Publication: Coelho et al., CRISPR GUARD protects off-target sites from Cas9 nuclease activity using short guide RNAs. Nature Communications (2020). https://doi.org/10.1038/s41467-020-17952-5
About: https://www.sanger.ac.uk/tool/crispr-guard-finder/
Online tool: https://www.sanger.ac.uk/science/tools/crisprguardfinder/crisprguardfinder/
Given an on-target CRISPR guide this tool will identify all of its off-targets and design “GUARD” sequences (short guides) to interfere with the off-target activity. This is the code referred to in Coelho et al [1].
The tool incorporates an off-target search based on the one used in the Sanger WGE website [2, 3] enhanced with the calculation of the probability of the off-target [4]. R is used to identify and score the GUARD designs, and the whole is coordinated by a nextflow script.
- Coelho et al., CRISPR GUARD protects off-target sites from Cas9 nuclease activity using short guide RNAs. Nature Communications (2020). https://doi.org/10.1038/s41467-020-17952-5
- WGE: a CRISPR database for genome engineering. - PubMed - NCBI
- WTSI Genome Editing
- Repurposing CRISPR as an RNA-guided platform for sequence-specific control of gene expression. - PubMed - NCBI
Create a new directory for the installation which I will denote guard_root in what follows.
Clone out the GitHub repo:
mkdir guard_root
cd guard_root
git clone https://github.com/MatthewACoelho/CRISPRGUARDfinder.git .
The pipeline requires R (version 4.0.0) with optparse (version 1.6.6) and BSgenome (version 1.56.0) packages for each of the genomes required, and nextflow (version 20.01.0). One way to get these if you don’t already have them is using conda:
conda create --prefix guard_root/ext
conda activate guard_root/ext
#install R
conda config --add channels conda-forge
conda config --set channel_priority strict
conda install -c conda-forge r-base
conda install -c bioconda bioconductor-bsgenome.mmusculus.ucsc.mm10
conda install -c bioconda nextflow
#Open R to install BSgenome for human (version1.4.3), and optparse
#When asked to update other packages, all/some/none - select "none"
R
if (!requireNamespace("BiocManager", quietly = TRUE))
install.packages("BiocManager")
BiocManager::install("BSgenome.Hsapiens.UCSC.hg38")
BiocManager::install("optparse")
quit()
#deactivate the environment
conda deactivate
The off-target search is a C program requiring 64-bit architecture, which can be compiled as follows:
gcc -mcmodel=medium -O4 -o bin/ot -pthread src/ot.c
The genome-related files and indexes are too large to include here, so before you can run the tool you will need to generate them. Please note that these url sources will change, so substitute with the new release link appropriately. Download the genome fasta file and GTF from Ensembl for each organism in which you are interested. For example Human:
curl http://ftp.ensembl.org/pub/current_fasta/homo_sapiens/dna/Homo_sapiens.GRCh38.dna.toplevel.fa.gz > guard_root/data/hg38.fa.gz
curl http://ftp.ensembl.org/pub/current_gtf/homo_sapiens/Homo_sapiens.GRCh38.100.chr.gtf.gz > guard_root/data/hg38.gtf.gz
and Mouse:
curl http://ftp.ensembl.org/pub/current_fasta/mus_musculus/dna/Mus_musculus.GRCm38.dna.toplevel.fa.gz > guard_root/data/mm10.fa.gz
curl http://ftp.ensembl.org/pub/current_gtf/mus_musculus/Mus_musculus.GRCm38.96.gtf.gz > guard_root/data/mm10.gtf.gz
To create an off-target index for a genome $genome for PAM $pam:
mkdir -p guard_root/data/$genome/$pam
gunzip guard_root/data/$genome.fa.gz
guard_root/bin/ot index -out guard_root/data/$genome/$pam -pam $pam guard_root/data/$genome.fa
You can remove some of the indexes to reduce noise in the results, we normally only keep the major chromosome indexes:
rm guard_root/data/$genome/$pam/chrCHR*
rm guard_root/data/$genome/$pam/chrGL*
rm guard_root/data/$genome/$pam/chrK*
To convert an Ensembl GTF into the format required by the off-target search (which requires perl):
mkdir -p guard_root/data/$genome/info
gunzip guard_root/data/$genome.gtf.gz
# make scripts executable
chmod +x bin/process_gtf.pl
chmod +x bin/find_guards.sh
chmod +x bin/find_guards.R
chmod +x bin/find_guards.nf
chmod +x bin/local.nf
chmod +x bin/slurm.nf
guard_root/bin/process_gtf.pl guard_root/data/$genome/info < guard_root/data/$genome.gtf
Activate the environment:
conda activate guard_root/ext
if required, make sure an environment variable guard_root is set appropriately and R and nextflow are in your $PATH.
if required and env not set correctly, the error will be "Fatal No guard_root"
export guard_root=$PATH/guard_root
For example:
export guard_root=/Users/your-username/guard_root
Create a directory for the run, called results
mkdir results
cd results
and within it, make a params.nf file like:
params {
/* Name of this run - for file naming */
id = "VEGFA"
/* On-target guide sequence without PAM */
guide = "GACCCCCTCCACCCCGCCTC"
/* On-target guide location if there is > 1 with 0 mismatches */
chr = "" /* chr4 for example */
start = ""
end = ""
strand = "" /* + or - */
/* Genome version - hg38 or mm10 */
genome = "hg38"
/* PAM sequence - must correspond to one of the off-target indexes */
pam = "NGG"
/* Mismatches for the guide off-target search */
guide_mismatches = 5
/* Filter for which off-targets to take forward to guard design */
guide_min_pvalue = 0.1
/* Desired length of guard */
guard_length = 14
/* Mismatches for the guard off-target search */
guard_mismatches = 3
/* Maximum distance from a designed guard to its off-target guide */
max_guard_distance = 10
}
In the results directory, run:
guard_root/bin/find_guards.sh
which will run the process locally, or:
guard_root/bin/find_guards.sh --submit
which will submit jobs to slurm.
In both cases the script will pick up cached results if they exist, so add --force to force recalculation.
When complete you should have a text file named id_final.txt, e.g. VEGFA_final.txt.
We have developed a Shiny app for ease of use with user inputs for parameters to avoid making a parameter file manually for each query. The Shiny app requires further installation of the R packages "shiny" (version 1.4.0.2), "tidyverse" (version 1.3.0), "digest" (version 0.6.25) and "DT" (version. 0.13), which can be installed in R with 'install.packages("package_name")'. The app calls the scripts from your '/guard_root/bin' directory and generates results dataframes and a results download link.
The output is a tab-delimited text file with the following columns:
- Guard: sequence of the GUARD
- ID: unique identifier for the off-target this guard relates to - a unique off-target number followed by the gene, it hits if there is one
- GuardChr, GuardStart, GuardStop, GuardStrand: genomic location and strand of the GUARD
- GuardWithPAM: GUARD with PAM from the + strand view
- GuardGC: GC content of the GUARD
- ForwardGuardPAM: GUARD PAM on the GUARD strand
- ForwardGuardWithPAM: GUARD with PAM on the GUARD strand
- OffGuideOverlap, OffGuideSeedOverlap, OffGuidePAMOverlap: number of bases the GUARD overlaps with the off-target guide_seed region_PAM
- OffGuide: sequence of the off-target guide
- OffGuideGC: GC content of the off-target guide
- OffGuideStrand: Strand of the off-target guide
- GuardToOffGuidePAMDistance: base distance between the guard and off-target guide PAMs
- pCoding, nCoding: p-value and number of off-targets of the GUARD hitting coding (CDS) regions
- pNonCoding, nNonCoding: p-value and number of off-targets of the GUARD hitting non-coding regions
- nBadSeed: number of off-targets for the GUARD where the seed region is identical to the on-target
- X0, X1, X2, X3: number of genomic hits of the GUARD with 0, 1, 2, 3 mismatches.
- OnTargetHit, OnTargetMismatches, OnTargetSeedMismatches: sequence of the potential on-target hit of the GUARD
- PAMAndSeedScore, OnTargetScore, OffTargetScore, GCScore: components of the final GUARD score
The code is freely available under the MIT license .
#ngs/crispr