Research On Plant NcRNA Interactions Based On Ensemble Learning

Non-codingRNA plays a vital role in many animal and plant life activities,and the representative ones are miRNA and lncRNA.More and more studies have shown that miRNA can not only interact with mRNA,but also interact with lncRNAto affect biological processes.At present,the research on miRNA and lncRNA is mainly focused on humans and animals.There are relatively few studies on miRNA and lncRNA in plants,and there are fewer studies on the interaction of miRNA and lncRNA in plants,and they are scattered in different plant species.The interaction of miRNA and lncRNA plays an irreplaceable role in the growth and development of plants.It affects plant vernalization,flowering,fruiting,cell differentiation and other important life processes.Exploring its mechanism and in-depth analysis of genomics functions are not only of great significance in increasing crop yields,enhancing crop resistance to cold,drought,and disease and insect pests,but also providing new ideas and support for further exploring the molecular structure and function of plant genes.With the continuous deepening of non-codingRNA research,various research methods have been proposed successively.These methods are mainly divided into two categories,biological experimental identification methods and computational methods.Biological genomics identification methods are accurate and reliable,but they are costly and have a long period.They have high requirements for experimental instruments and identification personnel,and it is difficult to apply them on a large scale.The calculation methods are mainly divided into two categories: feature engineering and deep-learning.Feature engineering uses manual acquisition of features,which can add some prior knowledge to help the model learn,but at the same time,it is subject to more human intervention,feature extraction is difficult,and the process is complex,and comprehensive acquisition Effective feature.Deep learning can automatically learn and obtain sequence features for classification prediction,but cannot obtain the structural features of gene sequences,which play a significant role in the function of noncodingRNA.This paper combines feature engineering and deep-learning methods to propose a new plant non-coding ncRNA interaction prediction model.The model acquires sequence features through deep-learning,while manually extracting structural features,and introduces structural features into the model through machine learning to help the model classify data more accurately.In deep-learning,the convolutional neural network is used to obtain and filter the sequence features,and then input them into the independent recurrent neural network to learn the long-term dependence between the features before and after the sequence,and effectively combat the problem of gradient disappearance and explosion.Machine-learning uses Random Forest as the base model for learning classification,and divides the model into two modes according to different ways of combining deep learning and machine learning.In order to verify the performance of the model,experiments were carried out on the data sets of Glycine max,Brachypodium distachyon,Medicago truncatula,and Setaria italica.The machine-learning methods based on feature engineering and different model architectures based on deep-learning were compared The real 21 miRNA-lncRNA interaction data of Solanum lycopersicum verified by biological experiments have been tested,the proposed model was able to accurately identify these data,which proves the validity and accuracy of the model.