RNA Secondary Structure Prediction Based On The Minimum Energy Model

RNA molecules are important cellular components involved in manyfundamental biological processes. Understanding the mechanisms behind theirfunctions requires RNA tertiary structure knowledge. The current, direct theoreticalprediction of the tertiary structure does not progress smoothly, thus through predictingtheir secondary structure to achieve the purpose for their tertiary structure become aninevitable trend.Firstly, this paper has illustrated and discussed the relevant theories in thefollowing aspects of RNA: the constitution, classification, structure, functions and soon. The thermodynamic model has been introduced to predict the model. It alsoprovided an in-depth analysis on the prediction algorithms, includes two categories:one is based on the sequence alignment, the other is based on the free energyminimization. Through comparative analysis, based on the sequence alignmentalgorithm is limited by the prior knowledge of the existing sequence, and thecomplexity of the time and space is high; based on the free energy minimization’saccuracy is not high enough or time complexity, such as RNA secondary structureprediction based on the discrete Hopfield Neural Network (DHNN) which has quicklyconvergence speed but is easy to fall into local optimum, that affect the predictionaccuracy of the algorithm.In response to these problems, a hybrid algorithm was proposed based on thefusion of Discrete Hopfield Neural Network (DHNN) and Tabu Search (TS). DHNNis the main method which could quickly obtain a feasible solution of partitioning, andthe TS algorithm could “taboo” the current solution and transferred to the otherminimum points that could jump out from the local optimal solution. In the process ofthe algorithm implementation, as the DHNN algorithm depends on the initial value,taken for an improved method. Consider the different neurons different initial valuesaccording to the distance function in the initialization of neurons, in order to reducethe time and space complexity of the algorithm, we used a pretreatment method tocalculate the overall energy. It has been demonstrated through the results of experiments: testing TS_DHNNalgorithm’s prediction accuracy from the stalk level and base level and comparing theprediction accuracy with a separate DHNN algorithm, TS algorithm andRNAStructure comparison, the following statements can be concluded:①TS_DHNNalgorithm is stable and effective, the number of zones and the actual number of thestalk region of the correct prediction stems basically consistent accuracy of80%canbe achieved.②the algorithm either from the point of view predict a tRNA or RNasepRNA sequence alone, or from the overall prediction situation are significantly betterthan the other separate algorithm to prove the prediction accuracy of the hybridalgorithm for RNA play an active effect.