The overall objective of the effort is to develop a comprehensive electrostatic molecular modeling software package to be implemented upon both serial and parallel computer platforms for application to DNA supercoiling. A central feature of the proposed method is an adaptive fast algorithm based on multipole and Taylor series expansions for computation of electrostatic energy and forces as well as the writhing number. In Phase I, a preliminary version of the method led to two orders of magnitude CPU reductions compared to direct summation for a 10,000 particle configuration. A parallel implementation was also developed and shown to scale with the number of processors and minimize interprocessor communication time. In Phase II, the fast analysis will be extended by developing parallel spatial partitioning techniques to maximize load balancing and optimize the fast algorithm performance, formally accounting for Debye-Huckel screening effects, and developing a suitable time stepping methodology for molecular dynamics applications. Enhancements to the elastic structural modeling of the DNA chain are also contemplated including direct computation of elastic constants from all-atom models. The analysis will be coupled to Monte Carlo/simulated annealing, deterministic optimization, and suitable time stepping codes to study DNA behavior.
The serial and parallel software ensuing from successful completion of the Phase II effort should be of general interest to the pharmaceutical and biotechnology industries. Efficient and reliable calculation of electrostatic interactions, which are frequently ignored or inappropriately approximated due to computational limitations, and improved structural representation of the DNA molecule, will make possible, among other applications, the routine study of highly charged ligand-receptor complexes, charged intercalative and groove binding drugs, and sequence-specific third oligonucleotide strands.
