Research On Plant MicroRNA Prediction Algorithm And Its Target Prediction Algorithm

MicroRNA(miRNA)is a kind of endogenous single-stranded non-coding smallRNA with a length of about 20-24 nucleotides,which has high evolutionary conservation and expression specificity.In addition,miRNA plays an extensive and important regulatory role in various physiological and pathological processes in organisms.So far,scientists have detected tens of thousands of miRNAs in organisms such as animals,plants,viruses,and fungi,but there are still a large number of unknown miRNAs waiting to be discovered.The discovery and identification of more new miRNAs will facilitate deeper comprehensive research and analysis on their functions and their regulatory roles in complex biological processes.The discovery of new miRNAs mainly includes biological experimental detection and computational prediction.Although the former is more direct and accurate,it has a long experimental cycle and high cost,and it is difficult to clone miRNAs expressed in specific tissues and periods.Computational prediction methods can make up for the lack of experimental methods.With the continuous integration and development of bioinformatics and machine learning,computational prediction methods based on machine learning have become a current research hotspot.Plant miRNA precursors(premiRNA)have more complex secondary structures than animals,and there are relatively few studies on plant miRNA prediction,so it is necessary to design a plant-specific miRNA prediction algorithm.In the first part of the research,the algorithm plant Mir P2 was designed for plant premiRNA and miRNA prediction.Using the latest mi RBase database data,the knowledgebased energy features are extensively combined with other excellent features,and the plant pre-miRNA prediction model is established based on the support vector machine algorithm.The prediction accuracy of plant Mir P2 on the plant pre-miRNA dataset reached 97.58%,showing good prediction performance.Through computational analysis of plant next-generation sequencing data,plant Mir P2 can also predict new miRNAs.In addition,the usability of the algorithm is enhanced by building a web server and other methods.The next step of miRNA prediction is miRNA target prediction,which is also an important step in the entire miRNA analysis research.Computational-based target prediction and identification methods will be used first to narrow the scope of experimental verification,thereby reducing the energy and research costs of target verification.Therefore,a prediction algorithm with higher accuracy and lower false positive rate is very important.In addition,the existing methods lack the utilization of increasingly abundant experimentally verified miRNA target pair data,and considering the significant differences between plants and animals in the miRNA target recognition process,it is necessary to design target prediction algorithms for plants alone.In the second part of the research,for the model plant Arabidopsis thaliana,a miRNA target prediction algorithm ath Mir TP was designed based on traditional target calculation prediction rules and deep residual network.Through the mining and utilization of experimentally validated Arabidopsis miRNA target data,higher prediction performance than traditional rule-based target prediction algorithms was achieved.In addition,ath Mir TP also showed excellent predictive ability and generalization in cross-species plant miRNA target prediction.