APPLICATION MACHINE LEARNING TO CHARACTERIZE MIRNAS IN WHEAT PANGENOME Abstract uri icon

abstract

  • Core and dispensable genomes are carefully examined in order to reveal/improve agronomically traits that might be giving plants an advantage for tolerance to a/biotic stresses and the yield increase. In this study, we performed in silico miRNA identification from the genomes of six hexaploid wheat cultivars to find miRNA families that were present in the core and dispensable genomes. The results revealed 93 miRNA families that were commonly found in 6 lineages. On the other hand, 3 miRNA families were lineage-specific. Target analysis for the identified miRNAs was performed in order to find out if they regulate any genes related to stress resistance. One of the lineage-specific miRNAs, miR384, was found in the genome of Jagger that is known to have stem rust resistance, and targeted 5 transcripts that coded for NBS-LRR disease resistance proteins. The remaining two miRNAs, miR8175 and miR9675, were identified from Stanley genome and notably, they were also targeting several transcripts coding disease resistance proteins as well as some transcripts coding abiotic stressrelated proteins such as MYB and AP2. In addition, we further applied Machine Learning (ML) based approaches to classify miRNAs based on their genomic origins. Using more than 170 features, we achieved >90% accuracy in classification of six genome miRNAs using three different algorithms. Prediction accuracy was >90% in only naive Bayes. Finding regulatory elements like miRNAs behind these advantageous traits that might be lying on dispensable genomes will help us gain more insight about a/biotic stress resistance and regulation in plants and will fasten the efforts to develop stress-tolerant cultivars.

publication date

  • July 2019