Novel molecular computational methods and their quantitative assessment

Molecular computational methods and their efficiency estimation are discussed in this thesis.Computational simulations in biological systems help to understand biological process, drug design and many other crucial applications, thus plays more and more important roles in biological system studies. First, the new algorithm based on polymer growth strategies isintroduced. The main novel feature of this approach is the use of pre-calculated statisticallibraries of molecular fragments. A molecule is sampled by combining fragment configurationsof single residues in this study which are stored in the libraries. This method isdemonstrated to be accurate and can generate large peptides (i.e., 16 residues) in less than aminute of single-processor computing. As an application to this growth algorithm, a practicalmethod is developed to calculate absolute free energy that stages such calculation in severalsteps through growing a molecule. Significant computer time is saved by pre-calculatingfragment configurations and interactions for reuse in a variety of molecules. Then the question¡°how faster is a method than standard molecular dynamics?¡± is addressed. To quantifythe progress in the development of algorithms and forcefields used in molecular simulations,a general method for the assessment of the sampling quality is needed. I therefore developan approach for analyzing the variances in state populations, which quantifies the degree ofsampling in terms of the effective sample size (ESS). This procedure is tested in a variety ofsystems from toy models to atomistic protein simulations. At last, a simple automated procedurewas introduced to obtain approximate physical states from dynamic trajectories: thisallows samplesize estimation in systems for which physical states are not known in advance