Free online tool for designing guides to edit DNA using CRISPR base editing

The system verifies the user-provided list of official gene names or the validity of the expected format for genomic regions based on the National Center for Biotechnology Information (NCBI) annotation standards. Genome assemblies are obtained from the International Nucleotide Sequence Database Collaboration (INSDC) (Supplementary Table S1). The gffutils and PyRanges Python packages are used to parse Ensembl genomic annotation databases [47], ensuring that queried regions are valid within the genome assembly. For Homo sapiens (hg38), Mus musculus (mm39), Caenorhabditis elegans (Ce11), and Drosophila melanogaster (dm6) genomes, ENCODE blacklisted regions are excluded [48]. All coding sequences associated with the provided gene names are extracted and validated, with flanking regions adjusted per user input. Upon users’ request, only the coordinates corresponding to Ensembl’s canonical isoform will be validated and retrieved. When the user inputs genomic coordinates, the pipeline validates the complete region and adds flanking regions as specified. For overlapping genomic regions, CRISPR-BEasy will return an error to prevent the design and downstream elimination of duplicated sgRNAs. Finally, genomic coordinates are written into a BED file [49].

The sgRNA design starts with inputting the genomic coordinates in the BED file and the selected Cas variant into the crisprDesign package within the crisprVerse Bioconductor ecosystem using R (version 4.4.3) [38]. Off-target alignment is determined using the crisprBowtie package within crisprVerse, and off-targets are filtered based on user-specified thresholds for the maximal count of off-targets and Cutting Frequency Determination (CFD) score [1]. Off-target filtering is only available for SpCas9 variants: The CFD score was developed for SpCas9, and we adjusted it for engineered SpCas9 variants with non-NGG PAM sequences. We assumed all SpCas9 variants would present similar tolerance to sgRNA–target sequence mismatches. Following a conservative approach, we adjusted the PAM mismatch penalty matrix by marking allowable PAMs as 100% cutting efficiency (e.g. SpG Cas9 will have 100% cutting efficiency in NGG, NGA, NGC, and NGT PAMs and 0% cutting efficiency in all other PAMs). The PAM mismatch penalty matrices are continuously refined with experimentally determined cutting efficiencies for Cas9 variants, and the changes are reported in the manual. We calculate on-target sgRNA efficiency using the rule set 3 (RS3) score [50], also optimised for SpCas9 and enabled for all SpCas9 variants. The output of this step includes a final list of sgRNAs with off/on-target scores and genomic locations.

Predicted mutational outcomes are determined for each sgRNA–base editor pair, following user-defined base editor characteristics, i.e. the window of activity and base substitution inserted. For each sgRNA, the sequence within the base editor window of activity is retrieved, and all possible sites are mutated unless otherwise specified; Cs in GC motifs can remain unmutated if selected [10]. Mutated sequences are annotated using Ensembl’s VEP (version 112) [39]. Annotations are reported for all isoforms of the requested gene, highlighting Ensembl’s canonical isoform and the isoform retrieved using the VEP’s –pick option, enabling the users to identify the most relevant annotation per sgRNA. Results are output in an Excel format, directly compatible with the provided CRISPR-BEasy’s Oligo-assembly tool. For simplicity, Ensembl’s VEP annotations are aggregated in mutational outcomes: splice, nonsense, missense, synonymous, and noncoding (Supplementary Table S2).

The user first inputs the official gene names or genomic coordinates of their feature of interest, one per line, and selects the genome of the organism to be targeted from a drop-down menu. Genome assemblies for widely used model organisms are included in this release (); additional genome assemblies with annotations can be added upon user request. By default, CRISPR-BEasy will design sgRNA libraries to target all possible isoforms for a given gene. Alternatively, a feature is included that allows the design sgRNA libraries targeting canonical isoforms as determined by Ensembl. Custom flanking regions can be added to the genomic features of interest to ensure sgRNAs fully mutate the feature boundaries (e.g. splice sites). Supplementary Table S1

Then, the user selects the Cas variant from a drop-down menu, which includes all variants reportedly used in base editors to date. To enhance CRISPR-BEasy’s ability to integrate newly developed base editor architectures, base editor parameters are input in a text box, including the name, start and end position of the window of activity, and the base substitution inserted. Since APOBEC-1-based cytosine base editors disfavour editing in Cs in GC motifs [9], CRISPR-BEasy provides an option to exclude GC sites. Although base editor parameters have no direct impact on the design of the sgRNA library at this stage, they dictate the downstream annotation of predicted mutational outcomes per sgRNA.

Generating an sgRNA library free of potential off-target-induced artifacts can facilitate the interpretation of downstream BE screen results. When selecting SpCas9 variants, the user can customise the number of allowed off-targets per sgRNA, meeting a threshold for the CFD off-target score [1]. Although productive BE off-targets have additional requirements to those of Cas9—such as the presence of targetable DNA bases in the window of activity or the deleterious outcome of an inserted substitution—Cas9 off-target metrics are a good indication of promiscuous sgRNAs.

Finally, users can consider a set of genomic loci to act as positive and negative controls, selecting sgRNAs with predicted deleterious and neutral behaviour in their BE screen. For instance, in a viability screen, sgRNA inserting nonsense mutations in known essential genes and mutations in neutral genomic regions (e.g. the AAVS1 safe harbour locus) can act as positive and negative controls, respectively. Including high-confidence positive and negative controls is critical to evaluating the performance of BE screens. The sgRNA ranking for positive and negative controls within a BE screen can be used to build receiver operator characteristic (ROC) curves and calculate the area under the curve (AUC) to evaluate the screen’s sensitivity and specificity [15, 16]. Further, the distribution of phenotypic scores of negative controls can assist in determining thresholds of biological relevance [15]. Nonetheless, we marked these fields as optional, considering that the user’s target library may already contain the internal positive (e.g. sgRNA inserting nonsense/splice mutations) and negative controls (e.g. empty-window sgRNAs) required to conduct optimal BE screens. Finally, the modular structure of CRISPR-BEasy also enables the user to design a target library and positive and negative controls with different parameters. For instance, users may want more stringent off-target thresholds for negative controls, or positive controls targeting only highly efficient NGG-PAMs in an NG-PAM-targeting BE screen.

Once the sgRNA design is completed, CRISPR-BEasy provides downloadable Excel files for the target library (Target_Library.xlsx), and the positive and negative control libraries (Positive_Control_Library.xlsx, Negative_Control_Library.xlsx) if applicable. Each file contains a spreadsheet with basic information on the sgRNA library, including unique sgRNA identifiers and sequences, PAM sequences, details on their genomic position, and a column indicating the library type (i.e. target, positive, and negative). A spreadsheet with the variant annotation results per sgRNA from Ensembl’s VEP is added for each base editor specified in the input. These spreadsheets have been customised to facilitate downstream data analyses, indicating the canonical transcript, aggregating sgRNAs that insert the same variant, and indicating the number of base changes per sgRNA. Genomic loci acting as positive and negative controls follow the same processing as the target library in terms of sgRNA design and annotation. This workflow enables versatility in control selection across experiments and allows manual curation by the principal investigator before their integration into the assembled oligonucleotide library.

To verify the correct design of the library, CRISPR-BEasy also provides multiple auxiliary files with an overview of the sgRNA library composition. An interactive report includes a summary of the number of sgRNAs per gene and a per-base-editor breakdown of predicted mutational outcomes, i.e. splice, nonsense, missense, synonymous, and noncoding (Fig. 2). Users can click on each genomic feature to reveal a genome browser drop-down where a track for the designed sgRNA library marks sgRNA location and sgRNA coverage (Fig. 2). At a glance, users can check if all their features of interest (e.g. exons in all isoforms) are being targeted and to what extent.

Custom sgRNA libraries require synthesis as DNA oligonucleotides with suitable overhangs for downstream amplification and cloning.

The user uploads the Excel files for the target library obtained in Step I. If applicable, the user can upload the Excel files for positive and/or negative controls and indicate the desired number of sgRNAs to include in each category. Furthermore, the user can select editor-specific positive controls to create libraries compatible with different base editors. For instance, cytosine base editors can insert stop codons (nonsense) and splice-acceptor/donor-altering variants (splice), while adenosine base editors can cause start-loss codons (nonsense) and splice-acceptor/donor-altering variants (splice).

Now, the user provides the primer sequences for library amplification. If users do not have preferred primer sequences, the searchable CRISPR-BEasy’s manual lists primer pairs we have validated in past and ongoing BE screens ([15] and Manual). The selection of different primer pairs enables multiplexing of different sgRNA libraries within DNA oligonucleotide pools, reducing DNA synthesis costs.

Finally, users can select to assemble their sgRNA libraries as oligonucleotides for “traditional” BE screens [15, 16], or for sensor-based base editing screens with reporter-based monitoring of mutational outcomes and efficiencies [29]. For “traditional” BE screens, the user simply indicates the number of sgRNAs to assemble per DNA oligonucleotide to fit their DNA synthesis needs. The DNA oligonucleotides assembled are designed for cloning into sgRNA expression vectors using the type II restriction enzyme BsmBI. If opting for sensor-based BE screens, sgRNAs will be assembled in DNA oligonucleotides with an optimised sgRNA scaffold, the ∼40-nt sensor sequence corresponding to the sgRNA target site, and overhangs for cloning into sgRNA expression vectors using BsmBI and EcoRI [29].

The main output is a text file ready to be submitted for DNA oligonucleotide ordering. A second file (Prepared_library.csv) contains the sgRNAs included in the final library assembly, essential for downstream BE screen analyses. Of note, the number of sgRNAs in this file may differ slightly from the initial input target library file since duplicated sgRNAs and sgRNAs containing additional cloning restriction sites (BsmBI or BsmBI/EcoRI) are eliminated during the DNA oligonucleotide assembly. The generated DNA oligonucleotide library is ready to be cloned using the detailed protocol provided by CRISPR-BEasy in the searchable online manual [15, 53] or the protocol described for sensor-based libraries [29]. We generated four different sgRNA libraries of varying sizes using CRISPR-BEasy and cloned them with the protocol provided (Fig. 3A). Quality control analyses revealed Gaussian-like distributions for all libraries, with a low percentage of sgRNA loss when sequencing with coverage >200× (Fig. 3A and B). Since libraries were designed as 3-sgRNA oligonucleotides, we tested potential positional biases in sgRNA library amplification. A bias was observed toward increased representation of sgRNAs placed in later positions (3>2>1) in early cloned libraries (Library 1, Fig. 3C). This bias was diminished in libraries cloned with small protocol adjustments (Library 4, Fig. 3C, see “Materials and methods” section): We encourage a BsmBI pre-digestion step for >2-sgRNA-containing oligomers, as reflected in the manual’s protocol.