Back to Search
Start Over
Deep sampling of gRNA in the human genome and deep-learning-informed prediction of gRNA activities.
- Source :
- Cell Discovery; 5/16/2023, Vol. 9 Issue 1, p1-20, 20p
- Publication Year :
- 2023
-
Abstract
- Life science studies involving clustered regularly interspaced short palindromic repeat (CRISPR) editing generally apply the best-performing guide RNA (gRNA) for a gene of interest. Computational models are combined with massive experimental quantification on synthetic gRNA-target libraries to accurately predict gRNA activity and mutational patterns. However, the measurements are inconsistent between studies due to differences in the designs of the gRNA-target pair constructs, and there has not yet been an integrated investigation that concurrently focuses on multiple facets of gRNA capacity. In this study, we analyzed the DNA double-strand break (DSB)-induced repair outcomes and measured SpCas9/gRNA activities at both matched and mismatched locations using 926,476 gRNAs covering 19,111 protein-coding genes and 20,268 non-coding genes. We developed machine learning models to forecast the on-target cleavage efficiency (AIdit_ON), off-target cleavage specificity (AIdit_OFF), and mutational profiles (AIdit_DSB) of SpCas9/gRNA from a uniformly collected and processed dataset by deep sampling and massively quantifying gRNA capabilities in K562 cells. Each of these models exhibited superlative performance in predicting SpCas9/gRNA activities on independent datasets when benchmarked with previous models. A previous unknown parameter was also empirically determined regarding the "sweet spot" in the size of datasets used to establish an effective model to predict gRNA capabilities at a manageable experimental scale. In addition, we observed cell type-specific mutational profiles and were able to link nucleotidylexotransferase as the key factor driving these outcomes. These massive datasets and deep learning algorithms have been implemented into the user-friendly web service http://crispr-aidit.com to evaluate and rank gRNAs for life science studies. [ABSTRACT FROM AUTHOR]
Details
- Language :
- English
- ISSN :
- 20565968
- Volume :
- 9
- Issue :
- 1
- Database :
- Complementary Index
- Journal :
- Cell Discovery
- Publication Type :
- Academic Journal
- Accession number :
- 163740452
- Full Text :
- https://doi.org/10.1038/s41421-023-00549-9