The results with correlation analysis are almost the same as those found by discretization.