Using a modelling technique specifically adapted to DNA helices, we have systematically studied the impact of base sequence on the geometry of the double helix. The results obtained show that each repetitive base sequence leads to several stable conformations belonging to the B-DNA family. These conformational sub-states generally have similar stabilities, but often differ considerably in terms of their helical and backbone parameters. Each sub-state can be characterised by the puckering of its sugar rings. Surface energy mapping and combinatorial search techniques are used to further understand the DNA conformational hypersurface and to extend our study from dinucleotide repeats to tetranucleotide sequences. The resulting structural database should be useful for predicting the properties of longer and more irregular base sequences and thus should contribute to understanding how DNA target sites are recognised.