Efficient Biased Sampling Methods for Biomacromolecules: Protein Loops and RNA Thermodynamic Prediction

Biomacromolecules are fundamental structural and functional units of cell. Nucleic acids and proteins are the most common biomacromolecules. The relationship between sequences and structures of these biomacromolecules is one of the most important problem in biology for decades. Computational approaches can provide novel and efficient ways to study this problem and other structural-related problems, e.g. thermodynamics of nucleic acids. As nucleic acids and proteins are both biopolymers, sampling structures using chain-growth method with certain distribution is an effective approach to study the sequence-structure relationship of biomacromolecules. In this thesis, I develop a fast chain-growth method to efficiently predict protein loop conformations, and a coarse-grained chain-growth model to study thermodynamics of pseudoknotted RNA molecules. With an energy function designed specifically for loops, my method can efficiently generate high quality protein loop conformations with low energy that are enriched with near-native loop structures. I further applied this method to study multiple loop structures modeling problem as the interactions among loops in spatial proximity can be rather complex, and very few studies worked on this challenging problem. It shows better performance in accuracy compared to other methods. This method also succeeded in sampling and predicting conformations of antibody H3 loop while takes less computational time compared to other methods. For RNA pseudoknots, a coarse-grained chain-growth model is used to study the thermodynamics and folding stabilities of mouse mammary tumor virus pseudoknot – VPK. My results show that the melting temperature of VPK and its two subsequences can be correctly predicted. The melting temperature calculated from the heat capacity is in better agreement with the available experimental data than previous computational studies. My study also provides detailed information about the unfolding pathways of pseudoknots by analyzing the distribution of base pairing probability. The results favor the parallel melting pathway hypothesis of VPK folding over a simple sequential unfolding pathway. Overall, the above studies address two challenging problems of modeling three dimensional structures of proteins and RNAs, and have deepened our understanding of the relationship between sequences and structures of biomacromolecules.