Novel Encoding of sgRNA-DNA Sequences for Accurate Deep Learning Off-Target Predictions

Off-target predictions are crucial in gene editing research to improve existing prediction methods. Recently, significant progress has been achieved in the field of prediction of off-target mutations, particularly with CRISPR-Cas9 data, thanks to the use of deep learning. CRISPR-Cas9 is a precise gene editing technique allowing manipulations of DNA fragments. The encoding of sgRNA-DNA sequences for deep neural networks is a complex process, which impacts significantly the prediction accuracy. In this context, we propose a novel encoding of sgRNA-DNA sequences that is capable to aggregate the involved sequence data without any loss of information. In our experiments, we compare our novel encoding with the state-of-the-art sgRNA-DNA encoding. We demonstrate the superior accuracy of our approach in our simulations involving Feedforward Neural Networks (FFNs) and Convolutional Neural Networks (CNNs). We highlight the universality of our results by building several FFNs and CNNs with various layer depths and performing predictions on two popular public gene editing data sets, the CRISPOR data set and the GUIDE-seq data set. In all our experiments, the new encoding led to more accurate off-target prediction results, providing an improvement of the area under the Receiver Operating Characteristic (ROC) curve up to 35\%.