Development of cnn model for prediction of crispr/cas12 guide rna activity

Abstract

CRISPR technology is a powerful gene modifying tool that was derived from bacteria and archea. This CRISPR tool recognizes and cleaves any target DNA with the usage of a guide RNA and Cas endonuclease enzyme. Cpf1 endonuclease enzyme is a type of Cas enzyme from class 2 in the CRISPR-Cas system used for cutting DNA. The success of this genetic engineering tool depends on design of effective single guide RNA (sgRNA) to help Cas endonuclease enzyme carry out DNA cleavage. Different computational tools have been created to help with design of sgRNA however, it still remains a challenge. The application of deep learning has been extended to CRISPR technology. Here, we developed a deep learning algorithm using convolutional neural network which predicts activity of Cpf1 guide RNA. We built different models using different hyperparameters and identified the best model capable of predicting Cpf1guide RNA activity. ConvCpf1 will help in design of sgRNAs that will be highly active for genome editing using Cpf1 endonuclease enzyme, thereby saving time as well as experimental cost.