Prediction Research Of Plant LncRNA-Protein Interaction Based On Sequence Information

Long non-coding RNA(lncRNA)is widely involved in plant growth,development and stress,which has become an important topic in gene regulation,development and environmental response around the world.The interaction of lncRNA-protein plays an important regulatory role in plant immunity and life activities.Researchers are increasingly using computer technology to help and support related researches,because laboratory methods are time-consuming and labor-intensive.However,research in plants is immature compared to research in animals in this area.Due to the lack of experimental verification of interaction data,there are large differences between known and unknown interaction samples in plant datasets and a large number of noisy data,which can affect the performance of prediction models.In order to better solve the above problems,this study attempted to establish two effective prediction models for the interaction plant lncRNA-protein interaction based on sequence information,so as to improve the accuracy of prediction and promote the related research on plant molecular biology.Firstly,in this study,different feature extraction methods were used to the sequence information of lncRNA and protein downloaded from the plant database and establish the relevant feature matrix.A multi-feature fusion method based on linear neighborhood propagation is proposed to predict plant lncRNA-protein interactions.The linear neighborhood similarity of feature space is calculated and the result is predicted by label propagation method.The potential interactive information in data can be better explored by integrating multiple feature training models.Then,in order to reduce the impact of noise data and better predict unlabeled data,this paper further proposed a non-negative matrix decomposition algorithm based on axiomatic fuzzy set theory to predict plant lncRNA-protein interaction.The fuzzy similarity matrix is constructed based on axiomatic fuzzy set theory,and the obtained fuzzy similarity matrix is treated with graph regularization to constrain the decomposition process of matrix,which effectively reduces the influence of noise data and has good prediction performance.In this paper,the performance of the proposed method was evaluated by using the five-fold cross validation method.Experimental results show that the proposed method for predicting plant lncRNA-protein interactions is superior to other related prediction methods.The proposed prediction method has certain generalization ability for plant data sets,and can be used as an auxiliary tool to explore the plant lncRNA-protein interaction relationship,and help to further explain the function and mechanism of plant lncRNAs,and accelerate the study of plant molecular biology related fields.