Abstract: In order to research the recognition method for E. coli promoter, E. coli promoter feature elements are researched combining with molecubiology theory and statistical facts of E. coli gene promoters. Two-structure neural network methods were applied to analysis the promoter sequence elements in E. coli gene promoters by selecting different promoter conservative sequences. As a result, we found that the recognition rate of the positives and negatives have the best performance when the canonical elements and the non-canonical sequence elements are all included, which is 77.67% and 88.45% respectively. This result can provide help to the feature selection and the recognition algorithm research of the promoters.