Prediction of regulatory proteins in the E.coli genome. The method is divided into two stages. In the first stage, a set of putative transcription factors is generated. This results from searches in the SwissProt database (keywords and regular expressions were used) and from scanning all 4283 ORFs of the E.coli genome using Dodd and Egan, PROSITE patterns and Gibbs sampler algorithms. In addition, a literature search was performed for evidence concerning new transcription factors. We filtered proteins that are not transcription regulators (a transcription factor has a DNA-binding functional motif). In the second part, a sequence comparison was performed using Blast 2.0 to detect additional proteins. Additionally, annotations of the E.coli genome were used together with the collection in RegulonDB (3,32). The DNA-binding protein structures used in the Gibbs sampler calibration were: ArgR (58), BirA (59,60), Crp (61), Fis (62), FruR (63), LacI (64), LexA (65), NarL (66), OmpR (67), PurR (68), TetR (69,70) and TrpR (71).