The biological functions of DNA-information storage, replication, and transcription-are ultimately linked to the chemical elements of energetics, structure, and molecular motions. Understanding and interpreting the experimental results requires molecular models that are consistent with observed data. Molecular dynamics (MD) computer simulation is a quantitative, deterministic technique available for modeling macromolecular systems, and in principle produces a complete microscopic description of the molecular structure and motions of proteins, nucleic acids, and other macromolecules. In practice, MD modeling of DNA is a work in progress, with results mitigated by approximations in the underlying empirical force field, simulation protocols, system preparation, and digital computer capacity. Recent developments, particularly “second generation” force fields designed for use with explicit solvent models, treatment of long range interactions using the particle-mesh Ewald method, and advances in PC Cluster technology, have combined to produce considerably more accurate MD models of DNA than those obtained with first generation methods. These developments position MD modeling at a vantage point to make important and timely contributions to structural biology. The specific problems we are concerned with are:
- Characterization Studies on DNA and RNA oligonucleotides: MD simulations on oligonucleotide sequences, well-characterized experimentally, form a basis for assessments of MD methods as well as investigations into the nature of the dynamical structures. Studies of a broad range of systems involving diverse sequences in both solution and crystalline conditions are required to fully elaborate this problem area, and focusing on failures or limitations of the methodology is essential to determine where improvements need to be made. The development of general procedures for comparison of MD modeling results systematically with experimental results from crystallography and NMR spectroscopy is a part of this initiative.
- DNA and RNA Solvent Effects: MD studies in this problem area are divided into two sub-areas: a) the effect of water, counterions, and coions on nucleic acid structure and b) the effect of nucleic acids on the structure of water, counterions and coions. Patterns in the structure of water and ion atmosphere around DNA are delineated and compared with available data to assess the accuracy of the solvent models, residence times, and other dynamical elements of water and mobile ions. The development of methods for estimating solvation free energies from simulation is important to future developments in the field.
- Sequence Effects on DNA Structure and Axis Bending: MD models of the structure and motions of all 10 unique base-pair steps will serve as a basis for inquiring as to whether the average helicoidal parameters incorporated in a dinucleotide regression model can account for the variance in the experimental data such as gel retardation related to DNA bending. This project will be extended to consider context effects and each step will be analyzed with respect to solvation structure and the water and ion residence times to investigate further sequence dependent YR and RY steps and to explain the results in terms of a trade-off between steric clashes and hydrophobic penalties on unstacking. We participate in a consortium designed to produce MD models of all base pair quadruples in order to study context effects on structure.
- DNA and RNA Recognition and Ligand Binding: We are in the process of intensive MD studies of a small selection of DNA-drug, DNA-protein, and RNA-protein complexes. Specific issues we are interested in are the problems of structural reorganization and adaptation of nucleic acid and ligand in binding, the nature of counterion reorganization and release on complex formation, and the entropic effects on complex formation related to the conversion of rotational and translational degrees of freedom into low frequency vibrations. These studies all relate to the functional energetics of ligand binding, and emphasize aspects uniquely accessible to MD simulation. In parallel, we are concerned with the prospects of a general unified theory of molecular and macromolecular ligand binding to DNA and RNA and the difficult problems and limitations of computational theory in the sector.
- MD Simulations, Hidden Markov Models, and Structural Genomics: MD studies of DNA provide a potentially novel vantage point on the field of structural genomics. Experiments such as gel retardation and cyclization kinetics measure a combination of the properties of curvature, bending, and bendability, impossible to deconvolute since there is no unequivocal molecular model for the process. From MD these properties can be deconvoluted by step. We are pursuing the idea that probabilistic measures of DNA deformations of various sorts from MD can be used as a basis for Hidden Markov Models and used to identify structural homologies in genomic problems. If successful, the resulting HMMs will provide a tool for genomics research with the potential to score properly on structural as well sequence characteristics. HMM/MD methodology is expected to be particularly useful in elucidating multiple binding sites of regulatory proteins and in exploring the nature of direct and indirect readout in protein DNA and RNA recognition.