The Research On Prediction Of RNA Secondary Structure

The function of a biomolecule is closely related to its structure,even in very different species,molecules with the same function exhibit very similar structure.For RNA molecules,on one hand,because their secondary structures construct main framework of their tertiary or quaternary structures and cover most of their free energy of folding,they can be used to partially even fully interpret RNA function.On the other hand,because of fast decomposition and hard crystallization of RNA,it will result in high payments on both money and time to use experimental methods to determine RNA 3D-structures.Therefore,the use of computational methods and mathematic theory to predict RNA structures is highly desired.This thesis mainly includes three parts of following contents and conclusions.In the first part,some current important methods of RNA secondary structure prediction are introduced,and the scope of their application as well as their merits and demerits are analyzed.Because of conservation of homology RNA secondary structures, it is an effective means to predict their structures via comparative sequence analysis when there are sufficient homology RNA sequences.On the contrary,the method of free energy minimization is a conventional choice.It's frequently to combine them together to take both advantages.In the second part,we study the feasibility to predict an RNA secondary structure by using base pair probability taking from the energy contribution of all possible secondary structures.At first,the concepts and computation processes of partition function and base pair probability for RNA molecules are introduced in detail.Second, it is discussed how to use a base pair probability matrix to predict a RNA secondary structure with maximum expectation accuracy according to its definition.Third,the new method is tested by some RNA sequences and the test results are compared with that of RNAfold implemented on the basis of free energy minimization algorithm.The initial comparison indicates that the accuracy of the new method is slightly better than RNAfold at most cases.In the third part,we study how to combine minimum of free energy and covariation information to predict the consensus secondary structure of homology RNAs and the application of iteration for predicting pseudoknots.In the first place,we describe the formation for calculation of meanly minimum free energy and covariation value and how to combine them to predict the consensus secondary structure of homology RNAs.In the second place,by analyzing the ability of iterating method on pseudoknots prediction,we present a new iterating method to predict the consensus structure of homology RNAs including pseudoknots.Finally,the testing and comparing results demonstrate that it is superior to other methods,because its sensitivity and specificity are closed to or are the best after the prediction results are simply processed.