Function Prediction Of LncRNAs Based On Deep Learning Architecture

RNA-protein interactions play important roles in many important biological processes.Based on the innovation and development of next-generation sequencing technology,hundreds of RNA-binding proteins(RBPs)and their corresponding RNAs have been gradually discovered.By summarizing and analyzing its biological process,it is possible to make large-scale prediction of RNA-protein interactions using machine learning methods in computational biology.So far,scholars in the field of computational biology have explored and developed a variety of computational tools and methods on this issue,including deep learning models,which have also made remarkable achievements in the recognition of RNA-protein binding affinity and sites based on deep learning models.RNA-binding proteins(RBPs)are closely involved in many important biological processes.The resulting complex of RNA binding to proteins plays an important role in many biological processes,such as gene expression regulation,splicing,translation rules,and viral replication.Therefore,understanding protein-RNA binding may provide important insights into the function and dynamics of many cellular processes.This has stimulated interest in experimental studies and computational predictions of protein-RNA binding.Long non-coding RNAs(lncRNAs)are non-coding RNAs with a length of more than 200 nucleotides.Although the research on lncRNAs has made rapid progress,it is still unclear what functions most lncRNAs should have.With the advance of biological research,a large number of lncRNAs have been discovered successively.Therefore,the study of a wide variety of lncRNAs has gradually become a very noteworthy direction in the study of RNA genomes.Lnc RNA plays an important role in a large number of biological processes.They also contribute to the development and development of some serious genetic diseases,such as cancer.However,many lncRNAs lack functional annotations to explain such problems.Therefore,understanding the function of lncRNAs is a necessary step for biological and medical research.Because the calculation method is more suitable for large-scale experiments than the biological experiment method in terms of time and cost,the prediction and analysis by the calculation method can provide an important reference and basis for the experimental method.However,the method to predict the function of lncRNAs by means of computation and to detect new lncRNAs is still very challenging from the current research level.Some lncRNAs work on the basis of binding to proteins,so calculating the interaction preference between lncRNAs and proteins may provide a new breakthrough in the development of new computational methods.In this paper,we firstly improve and upgrade the convolutional neural network model in DLPRB,a highperformance deep neural network with the best performance at present,which is RNA and protein binding preference prediction.This model was used to predict the binding preference of all 244 experimental datasets of RNA-protein binding preference in the RNAcompete dataset.We summarized and classified the improved model related to protein functions,and put the sequence and characteristics of lncRNAs into the model through transfer learning method as experimental input data,and predicted the function of lncRNAs through binding strength value.Through the construction of the experimental model in this paper,we have improved the effect of the existing experimental model with the best performance.In the interval model with Pearson correlation coefficient greater than 0.6,the improvement effect of this experiment accounted for 90.8%.At the same time,we proposed a voting algorithm to predict the function of lncRNAs based on the designed model.This is the first computational method to use RNA-protein binding strength and deep learning techniques to predict lncRNA function.Finally,we used 25 lncRNAs from the Lnc Book data set to test our prediction method,and the recall rate of the experimental results reached 88%,which was significantly better than the previous experimental methods.The results confirmed that our method was meaningful.This method can be used to predict the function of newly detected lncRNAs only by using sequence-based features.We believe that this work can provide a new perspective for the study of lncRNAs.