A. V. Starostin1, D. D. Gavrilova1 , Y. L. Orlov1
1 I.M. Sechenov First Moscow State Medical University, Moscow, Russia
staral.ru [at] yandex.ru
Abstract
To assess the effectiveness of neural network based search for target genes for miRNA regulation of atherosclerosis development. The list of mitophagy associated genes was selected by analyzing scientific databases. Then, each of the genes was passed through a specialized miRDB database in order to find corresponding miRNA. The resulting miRNA list was analyzed through the web interface, which transmitted data to the miTAR neural network, in order to estimate affinity values. The final stage of the experiment included the identification of the obtained through miTAR sequences through the NCBI Blast web service. Next, the list of obtained genes was compared with the list of genes found manually using miRDB. Based on the results of the comparison we came to the conclusion that the neural network based miTAR surpassingly copes with the task of predicting microRNA target genes, albeit the results for miRNA with lower affinity index significantly varied from the ones found manually. However, this reservation could be attributed to upsides of the model as it reveals new potential target genes for further evaluation. Our results indicate that neural network models can serve as an effective tool in the search for miRNA target genes. In addition, the found genes could be furtherly studied to identify their linkage to atherosclerosis. This will also assist in the study of possible drug substances in the treatment of atherosclerosis.