A deep neural network for G-quadruplexes binding proteins classification

In the last years, a plethora of studies unveiled the fundamental roles of RNA G-quadruplexes (RG4s), unique structural RNA strands featuring guanine-rich nucleic acid sequences, in basic cellular processes, as well as in the pathogenesis of important diseases, such as cancer and neurodegeneration. As the knowledge of the pathological roles played by RG4s has grown wider, the involvement of RG4-binding proteins in designing new diagnostic and therapeutic strategies has become increasingly recognized, but the classification of these proteins is still challenging. In this paper, we describe the architecture and the training procedure of a deep neural network based on Long Short-Term Memory layers to classify RG4-binding proteins. The impressive classification accuracy achieved on the test set provides a strong foundation for future investigations on more data samples and several experimental purposes.