Using Attribution Sequence Alignment to Interpret Deep Learning Models for miRNA Binding Site Prediction

SIMPLE SUMMARY: MicroRNAs are small non-coding RNAs that play a central role in many molecular processes, but the exact rules of their activity are not known. In recent years, deep learning computational methods have revolutionized many fields, including the microRNA field. While making accurate predictions is important in biomedical tasks, it is equally important to understand why models make their predictions. Here, we present a novel interpretation technique for deep learning models that produces human readable visual representation of the knowledge learned by the model. This representation is useful for understanding the model’s decisions and can be used as a proxy for the further interpretation of biological concepts learned by the deep learning model. Importantly, the presented method is not tied to the model or biological domain and can be easily extended. ABSTRACT: MicroRNAs (miRNAs) are small non-coding RNAs that play a central role in the post-transcriptional regulation of biological processes. miRNAs regulate transcripts through direct binding involving the Argonaute protein family. The exact rules of binding are not known, and several in silico miRNA target prediction methods have been developed to date. Deep learning has recently revolutionized miRNA target prediction. However, the higher predictive power comes with a decreased ability to interpret increasingly complex models. Here, we present a novel interpretation technique, called attribution sequence alignment, for miRNA target site prediction models that can interpret such deep learning models on a two-dimensional representation of miRNA and putative target sequence. Our method produces a human readable visual representation of miRNA:target interactions and can be used as a proxy for the further interpretation of biological concepts learned by the neural network. We demonstrate applications of this method in the clustering of experimental data into binding classes, as well as using the method to narrow down predicted miRNA binding sites on long transcript sequences. Importantly, the presented method works with any neural network model trained on a two-dimensional representation of interactions and can be easily extended to further domains such as protein–protein interactions.