Comprehensive computational analysis of epigenetic descriptors affecting CRISPR-Cas9 off-target activity

Abstract

Background A common issue in CRISPR-Cas9 genome editing is off-target activity, which prevents the widespread use of CRISPR-Cas9 in medical applications. Among other factors, primary chromatin structure and epigenetics may influence off-target activity. Methods In this work, we utilize crisprSQL, an off-target database, to analyze the effect of 19 epigenetic descriptors on CRISPR-Cas9 off-target activity. Termed as 19 epigenetic features/scores, they consist of 6 experimental epigenetic and 13 computed nucleosome organization-related features. In terms of novel features, 15 of the epigenetic scores are newly considered. The 15 newly considered scores consist of 13 freshly computed nucleosome occupancy/positioning scores and 2 experimental features (MNase and DRIP). The other 4 existing scores are experimental features (CTCF, DNase I, H3K4me3, RRBS) commonly used in deep learning models for off-target activity prediction. For data curation, MNase was aggregated from existing experimental nucleosome occupancy data. Based on the sequence context information available in crisprSQL, we also computed nucleosome occupancy/positioning scores for off-target sites. Results To investigate the relationship between the 19 epigenetic features and off-target activity, we first conducted Spearman and Pearson correlation analysis. Such analysis shows that some computed scores derived from training-based models and training-free algorithms outperform all experimental epigenetic features. Next, we evaluated the contribution of all epigenetic features in two successful machine/deep learning models which predict off-target activity. We found that some computed scores, unlike all 6 experimental features, significantly contribute to the predictions of both models. As a practical research contribution, we make the off-target dataset containing all 19 epigenetic features available to the research community. Conclusions Our comprehensive computational analysis helps the CRISPR-Cas9 community better understand the relationship between epigenetic features and CRISPR-Cas9 off-target activity.

piCRISPR: Physically Informed Deep Learning Models for CRISPR/Cas9 Off-Target Cleavage Prediction

“…All underlying distributions are shown in Figure S9. tivity, supporting what has been observed in preliminary, non-sequence based models in [10].…”

“…When considering nucleosome positioning-related feature channels, we see that the 147 bp GC content around each nucleotide has a net positive influence on cleavage activity. Similar to the argument in [10], this can be attributed to the increased bendability of GC-rich DNA [37] which is beneficial to Cas9 sequence readout during binding [38]. We further observe that the NuPoP Affinity score ranks higher in terms of global SHAP value than most sequence features.…”

“…The negative influence of nucleosome affinity can be explained by the reduced accessibility of high-affinity DNA regions, and is observed strongly between nucleotides 5 and 19. This effect has been observed in [10] as well. We further observe an overall negative influence of low Nucleotide BDM values on cleavage ac- Comparison between piCRISPR, CRISPR-Net [13] and CRISPR-IP [15], with comparison values for the latter two models taken from [15].…”

“…Besides these established features, we propose the usage of nucleosome organisation-related features [10] which add an unprecedented level of sensitivity towards the biological environment of the cleavage site (see Figure 1C). In this publication, we trained a preliminary cleavage prediction model on 13 distinct nucleosome organisation-related scores all based on the 147 bp context around each (off-)target nucleotide (see Supplementary Material) as well as the four literature-standard epigenetic markers named above and include the three scores of highest feature importance: GC count, Nucleotide BDM [19] and NuPoP Affinity [8].…”

“…Capturing this environment is therefore instrumental in providing accurate predictions of cleavage activity. We recently found that computed nucleosome organisation-related features correlate better with cleavage frequency values than experimental epigenetic markers (DNase I, RRBS, CTCF, H3K4me3) [10] which have heretofore been the literature standard for cleavage prediction models. These computed features also surpassed epigenetic markers in terms of their feature importance in preliminary cleavage activity prediction models which had access to both computed and experimental epigenetic features.…”