Double-helical polynucleotide conformations, poly(dA).poly(dT), poly(d(A-T)).poly(d(T-A)),poly(dG).poly(dC), and poly(d(G-C)).poly(d(C-G)) are analyzed by the atom-atom potential method. The energy optimization is carried out in the space of eight independent geometric parameters using analytical procedures for the constraints, taking into account the flexibility of the P-D-deoxyribose rings. At the first stage, the full screening of atomic partial charges was assumed. The structures of the calculated B and the A forms of DNA are characterized by low energy and absence of short contacts; the dihedral angles are near the average values in the monomers. With the typical energy difference of 3-5 kcal/mol nucleotide pairs in all cases, the B form is more preferable as compared to the A form. At the final step the effect of the Coulomb term is evaluated for poly(dA).poly(dT) using various values of the effective dielectric constant ( E = 28, 24, 20, 18, 14, 12, 10, 8, 6 , 4, and 1). If F 5 2 4 , the energy optimization leads A to B. We discuss the stereochemical details of the intermediate conformations on the A-B path and hypothesize the nature of stability of the A and the B forms and the mechanism of the A-B transition.