Biological Features Between miRNAs and their Targets are Unveiled from Deep Learning Models

miRNAs are a kind of small Ribonucleic acid (RNAs) that have an average length of ~22 nucleotides. They typically form base-pairs with their targets to repress gene expression post-transcriptionally. miRNAs play key roles in a variety of biological processes and human diseases and several miRNA-targeted therapeutics have undergone clinical trials for treating human cancers.

Ji-Hyun Lee, Dr.Ph., Mingyi Xie, Ph.D., Brad Barbazuk, Ph.D. and Tongjun Gu, Ph.D., study the importance of identifying the targets of the miRNAs to better understand the function and regulation of miRNAs in a new Scientific Reports publication published November 10, 2021. In their earlier work, an advance Artificial Intelligence method, named miTAR (a short name for miRNA Target), was developed for predicting miRNA targets. The approach integrates two major types of deep learning algorithms: convolutional neural networks (CNNs) and recurrent neural networks (RNNs). They found that miTAR performs better than the existing methods.

In this study, the researchers demonstrated that miTAR captures known features, including the involvement of seed region and the free energy, as well as multiple novel features, in the miRNA:target interactions.

Interestingly, the CNN and RNN layers of the model perform differently at capturing the free energy feature: the units in RNN layer is more unique at capturing the feature, but collectively the CNN layer is more efficient at capturing the feature.

“Not all AI or Deep learning algorithms can be explainable biologically,” said Ji Hyun Lee, Dr.Ph., “However, the black boxes in the middle need to be opened if we can.┬áIt is because the interpretation of deep learning models is not only critical for developing robust and reliable new models, but also important for understanding the underlying biological mechanisms.”

READ THE PAPER IN SCIENTIFIC REPORTS

Tongjun Gu, Ph.D., conceived the study, performed the analysis, and drafted the manuscript. Ji-Hyun Lee, Dr.Ph, supervised the study and contributed to the statistics analysis. Tongjun Gu, Ph.D., Mingyi Xie, Ph.D., Brad Barbazuk, Ph.D, and Ji-Hyun Lee, Dr.Ph., interpreted the results and edited the manuscript. Tongjun Gu, Ph.D., and Ji-Hyun Lee, Dr.Ph, are co-corresponding authors.