246
V. Poltev et al.
with Watson-Crick nucleoside pairs but with changed conformation of one or both
SPB chains.
The regularities mentioned above have been additionally confirmed by ab initio
computations of selected structures using MP2 level of theory and various basis sets
up to 6-311++G** [6, 7]. Nevertheless, both QM approaches failed to reproduce the
fine aspects of dDMP and cdDMP structures. DFT computations using PW91 and
PBE functional systematically overestimate the distance between bases of dDMPs
and between the stacked base pairs of cdDMPs [6, 7]. This error may be explained by
underestimation of dispersion attraction between the aromatic rings in the DFT level
of theory. This limitation of the DFT method however additionally strengthened our
conclusion about the critical contribution of SPB to the formation of dDMP structure reproducing the conformational characteristics of WCD and to the nucleoside
sequence dependence of 3D structure. This conclusion has been validated by exploring the local energy minima of SPB corresponding to BI, BII, AI, and AII families
of DNA. This analysis demonstrated that the conformations of SPB in these DNA
structures correspond to those of the local energy minima of the free SPB.
The MP2 computations of dDMPs and cdDMPs produce structures with shortened distances (up to 2.9 Å) between the stacked bases in comparison to the NDB
data for DNA fragments due to the known effect of overestimation of dispersion
interactions at this level of theory. This limitation of the MP2 method shows up
in various combinations of bases [21]. The use of counterpoise correction to the
basis set superposition error, BSSE, suggested by Boys and Bernardi [34] reduces
the absolute value of base-base stacking energies and improves the interatomic distances. With that, only a few moderately shortened contacts (up to 3.2 Å) remain in
disagreement with the experimental data [21]. Unfortunately, the application of this
method to cdDMPs remains computationally prohibitive.
The results obtained with the use of MP2 method demonstrate that the application
of higher level of theory does not automatically improve the agreement with experimental data for WCD. Likewise, physically unrealistic structures having shortened
atom-atom contacts arise when using DFT-D method [4] and the functionals specifically designed for computation of stacking complexes of organic molecules (such
as M05-2X and M06-2X) [7]. The overall conclusion from the application of QM
methods to minimal fragments of DNA is that they can reproduce and predict important conformational regularities of WCD, but neither of the methods reproduces the
complete set of important characteristics of these systems.
Another topic of interest is the capability of MM method to reproduce not only
the overall structure of WCD but also the regularities in sequence dependence of
3D structure revealed in DFT computations, especially on base superposition. Our
calculations of selected cdDMPs with both dDMPs belonging to the BI family by
using AMBER force field demonstrated that many but not all AMBER optimized
structures reproduce these regularities [7]. Some of the optimized structures incorrectly show considerable base ring superposition in Pyr-Pyr or Pyr-Pur sequences.
We suppose that it is because AMBER force field overestimates base stacking [7].
Systematic computations and the search for local energy minima for pairwise base
stacking interactions by using AMBER and CHARMM additive force fields support
V. Poltev et al.
with Watson-Crick nucleoside pairs but with changed conformation of one or both
SPB chains.
The regularities mentioned above have been additionally confirmed by ab initio
computations of selected structures using MP2 level of theory and various basis sets
up to 6-311++G** [6, 7]. Nevertheless, both QM approaches failed to reproduce the
fine aspects of dDMP and cdDMP structures. DFT computations using PW91 and
PBE functional systematically overestimate the distance between bases of dDMPs
and between the stacked base pairs of cdDMPs [6, 7]. This error may be explained by
underestimation of dispersion attraction between the aromatic rings in the DFT level
of theory. This limitation of the DFT method however additionally strengthened our
conclusion about the critical contribution of SPB to the formation of dDMP structure reproducing the conformational characteristics of WCD and to the nucleoside
sequence dependence of 3D structure. This conclusion has been validated by exploring the local energy minima of SPB corresponding to BI, BII, AI, and AII families
of DNA. This analysis demonstrated that the conformations of SPB in these DNA
structures correspond to those of the local energy minima of the free SPB.
The MP2 computations of dDMPs and cdDMPs produce structures with shortened distances (up to 2.9 Å) between the stacked bases in comparison to the NDB
data for DNA fragments due to the known effect of overestimation of dispersion
interactions at this level of theory. This limitation of the MP2 method shows up
in various combinations of bases [21]. The use of counterpoise correction to the
basis set superposition error, BSSE, suggested by Boys and Bernardi [34] reduces
the absolute value of base-base stacking energies and improves the interatomic distances. With that, only a few moderately shortened contacts (up to 3.2 Å) remain in
disagreement with the experimental data [21]. Unfortunately, the application of this
method to cdDMPs remains computationally prohibitive.
The results obtained with the use of MP2 method demonstrate that the application
of higher level of theory does not automatically improve the agreement with experimental data for WCD. Likewise, physically unrealistic structures having shortened
atom-atom contacts arise when using DFT-D method [4] and the functionals specifically designed for computation of stacking complexes of organic molecules (such
as M05-2X and M06-2X) [7]. The overall conclusion from the application of QM
methods to minimal fragments of DNA is that they can reproduce and predict important conformational regularities of WCD, but neither of the methods reproduces the
complete set of important characteristics of these systems.
Another topic of interest is the capability of MM method to reproduce not only
the overall structure of WCD but also the regularities in sequence dependence of
3D structure revealed in DFT computations, especially on base superposition. Our
calculations of selected cdDMPs with both dDMPs belonging to the BI family by
using AMBER force field demonstrated that many but not all AMBER optimized
structures reproduce these regularities [7]. Some of the optimized structures incorrectly show considerable base ring superposition in Pyr-Pyr or Pyr-Pur sequences.
We suppose that it is because AMBER force field overestimates base stacking [7].
Systematic computations and the search for local energy minima for pairwise base
stacking interactions by using AMBER and CHARMM additive force fields support