Abstract
A new algorithm is developed which implements the Monte Carlo annealing method and is used to model and minimize linear, single and double stranded B-DNA sequences. The preliminary B-DNA structures are modeled using initial structures obtained from the Brookhaven database. The model contains structural details at the atomic level and is therefore more elaborate and accurate than the pseudo-atomic and elastomechanical models. The minimization concerns the entire chain length and not only local nucleotide complexes. A variety of DNA sequences (coding or non-coding, random or real, homogeneous or heterogeneous) are investigated in the range of 20-40 base pairs. The potential energy function is written in terms of a set of internal coordinates defined to account for the helical parameters such as twist, tilt and rise, which are important parameters for the description of the global shape of any type of DNA or RNA molecule. The force field used is composed of a limited number of bonded and non-bonded interactions such as bond stretch, angle bend and Lennard-Jones interactions with the Dreiding II force field parameter set used for these interactions. From the minimized structures the angles between Phosphate-Oxygen-Carbon " A 1 " and Oxygen-Phosphate-Oxygen " A 2 " and the average helical twist were calculated. For single strands it is shown that the bond angles are A 1 =107 - 1° and A 2 =122 - 1°, while the helical twist is 37.8 - 1°. For double stranded DNA our model predicts the helical twist of h =35.5 - 2° well, in the A strand, while the prediction is less accurate, h =47 - 2° for the complementary strand B. The average values for the angles A 1 and A 2 are 130 - 1° and 150 - 3° for strand A and 102 - 4° and 123 - 5° for strand B. The reason for this discrepancy is attributed to the different conformations initially adopted by the sugars in the A and B strands.