Assessment Of RNA Tertiary Structures And Analysis Of The Distribution Of Three Way Junctions

RNA is a kind of macromolecule that plays a very important role in living things.If we want to understand a series of biological functions of RNA such as protein synthesis,information heredity and biocatalysis,we must know the exact structure of RNA.So People are becoming more conscious of 2D structure and 3D structure of RNA.At present,the study of secondary structure is quite mature,but the experimental study of RNA tertiary structure has been greatly restricted by the difficulty of crystallization of RNA molecules in vitro,the limitation of experimental technical conditions and so on.Therefore,It is more and more popular to predict the tertiary structure of RNA molecule theoretically,which design an effective algorithm to search Conformation space and a reasonable score function to distinguish the natural state and the candidate structure of RNA.There is currently many RNA tertiary structure prediction method,such as 3d RNA,RNA2D3 D,FARNA/FARFAR,MC-Fold/MC-Sym,i Fold RNA and NAST,etc.However,these methods are quite difficult for modeling complex or large RNA molecules,especially for the prediction of the tertiary structure of junction,because the spatial distribution of junction usually depends on base pairing and base stacking interaction.In addition,there is little research on scoring functions used to distinguish natural structure from candidate structures.All are knowledge-based energy functions related to atomic distances and dihedral angles.For example,3d RNAScore and RASP.In order to solve the above problems,In this paper,the idea of desiging the energy function by using the characteristic of the static electricity in RNA solution as a scoring index is proposed,and score function Elec_DH based on Coulomb Debye potential and Elec_Delphi based on Delphi is constructed.It is found that electrostatic energy is not suitable as scoring index alone for designing energy function,so a combined scoring function using atomic distance,dihedral angle and electrostatic energy is designed.For randstr test set,compared with 3d RNAScore scoring function,there are 13 RNA scoring accuracy improved.However,the corresponding 7 RNA scoring accuracy is reduced.In addition,the spatial structure of the three junction are statistically analyzed,Using several current mainstream machine learning algorithm including neural networks,support vector machine(SVM),K nearest neighbor,Bias,regression,decision tree and randomforest.We do the prediction of three angles between the helics and the prediction of the coaxial stacking effect and family topological structure of three junction.This will help the improvement of prediction accuracy of RNA tertiary structures.