COMBINED WORMLIKE-CHAIN AND BEAD MODEL FOR DYNAMIC SIMULATIONS OF LONG LINEAR DNA

A carefully parameterized and tested simulation procedure for studying the dynamic properties of long linear DNA, based on a representation that combines features of both wormlike-chain and bead models, is pres ented. Our goals are to verify the model parameters and protocols with respect to all relevant experimental data and equilibrium simulations , to choose the most efficient algorithms, and to test different appro ximations that increase the speed of the computations. The energy of t he linear model chain includes stretching, bending, and electrostatic components. Beads are associated with each vertex of the chain in orde r to specify the hydrodynamic properties of the DNA. The value of the stretching rigidity constant is chosen to achieve a com prom ise betwe en the efficiency of the dynamic simulations (since the timestep depen ds on the stretching constant) and realistic modelling of the DNA (i.e ., small deviations of the input contour length); the bead hydrodynami c radius is set to yield agreement with known values of the translatio nal diffusion coefficient. By comparing results from both a first- and a second-order Brownian dynamics algorithm, we find that the two sche mes give reasonable accuracy for integration timesteps in the range 20 0-500 ps. However, the greater accuracy of the second-order algorithm permits timesteps of 600 ps to be used for better accuracy than the 30 0 ps used in the first-order method. We develop a more efficient secon d-order algorithm for our model by eliminating the auxiliary calculati ons of the translational diffusion tensor at each timestep. This treat ment does not sacrifice accuracy and reduces the required CPU time by about 50%. We also show that an appropriate monitoring of the chain to pology ensures essentially no intrachain crossing. The model details a re assessed by comparing simulation-generated equilibrium and dynamic properties with results of Monte Carlo simulations for short linear DN A (300, 600 base pairs) and with experimental results. Very good agree ment is obtained with Monte Carte results for distributions of the end -to-end distance, bond lengths, bond angles between adjacent links, an d translational diffusion measurements. Additionally, comparison of tr anslational diffusion coefficients with experimentally-measured values for DNA chains (of 367, 762, 1010, 2311 base pairs) shows excellent a greement as well. This lends confidence to the predictive ability of o ur model and sets the groundwork for further work on circular DNA. We conclude with results of such a predictive measurement, the autocorrel ation time, for the end-to-end distance and the bending angle as a fun ction of DNA length. Rotational diffusion measurements for different D NA lengths (300 to 2311 base pairs) are also presented. (C) 1997 Acade mic Press.