Computational Prediction Of Non-coding RNA Structures

Noncoding RNA (ncRNA) molecules are involved in various biological processes, such as catalytic and regulatory roles. For example, the small interference RNA has a remarkable role in gene silencing and cancer therapies; the signal recognition particle RNA (SRP RNA) can recognize and bind specific proteins. The tertiary structures of RNA molecules are the basis to perform the function. However, at present it is still difficult to obtain RNA tertiary structures experimentally, therefore, computational prediction and design of RNA tertiary structures are of great interest. The main contents of our work are described as follows.The automatic building of three-dimensional RNA structures. Building tertiary structures of non-coding RNA is required to understand their function and design new molecules. Current algorithms of RNA tertiary structure prediction give satisfactory accuracy only for small size and simple topology and many of them need manual manipulation. Here, we present an automatic and fast program,3dRNA, for RNA tertiary structure prediction with reasonable accuracy for RNAs of larger size and complex topology. In a benchmark test of RNA molecules with known tertiary structures, the predicted structures have an average heavy-atom RMSD value of2.8A for hairpins and duplexes,5.8A for larger RNAs with junctions from the experimental ones. The method gives high positive predictive values (PPV) and interaction network fidelity (INF). The performance of3dRNA is also compared with other popular prediction methods.The related problems of the computational prediction of non-coding RNA structures:(1) The comparison of the secondary structures between the native and predicted ones. The secondary structures of RNA are crucial to their functions and are also the basis of tertiary structure prediction. Recent investigations on riboswitches have shown that the secondary structures of RNA molecules in the functional states may not be the minimum free-energy states. To see the generality of this phenomenon, we compare the experimental secondary structures of RNA molecules in our RNA database with the predicted ones. The result shows that more than half of RNA molecules don't employ the minimum free-energy secondary structures. Our results indicate that the functional states of many RNA molecules may be locally stable states or kinetically controlled.(2) The characteristics of RNA tertiary structure modules. Conserved subunits in RNA tertiary structures are important to their structural organizations and biological functions. A preserved parallel helix-loop-helix conformation is frequently found in different RNA families. We investigate the mechanisms of stability of such parallel conformation by using secondary structure kinetics, all-atom molecular dynamics, and docking method. Our results indicate that this substructure itself alone is very stable. However, this parallel conformation is unstable in the absence of positive ions and without other tertiary interactions.(3) The improvments of RNA2D3D structure prediction method. Computational prediction of the RNA tertiary structures is a significant challenge, especially for the longer RNA and pseudoknots. One of possible approaches is through hierarchical steps: from sequence to secondary structure, then to tertiary structure. Here we present improvements of two key steps of this approach, the manual adjustment of atom overlapping and bond stretching using an energy function, and molecular dynamics refinement with a tested combination of Amber force field with implicit solvent model respectively.