Base functions and classes for CRISPR gRNA design
2022-11-01
1 Overview
The crisprBase package is a core package of the crisprVerse ecosystem that provides S4 classes for
representing CRISPR nucleases and base editors. It also provides arithmetic
functions to extract genomic ranges to help with the design and manipulation
of CRISPR guide-RNAs (gRNAs). The classes and functions are designed to work
with a broad spectrum of nucleases and applications, including
PAM-free CRISPR nucleases, RNA-targeting nucleases, and the more
general class of restriction enzymes. It also includes functionalities for
CRISPR nickases.
It provides a language and convention for our gRNA design ecosystem described in our recent bioRxiv preprint: “The crisprVerse: a comprehensive Bioconductor ecosystem for the design of CRISPR guide RNAs across nucleases and technologies”
2 Installation
2.1 Software requirements
2.1.1 OS Requirements
This package is supported for macOS, Linux and Windows machines. It was developed and tested on R version 4.2.
2.2 Installation
crisprBase can be installed from the Bioconductor devel branch by typing the
following commands inside of an R session:
if (!require("BiocManager", quietly = TRUE))
install.packages("BiocManager")
BiocManager::install(version="devel")
BiocManager::install("crisprBase")2.2.1 Getting started
We load crisprBase in the usual way:
library(crisprBase)3 Nuclease class
The Nuclease class is designed to store minimal information about the
recognition sites of general nucleases, such as restriction enzymes.
The Nuclease class has 5 fields: nucleaseName, targetType,
metadata, motifs and weights.
The nucleaseName field is a string specifying a name
for the nuclease.
The targetType specifies if the nuclease targets “DNA” (deoxyribonucleases)
or “RNA” (ribonucleases). The metadata field is a list of arbitrary length
to store additional information about the nuclease.
The motifs field is a character vector that specify one of several DNA
sequence motifs that are recognized by the nuclease for cleavage (always in
the 5’ to 3’ direction). The optional weights field is a numeric vector
specifying relative cleavage probabilities corresponding to the motifs
specified by motifs. Note that we use DNA to represent motifs irrespectively
of the target type for simplicity.
We use the Rebase convention to represent motif sequences (Roberts et al. 2010).
For enzymes that cleave within the recognition site,
we add the symbol ^ within the recognition sequence to specify
the cleavage site, always in the 5’ to 3’ direction. For enzymes that
cleave away from the recognition site, we specify the distance of the cleavage
site using a (x/y) notation where x
represents the number of nucleotides away from the recognition sequence on the
original strand, and y represents the number of nucleotides away from the
recognition sequence on the reverse strand.
3.1 Examples
The EcoRI enzyme recognizes the palindromic motif GAATTC, and cuts after
the first nucleotide, which is specified using the ^ below:
library(crisprBase)
EcoRI <- Nuclease("EcoRI",
targetType="DNA",
motifs=c("G^AATTC"),
metadata=list(description="EcoRI restriction enzyme"))The HgaI enzyme recognizes the motif GACGC, and cleaves DNA at 5
nucleotides downstream of the recognition sequence on the original strand,
and at 10 nucleotides downstream of the recognition sequence on the reverse
strand:
HgaI <- Nuclease("HgaI",
targetType="DNA",
motifs=c("GACGC(5/10)"),
metadata=list(description="HgaI restriction enzyme"))In case the cleavage site was upstream of the recognition sequence, we would
instead specify (5/10)GACGC.
Note that any nucleotide letter that is part of the extended IUPAC nucleic acid
code can be used to represent recognition motifs. For instance, we use Y
and R (pyrimidine and purine, respectively) to specify the possible
recognition sequences for PfaAI:
PfaAI <- Nuclease("PfaAI",
targetType="DNA",
motifs=c("G^GYRCC"),
metadata=list(description="PfaAI restriction enzyme"))3.2 Accessor functions
The accessor function motifs retrieve the motif
sequences:
motifs(PfaAI)## DNAStringSet object of length 1:
## width seq
## [1] 6 GGYRCCTo expand the motif sequence into all combinations of valid sequences with only
A/C/T/G nucleotides, users can use expand=TRUE.
motifs(PfaAI, expand=TRUE)## DNAStringSet object of length 4:
## width seq names
## [1] 6 GGCACC GGYRCC
## [2] 6 GGTACC GGYRCC
## [3] 6 GGCGCC GGYRCC
## [4] 6 GGTGCC GGYRCC
Figure 1: Examples of restriction enzymes
4 CrisprNuclease class
CRISPR nucleases are examples of RNA-guided nucleases. For cleavage, it requires two binding components. For CRISPR nucleases targeting DNA, the nuclease needs to first recognize a constant nucleotide motif in the target DNA called the protospacer adjacent motif (PAM) sequence. Second, the guide-RNA (gRNA), which guides the nuclease to the target sequence, needs to bind to a complementary sequence adjacent to the PAM sequence (protospacer sequence). The latter can be thought of a variable binding motif that can be specified by designing corresponding gRNA sequences. For CRISPR nucleases targeting RNA, the equivalent of the PAM sequence is called the Protospacer Flanking Sequence (PFS). We use the terms PAM and PFS interchangeably as it should be clear from context.
The CrisprNuclease class allows to characterize both binding components by
extending the Nuclease class to contain information about the gRNA
sequences.The PAM sequence characteristics, and the cleavage distance with
respect to the PAM sequence, are specified using the motif nomenclature
described in the Nuclease section above.
3 additional fields are required: pam_side, spacer_length and
spacer_gap. The pam_side field can only take 2 values, 5prime and
3prime, and specifies on which side the PAM sequence is located with
respect to the protospacer sequence. While it would be more appropriate to use
the terminology pfs_side for RNA-targeting nucleases, we still use the term
pam_side for simplicity.
The spacer_length specifies
a default spacer length, and the spacer_gap specifies a distance
(in nucleotides) between the PAM (or PFS) sequence and spacer sequence.
For most nucleases,spacer_gap=0 as the spacer sequence is located directly
next to the PAM/PFS sequence.
We show how we construct a CrisprNuclease object for the commonly-used Cas9
nuclease (Streptococcus pyogenes Cas9):
SpCas9 <- CrisprNuclease("SpCas9",
targetType="DNA",
pams=c("(3/3)NGG", "(3/3)NAG", "(3/3)NGA"),
weights=c(1, 0.2593, 0.0694),
metadata=list(description="Wildtype Streptococcus pyogenes Cas9 (SpCas9) nuclease"),
pam_side="3prime",
spacer_length=20)
SpCas9## Class: CrisprNuclease
## Name: SpCas9
## Target type: DNA
## Metadata: list of length 1
## PAMs: NGG, NAG, NGA
## Weights: 1, 0.2593, 0.0694
## Spacer length: 20
## PAM side: 3prime
## Distance from PAM: 0
## Prototype protospacers: 5'--SSSSSSSSSSSSSSSSSSSS[NGG]--3', 5'--SSSSSSSSSSSSSSSSSSSS[NAG]--3', 5'--SSSSSSSSSSSSSSSSSSSS[NGA]--3'Similar to the Nuclease class, we can specify PAM sequences using the
extended nucleotide code. SaCas9 serves as a good example:
SaCas9 <- CrisprNuclease("SaCas9",
targetType="DNA",
pams=c("(3/3)NNGRRT"),
metadata=list(description="Wildtype Staphylococcus
aureus Cas9 (SaCas9) nuclease"),
pam_side="3prime",
spacer_length=21)
SaCas9## Class: CrisprNuclease
## Name: SaCas9
## Target type: DNA
## Metadata: list of length 1
## PAMs: NNGRRT
## Weights: 1
## Spacer length: 21
## PAM side: 3prime
## Distance from PAM: 0
## Prototype protospacers: 5'--SSSSSSSSSSSSSSSSSSSSS[NNGRRT]--3'Here is another example where we construct a CrisprNuclease object for the
commonly-used Cas12a nuclease (AsCas12a):
AsCas12a <- CrisprNuclease("AsCas12a",
targetType="DNA",
pams="TTTV(18/23)",
metadata=list(description="Wildtype Acidaminococcus
Cas12a (AsCas12a) nuclease."),
pam_side="5prime",
spacer_length=23)
AsCas12a## Class: CrisprNuclease
## Name: AsCas12a
## Target type: DNA
## Metadata: list of length 1
## PAMs: TTTV
## Weights: 1
## Spacer length: 23
## PAM side: 5prime
## Distance from PAM: 0
## Prototype protospacers: 5'--[TTTV]SSSSSSSSSSSSSSSSSSSSSSS--3'4.1 CrisprNuclease objects provided in CrisprBase
Several already-constructed crisprNuclease objects are available
in crisprBase, see data(package="crisprBase").
5 CRISPR arithmetics
5.1 CRISPR terminology
The terms spacer and protospacer are not interchangeable. spacer refers to the sequence used in the gRNA construct to guide the Cas nuclease to the target protospacer sequence in the host genome / transcriptome. The protospacer sequence is adjacent to the PAM sequence / PFS sequence. We use the terminology target sequence to refer to the protospacer and PAM sequence taken together. For DNA-targeting nucleases such as Cas9 and Cas12a, the spacer and protospacer sequences are identical from a nucleotide point of view. For RNA-targeting nucleases such as Cas13d, the spacer and protospacer sequences are the reverse complement of each other.
An gRNA spacer sequence does not always uniquely target the host genome (a given sgRNA spacer can map to multiple protospacers in the genome). However, for a given reference genome, protospacer sequences can be uniquely identified using a combination of 3 attributes:
- chr: chromosome name
- strand: forward (+) or reverse (-)
- pam_site: genomic coordinate of the first nucleotide of the
nuclease-specific PAM sequence. For SpCas9, this corresponds to the genomic
coordinate of N in the NGG PAM sequence. For AsCas12a, this corresponds to the
genomic coordinate of the first T nucleotide in the TTTV PAM sequence.
For RNA-targeting nucleases, this corresponds to the first nucleotide of
the PFS (we do not use
pfs_sitefor simplicity).
Figure 2: Examples of CRISPR nucleases
5.2 Cut site
For convention, we used the nucleotide directly downstream of the DNA cut to represent the cut site nucleotide position. For instance, for SpCas9 (blunt-ended dsDNA break), the cut site occurs at position -3 with respect to the PAM site. For AsCas12a, the 5nt overhang dsDNA break occurs at 18 nucleotides after the PAM sequence on the targeted strand. Therefore the cute site on the forward strand occurs at position 22 with respect to the PAM site, and at position 27 on the reverse strand.
The convenience function cutSites extracts the cut site coordinates
relative to the PAM site:
data(SpCas9, package="crisprBase")
data(AsCas12a, package="crisprBase")
cutSites(SpCas9)## [1] -3cutSites(SpCas9, strand="-")## [1] -3cutSites(AsCas12a)## [1] 22cutSites(AsCas12a, strand="-")## [1] 27Below is an illustration of how different motif sequences and cut patterns translate into cut site coordinates with respect to a PAM sequence NGG:
Figure 3: Examples of cut site coordinates
5.3 Obtaining spacer and PAM sequences from target sequences
Given a list of target sequences (protospacer + PAM) and a CrisprNuclease object, one can extract protospacer and PAM sequences using the functions
extractProtospacerFromTarget and extractPamFromTarget, respectively.
targets <- c("AGGTGCTGATTGTAGTGCTGCGG",
"AGGTGCTGATTGTAGTGCTGAGG")
extractPamFromTarget(targets, SpCas9)## [1] "CGG" "AGG"extractProtospacerFromTarget(targets, SpCas9)## [1] "AGGTGCTGATTGTAGTGCTG" "AGGTGCTGATTGTAGTGCTG"5.4 Obtaining genomic coordinates of protospacer sequences using PAM site coordinates
Given a PAM coordinate, there are several functions in crisprBase that
allows to get get coordinates of the full PAM sequence, protospacer sequence, and target sequence: getPamRanges, getTargetRanges, and
getProtospacerRanges, respectively. The output objects are GRanges:
chr <- rep("chr7",2)
pam_site <- rep(200,2)
strand <- c("+", "-")
gr_pam <- getPamRanges(seqnames=chr,
pam_site=pam_site,
strand=strand,
nuclease=SpCas9)
gr_protospacer <- getProtospacerRanges(seqnames=chr,
pam_site=pam_site,
strand=strand,
nuclease=SpCas9)
gr_target <- getTargetRanges(seqnames=chr,
pam_site=pam_site,
strand=strand,
nuclease=SpCas9)
gr_pam## GRanges object with 2 ranges and 0 metadata columns:
## seqnames ranges strand
## <Rle> <IRanges> <Rle>
## [1] chr7 200-202 +
## [2] chr7 198-200 -
## -------
## seqinfo: 1 sequence from an unspecified genome; no seqlengthsgr_protospacer## GRanges object with 2 ranges and 0 metadata columns:
## seqnames ranges strand
## <Rle> <IRanges> <Rle>
## [1] chr7 180-199 +
## [2] chr7 201-220 -
## -------
## seqinfo: 1 sequence from an unspecified genome; no seqlengthsgr_target## GRanges object with 2 ranges and 0 metadata columns:
## seqnames ranges strand
## <Rle> <IRanges> <Rle>
## [1] chr7 180-202 +
## [2] chr7 198-220 -
## -------
## seqinfo: 1 sequence from an unspecified genome; no seqlengthsand for AsCas12a:
gr_pam <- getPamRanges(seqnames=chr,
pam_site=pam_site,
strand=strand,
nuclease=AsCas12a)
gr_protospacer <- getProtospacerRanges(seqnames=chr,
pam_site=pam_site,
strand=strand,
nuclease=AsCas12a)
gr_target <- getTargetRanges(seqnames=chr,
pam_site=pam_site,
strand=strand,
nuclease=AsCas12a)
gr_pam## GRanges object with 2 ranges and 0 metadata columns:
## seqnames ranges strand
## <Rle> <IRanges> <Rle>
## [1] chr7 200-203 +
## [2] chr7 197-200 -
## -------
## seqinfo: 1 sequence from an unspecified genome; no seqlengthsgr_protospacer## GRanges object with 2 ranges and 0 metadata columns:
## seqnames ranges strand
## <Rle> <IRanges> <Rle>
## [1] chr7 204-226 +
## [2] chr7 174-196 -
## -------
## seqinfo: 1 sequence from an unspecified genome; no seqlengthsgr_target## GRanges object with 2 ranges and 0 metadata columns:
## seqnames ranges strand
## <Rle> <IRanges> <Rle>
## [1] chr7 200-226 +
## [2] chr7 174-200 -
## -------
## seqinfo: 1 sequence from an unspecified genome; no seqlengths6 BaseEditor class
Base editors are inactive Cas nucleases coupled with a specific deaminase. For instance, the first cytosine base editor (CBE) was obtained by coupling a cytidine deaminase with dCas9 to convert Cs to Ts (Komor et al. 2016).
Figure 4: Examples of base editors
We provide in crisprBase a S4 class, BaseEditor, to represent base editors.
It extends the CrisprNuclase class with 3 additional fields:
baseEditorName: string specifying the name of the base editor.editingStrand: strand where the editing happens with respect to the target protospacer sequence (“original” or “opposite”).editingWeights: a matrix of experimentally-derived editing weights.
We now show how to build a BaseEditor object with the CBE base editor BE4max
with weights obtained from Arbab et al. (2020).
We first obtain a matrix of weights for the BE4max editor stored in
the package crisprBase:
# Creating weight matrix
weightsFile <- system.file("be/b4max.csv",
package="crisprBase",
mustWork=TRUE)
ws <- t(read.csv(weightsFile))
ws <- as.data.frame(ws)The row names of the matrix must correspond to the nucleotide substitutions Nucleotide substitutions that are not present in the matrix will have weight assigned to 0.
rownames(ws)## [1] "Position" "C2A" "C2G" "C2T" "G2A" "G2C"The column names must correspond to the relative position with respect to the PAM site.
colnames(ws) <- ws["Position",]
ws <- ws[-c(match("Position", rownames(ws))),,drop=FALSE]
ws <- as.matrix(ws)
head(ws)## -36 -35 -34 -33 -32 -31 -30 -29 -28 -27 -26 -25 -24 -23 -22 -21 -20 -19
## C2A 0.0 0.0 0.0 0.7 0.1 0.2 0.0 0.2 0.3 0.0 0.2 0.0 0.9 0.0 0.1 0.2 0.1 0.3
## C2G 0.9 0.1 0.1 0.0 0.3 0.7 0.1 0.1 0.7 0.0 0.4 0.1 0.1 0.1 0.1 0.1 0.0 0.5
## C2T 0.7 0.7 0.8 1.8 1.0 2.0 1.4 1.2 2.3 1.3 2.4 2.2 3.4 2.2 2.1 3.5 5.8 16.2
## G2A 0.0 0.0 0.5 0.0 0.0 0.3 0.4 1.1 0.9 0.6 0.3 1.7 0.7 0.8 0.1 0.3 0.1 0.0
## G2C 0.1 0.0 0.0 0.0 0.6 2.8 0.0 0.0 0.3 0.2 0.2 0.1 0.0 0.3 0.0 0.0 0.0 0.0
## -18 -17 -16 -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3
## C2A 1.0 2.0 2.7 3.00 2.7 1.9 0.8 0.6 0.3 0.0 0.1 0.1 0.1 0.0 0.0 0.0
## C2G 1.3 2.7 4.7 5.40 5.6 3.9 1.7 0.6 0.6 0.4 0.5 0.1 0.0 0.1 0.0 0.0
## C2T 31.8 63.2 90.3 100.00 87.0 62.0 31.4 16.3 10.0 5.6 3.3 1.9 1.8 2.4 1.7 0.5
## G2A 0.0 0.0 0.1 0.01 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.2 0.2 0.0
## G2C 0.0 0.0 0.2 0.00 0.0 0.1 0.1 0.2 0.2 0.0 0.0 0.0 0.1 0.0 0.0 0.0
## -2 -1
## C2A 0.0 0.0
## C2G 0.0 0.0
## C2T 0.2 0.1
## G2A 0.0 0.1
## G2C 0.0 0.0Since BE4max uses Cas9, we can use the SpCas9 CrisprNuclease object
available in crisprBase to build the BaseEditor object:
data(SpCas9, package="crisprBase")
BE4max <- BaseEditor(SpCas9,
baseEditorName="BE4max",
editingStrand="original",
editingWeights=ws)
metadata(BE4max)$description_base_editor <- "BE4max cytosine base editor."
BE4max## Class: BaseEditor
## CRISPR Nuclease name: SpCas9
## Target type: DNA
## Metadata: list of length 2
## PAMs: NGG, NAG, NGA
## Weights: 1, 0.2593, 0.0694
## Spacer length: 20
## PAM side: 3prime
## Distance from PAM: 0
## Prototype protospacers: 5'--SSSSSSSSSSSSSSSSSSSS[NGG]--3', 5'--SSSSSSSSSSSSSSSSSSSS[NAG]--3', 5'--SSSSSSSSSSSSSSSSSSSS[NGA]--3'
## Base editor name: BE4max
## Editing strand: original
## Maximum editing weight: C2T at position -15One can quickly visualize the editing weights using the function
plotEditingWeights:
plotEditingWeights(BE4max)
7 CrisprNickase class
CRISPR nickases can be created by mutating one of the two nuclease domains of a CRISPR nuclease. They create single-strand breaks instead of double-strand breaks.
For instance, the D10A mutation of SpCas9 inactivates the RuvC domain, and the resulting CRISPR nickase (Cas9D10A) cleaves only the strand opposite to the protospacer sequence. The H840A mutation of SpCas9 inactivates the HNN domain, and the resulting CRISPR nickase (Cas9H840A) cleaves only the strand that contains the protospacer sequence. See Figure below.
Figure 5: Examples of CRISPR nickases
The CrisprNickase class in crisprBase works similar to the CrisprNuclease
class:
Cas9D10A <- CrisprNickase("Cas9D10A",
nickingStrand="opposite",
pams=c("(3)NGG", "(3)NAG", "(3)NGA"),
weights=c(1, 0.2593, 0.0694),
metadata=list(description="D10A-mutated Streptococcus
pyogenes Cas9 (SpCas9) nickase"),
pam_side="3prime",
spacer_length=20)
Cas9H840A <- CrisprNickase("Cas9H840A",
nickingStrand="original",
pams=c("(3)NGG", "(3)NAG", "(3)NGA"),
weights=c(1, 0.2593, 0.0694),
metadata=list(description="H840A-mutated Streptococcus
pyogenes Cas9 (SpCas9) nickase"),
pam_side="3prime",
spacer_length=20)The nickingStrand field indicates which strand is being cleaved by the
nickase.
8 RNA-targeting nucleases
RNA-targeting CRISPR nucleases, such as the Cas13 family of nucleases, target single-stranded RNA (ssRNA) instead of dsDNA as the name suggests. The equivalent of the PAM sequence is called Protospacer Flanking Sequence (PFS).
For RNA-targeting CRISPR nucleases, the spacer sequence is the reverse complement of the protospacer sequence. This differs from DNA-targeting CRISPR nucleases, for which the spacer and protospacer sequences are identical.
We can construct an RNA-targeting nuclease in way similar to a
DNA-targeting nuclease by specifying target="RNA". As an example,
we construct below a CrisprNuclease object for the CasRx nuclease (Cas13d
from Ruminococcus flavefaciens strain XPD3002):
CasRx <- CrisprNuclease("CasRx",
targetType="RNA",
pams="N",
metadata=list(description="CasRx nuclease"),
pam_side="3prime",
spacer_length=23)
CasRx## Class: CrisprNuclease
## Name: CasRx
## Target type: RNA
## Metadata: list of length 1
## PFS: N
## Weights: 1
## Spacer length: 23
## PFS side: 3prime
## Distance from PFS: 0
## Prototype protospacers: 5'--SSSSSSSSSSSSSSSSSSSSSSS[N]--3'9 Additional notes
9.1 dCas9 and other “dead” nucleases
The CRISPR inhibition (CRISPRi) and CRISPR activation (CRISPRa) technologies uses modified versions of CRISPR nucleases that lack endonuclease activity, often referred to as “dead Cas” nucleases, such as the dCas9.
While fully-active Cas nucleases and dCas nucleases differ in terms of
applications and type of genomic perturbations, the gRNA design remains
unchanged in terms of spacer sequence search and genomic coordinates. Therefore
it is convenient to use the fully-active version of the nuclease throughout
crisprBase.
10 License
The project as a whole is covered by the MIT license.
11 Reproducibility
sessionInfo()## R version 4.2.1 (2022-06-23)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 20.04.5 LTS
##
## Matrix products: default
## BLAS: /home/biocbuild/bbs-3.16-bioc/R/lib/libRblas.so
## LAPACK: /home/biocbuild/bbs-3.16-bioc/R/lib/libRlapack.so
##
## locale:
## [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
## [3] LC_TIME=en_GB LC_COLLATE=C
## [5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
## [7] LC_PAPER=en_US.UTF-8 LC_NAME=C
## [9] LC_ADDRESS=C LC_TELEPHONE=C
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
##
## attached base packages:
## [1] stats graphics grDevices utils datasets methods base
##
## other attached packages:
## [1] crisprBase_1.2.0 BiocStyle_2.26.0
##
## loaded via a namespace (and not attached):
## [1] Rcpp_1.0.9 XVector_0.38.0 knitr_1.40
## [4] magrittr_2.0.3 GenomicRanges_1.50.0 zlibbioc_1.44.0
## [7] IRanges_2.32.0 BiocGenerics_0.44.0 R6_2.5.1
## [10] rlang_1.0.6 fastmap_1.1.0 highr_0.9
## [13] GenomeInfoDb_1.34.0 stringr_1.4.1 tools_4.2.1
## [16] xfun_0.34 cli_3.4.1 jquerylib_0.1.4
## [19] htmltools_0.5.3 yaml_2.3.6 digest_0.6.30
## [22] crayon_1.5.2 bookdown_0.29 GenomeInfoDbData_1.2.9
## [25] BiocManager_1.30.19 bitops_1.0-7 sass_0.4.2
## [28] S4Vectors_0.36.0 RCurl_1.98-1.9 cachem_1.0.6
## [31] evaluate_0.17 rmarkdown_2.17 stringi_1.7.8
## [34] compiler_4.2.1 bslib_0.4.0 magick_2.7.3
## [37] Biostrings_2.66.0 stats4_4.2.1 jsonlite_1.8.3References
Arbab, Mandana, Max W Shen, Beverly Mok, Christopher Wilson, Żaneta Matuszek, Christopher A Cassa, and David R Liu. 2020. “Determinants of Base Editing Outcomes from Target Library Analysis and Machine Learning.” Cell 182 (2): 463–80.
Komor, Alexis C, Yongjoo B Kim, Michael S Packer, John A Zuris, and David R Liu. 2016. “Programmable Editing of a Target Base in Genomic Dna Without Double-Stranded Dna Cleavage.” Nature 533 (7603): 420–24.
Roberts, Richard J, Tamas Vincze, Janos Posfai, and Dana Macelis. 2010. “REBASE—a Database for Dna Restriction and Modification: Enzymes, Genes and Genomes.” Nucleic Acids Research 38 (suppl_1): D234–D236.