Non-additivity of base-base interactions in all ten possible model dinucleotide steps were analyzed on MP2/aug-cc-pvDZ quantum chemistry level. Selected conformations of four nucleobases exactly matched to ones occurring in B-DNA crystals. In most of 162 analyzed tetramers both threeand four-body contributions are negligible except of d(GpG) steps. However, in these dinucleotides both contributions are always of opposite signs and in all cases the sum of all non-additive part of intermolecular interactions do not exceed 2.6 kcal/mol. This stands for less than 5% of the overall binding energy of dinucleotide steps. Also replacements of guanine with 8-oxoguanine in d(GpG) systems introduces non-additivity of the same magnitude as for canonical dinucleotides. It is observed linear relationships between values of the total binding energy obtained in the tetramer basis set and estimated energy exclusively in dimers basis sets with assumption of pairwise additivities. For all analyzed dinucleotides steps there are also linear correlations between amount of non-additive contributions and magnitude of pairs interactions. Based on differences in electrostatic contribution to the total binding energy of four nucleobases and polarity of dinucleotide steps three distinct classes of dinucleotide steps were identified.Response to Reviewers: 1) I recommend to do a careful proofreading, there are many typos and some sentences are difficult to understand. Thank you. Careful proofreading was applied and text was modified accordingly.2) Page 3, Results and discussion Alltogether 162 dinucleotide steps were analyzed. The number of individual steps is variable, why is it so? Why there was 48 d(GpG) steps, but only 10 d(GpC) steps? What were the PDB codes of structures from which steps were chosen, and which steps (chain ID, res ID) were chosen? What was the resolution of these structures? This information could go to the Supplementary material.In supplementary material details related to analyzed structures are provided now. The differences in dinucleotide populations have strictly technical reason. Since in my previous work accepted for publication in Int.J.Quantum Chem. (DOI:10.1002/qua.22435) detailed analysis of d(GpG) steps were presented I have just included these data in this work. Consequently this dinucleotide is overrepresented. Since it constitutes its own class I do not think that such overrepresentation changes papers conclusions. On the other hand it is not necessary extending the number of other dinucleotide steps since the non-additivity is quite small and they were selected from broad range of data according to IIE and structural parameters diversities.3) Page 3, Results and discussion The numbers of steps are large enough to perform some statistical analysis of resulting energies. E.g. ANOVA could be performed to test whether the mean energies in dinucleotide steps differ significantly each from other. Also, is the distribution of energies for individual steps Gaussian or not? If not, some non-parametric ...