The geometries of a 13 mer of a DNA double helix (5′-GCGTACACATGCG-3′) were determined by molecular dynamics simulations using a Cornell et al. empirical force field. The bases in the central base pair (shown in bold) were replaced (one or both) by a series of hydrophobic base analogues (phenyl, biphenyl, phenylnaphathalene, phenylanthracene and phenylphenanthrene). Due to the large fluctuations of the systems, an average geometry could not be determined. The interaction energies of the Model A, which consisted of three central steps of a duplex without a sugar phosphate backbone, taken from molecular dynamics simulations (geometry sampled every 1 ps), were calculated by the self-consistent charge density functional based tight-binding (SCC-DFTB-D) method and were subsequently averaged. The higher the stability of the systems the higher the aromaticity of the base analogues. To estimate the desolvation energy of the duplex, the COSMO continuum solvent model was used and the calculations were provided on a larger model, Model B (the three central steps of the duplex with a sugar phosphate backbone neutralised by H atoms), taken from molecular dynamics simulations (geometry sampled every 200 ps) and subsequently averaged. The selectivity of the base analogue pairs was ascertained (Model B) by including the desolvation energy and the interaction energy of both strands, as determined by the SCC-DFTB-D method. The highest selectivity was found for a phenylphenanthrene. Replacing the nucleic acid bases with a base analogue leads to structural changes of the central pair. Only with the smallest base analogues (phenyl) does the central base pair stay planar. When passing to larger base analogues the central base pair is usually stacked.