C-RNNCrispr: Prediction of CRISPR/Cas9 sgRNA activity using convolutional and recurrent neural networks

• DOI: 10.1016/j.csbj.2020.01.013
• Publication Date: 2020-02-12T11:13:55Z
• Authors: Guishan Zhang; Zhiming Dai; Xianhua Dai
• Keywords: Benchmark (surveying)

CRISPR/Cas9 is a hot genomic editing tool, but its success limited by the widely varying target efficiencies among different single guide RNAs (sgRNAs). In this study, we proposed C-RNNCrispr, hybrid convolutional neural networks (CNNs) and bidirectional gate recurrent unit network (BGRU) framework, to predict sgRNA on-target activity. C-RNNCrispr consists of two branches: branch epigenetic branch. The receives encoded binary matrix sequence four features as inputs, produces regression score. We introduced transfer learning approach using small-size datasets fine-tune model that were pre-trained from benchmark dataset, leading substantially improved predictive performance. Experiments on commonly used showed outperforms state-of-the-art methods in terms prediction accuracy generalization. Source codes are available at https://github.com/Peppags/C_RNNCrispr.