Identification of RNA pseudouridine sites using deep learning approaches

Pseudouridine(psi) is widely popular among various RNA modifications which have been confirmed to occur in rRNA, mRNA, tRNA, and nuclear/nucleolar RNA. Hence, identifying them has vital significance in academic research, drug development and gene therapies. Several laboratory techniques for psi identification have been introduced over the years. Although these techniques produce satisfactory results, they are costly, time-consuming and requires skilled experience. As the lengths of RNA sequences are getting longer day by day, an efficient method for identifying pseudouridine sites using computational approaches is very important. In this paper, we proposed a multi-channel convolution neural network using binary encoding. We employed k-fold cross-validation and grid search to tune the hyperparameters. We evaluated its performance in the independent datasets and found promising results. The results proved that our method can be used to identify pseudouridine sites for associated purposes. We have also implemented an easily accessible web server at http://103.99.176.239/ipseumulticnn/.

Automated summary

In this paper, a multi-channel convolution neural network using binary encoding was proposed for identifying pseudouridine sites. Tuning hyperparameters with k-fold cross-validation and grid search, promising results were obtained in independent datasets. This method proved to be effective for identifying pseudouridine sites and has been implemented as an easily accessible web server.