Use of Computational Technologies to Define Potential CRISPR/Cas9 Off-Target Effects of Anti-HIV-1 Grnas in the Human Genome

Despite antiretroviral therapy, human immunodeficiency virus type 1 (HIV-1) infection remains a life-long clinical problem due to reservoirs harboring proviral DNA in a latent/persistent form. Recently, gene-editing strategies utilizing the CRISPR/Cas9 system (CC9) have been developed to excise the HIV-1 genome from infected cells thereby achieving a single cell cure for HIV infection. However, due to the promiscuity of the guide RNA (gRNA) targeting, one necessary area of interest has focused on off-target activity that may cause unwanted DNA damage across the human genome. This is further complicated by the uncertainty of DNA accessibility such as the chromatin state in relevant cell populations as well as genetic variability across the human population. In order to predict the off-target effect, we have developed a screening method employing a bloom filter library constructed from a set of adapted human genomes that contains the information of all individuals within the 1000 Genomes project. In order to test the results of our method, publically available Genome-wide Unbiased Identification of Double-stranded-breaks Enabled by sequencing (GUIDE-seq) data was used from the use of HEK293T cells. Results of this analysis have shown that, in a given genomic region, the DNA accessibility implied by DNase-I hypersensitivity positively correlated to the frequency of CC9 activity. Integrating quantitative off-target data available from publications using a two-layer stacked regression model allowed better understanding of generalizable factors that contribute to off-target effect. Compiling these results will greatly increase the ability of researchers to design gRNAs that are effective with respect to targeting integrated HIV-1 proviral DNA while avoiding off-target effects.