Prediction Of Ires Elements In RNA Based On Sequence And Structural Features

The Internal Ribosome Entry Site(IRES)which mainly located on non-coding region of the mRNA,can be found in both viruses and eukaryotes and mediate a translation process that differs from the typical cap-dependent protein translation.IRES elements have great research significance and clinical applications in exploring protein translation mechanisms,gene therapy and cancer pathogenesis exploration.At present,however,IRES verification methods such as construction of expression vectors are expensive and difficult to operate.Based on the sequence and structure information of IRES elements,we constructed a prediction model using bioinformatics methods to predict potential IRES elements which provides help for relevant research.The main contents of this paper are as follows:Firstly,verified primary and secondary sequence information of IRES elements from relevant databases and literatures were collected for further research.Then,some effective features such as specific primary and secondary motifs,special secondary structure motifs,minimum free energy of sequence,complementary base pairing length which can effectively distinguish the IRES elements from the general untranslated region sequences we screen out by compared IRES sequence with the primary structure,secondary structure,sequence general characteristics and the probability of protein interaction with non-coding sequence.Finally,eight features were screened by the sequential search algorithm which make the largest contribution to the prediction result.Secondly,based on the diversity-based AdaBoostSVM algorithm,two IRES prediction models for virus and eukaryotic IRES by training the virus and eukaryotic datasets of IRES sequences information were constructed respectively.By comparing the corresponding k-NearestNeighbor algorithm,support vector machine(SVM)classification,Gaussian process classification,Decision tree algorithm,naive Bayes algorithm,Artificial neural network algorithm(ANN)and other prediction models of IRES,the advantages and disadvantages of these prediction model were analysised to verify the validity of the diversity-based AdaBoostSVM model can be used to predict IRES.Subsequently,the accuracy,sensitivity,specificity and precision of the prediction model on eukaryotic cells and virus test sets reached 79.4%,73.4%,91.8%,94.9% and 88.4%,84.7%,93.6% and 95.0%,respectively.The prediction model were also compared with the existing IRESPred method to prove that the prediction model constructed in this study has better predictive ability.Finally,the issue of coding potential of circular RNA were studyed by using the constructed prediction model.The translation of circRNAs has been predicted by predicting the absence of IRES elements based on circular RNA sequence information.The positive final results proved the significance of our IRES related research.