Discovery of a non-canonical GRHL1 binding site using deep convolutional and recurrent neural networks.

Our results show that convolutional recurrent neural networks may uncover unanticipated binding sites and facilitate quantitative transcription factor binding predictions.