BEscreen: a flexible tool for creating base editing libraries

BEscreen offers three different modes to design base editing guides for pre-defined variants, for coding sequences (CDS) of specific genes and transcripts, and for given genomic regions. Input can be provided manually or as a CSV file. An overview of the workflow using BEscreen is outlined in Fig. 1.

BEscreen offers extensive customization options for all three modes. Notably, BEscreen enables full customization of PAM site requirements and base editor characteristics, including adjusting the length of the desired guides and the position of the editing window. BEs are expected to exhibit high editing efficiency in their individually determined editing window, but edits may occur outside of this window, albeit at lower efficiency [20]. To account for this, users can define a safety region to annotate guides with possible off-target edits in regions of minor editing efficiency adjacent to the main editing window. For convenience, all parameters can be set using base editor presets [18, 20, 21, 23, 26].

In addition to customization options for input and base editor characteristics, BEscreen enables tailoring of the output. Importantly, as a part of this, BEscreen provides functionality to identify guides with given biological consequences. This feature enables convenient identification of screen controls, such as synonymously editing guides as negative controls and splice site disrupting or nonsense variant-mediating guides as positive controls.

The BEscreen web server (bescreen.ostendorflab.org↗) is a Shiny for Python-based web app (v1.2.1; Python v3.12.3). The main third-party modules used are Polars (v1.16.0; github.com/pola-rs/polars↗) for data processing, pyfaidx (v0.8.1.3; github.com/mdshw5/pyfaidx↗) to extract genomic sequences from a FASTA reference genome, and PyRanges (v0.1.2; [27]) to parse a GTF annotation file. Optional annotation is based on Ensembl’s Variant Effect Predictor (VEP; v113.3; [28]) and NCBI’s BLAST (v2.16.0; [29]). igv.js (v3.1.3; [30]) is used with BAM files created by pysam (v0.22.1; github.com/pysam-developers/pysam↗) to display the alignment of sgRNAs to the reference genome. BEscreen provides the Ensembl reference genomes for human (GRCh38, Release 112) and mouse (GRCm39, Release 112) and is running on Ubuntu (24.04.1 LTS) and Apache2 (version 2.4.58). A command-line version is available at github.com/ostendorflab/bescreen↗.

BEscreen's main output is a table providing detailed information of the identified guides, including the introduced variant(s), base change, guide sequence with and without PAM site, sequence of the editing window (and, if set, for the safety region), the genomic location, and the strand of guide alignment. In addition, the table details how many edits take place within the editing window, and the distance of these edits to the center of the editing window, where editing is most efficient. For the variant and gene modes, BEscreen annotates identified guides with the affected gene, transcript, and exon. In addition, BEscreen reports affected codons and their translation product pre- and post-editing with their positions. In the gene and region modes, output includes an additional table listing non-editing guides (sgRNAs predicted to introduce no edits).

Optionally, the output of BEscreen can be annotated using Ensembl’s VEP [28]. To identify additional binding sites of given guides, BEscreen can identify matches of the guide sequence in the reference genome using NCBI’s BLAST [29]. In addition to the tabular output, the BEscreen web server displays genomic alignment of the identified sgRNAs in an embedded version of the Integrative Genome Viewer (igv.js) [30].

To illustrate BEscreen's functionality in practice, we designed sgRNAs for variants in the APOE gene using the most commonly used ABE and CBE BEs. APOE variants are the largest monogenetic risk modifiers of Alzheimer's disease and have more recently been implicated in modulating additional phenotypes ranging from cardiovascular disease to tumor and infection immunity [31–35].

Three highly prevalent variants of APOE exist, termed APOE2, APOE3, and APOE4, which are defined by two SNPs. The first SNP (rs429358) is a T to C substitution on chromosome 19 at position 44908684 leading to a change of a TGC codon to CGC, resulting in a cysteine to arginine substitution (APOE^C130R). The second SNP (rs7412) corresponds to a C to T substitution at position 44908822 (APOE^R176C). The most common APOE3 variant is defined by the rs429358(T)/rs7412(C) genotype, while the APOE4 and APOE2 variants correspond to rs429358(C)/rs7412(C) and rs429358(T)/rs7412(T) genotypes, respectively. We used BEscreen's variant mode to identify guides that mediate the introduction of these mutations in the reference APOE3 background (Fig. 2A). BEscreen identified one guide suitable for the rs429358 edit. Notably, inspection of the guide's characteristics in the tabular output revealed that this guide introduces a second edit changing the upstream GTG codon to GCG, resulting in an additional missense mutation. For the rs7412 SNP, BEscreen identified two suitable guides mediating the desired APOE^R176C mutation. By entering variants at the protein instead of nucleotide level (i.e. APOE-C130R instead of rs429358), BEscreen searches for whether additional edits in any of the other two bases making up the affected codon achieve the intended protein consequences, which here was not the case.

The implications of other, less common APOE variants remain largely unexplored, rendering their systematic interrogation intriguing [36]. To design an sgRNA library introducing all possible edits into APOE, we next leveraged BEscreen's gene mode (Fig. 2B). Using “APOE” as input, BEscreen designed 1065 editing guides, 506 of which mediated only one edit and were thus annotated as specific (Fig. 2C). As detailed in BEscreen's output, the consequences mediated by all edits included 526 missense, 313 synonymous, 11 splice site, 48 nonsense, 2 start lost, and 2 stop lost edits (Fig. 2D). In addition, 402 binding but non-editing guides were identified. These non-editing guides and guides conferring synonymous edits are valuable as negative controls in a BE screen, while guides mediating splice site disruptions or nonsense mutations may serve as positive controls. Thus, BEscreen designed a complete library for evaluating known and unknown variants in the APOE gene with minimal and convenient input requirements.

Finally, we used the genomic region spanning the APOE gene as input for BEscreen's region mode (“19:44905791–44909393”), resulting in 3150 and 1247 editing and non-editing guides, respectively. Notably, these guides cover the entire gene region, including introns and 5′ and 3′ untranslated regions (UTRs) in contrast to the output of the gene mode.

The BEscreen web server is available at bescreen.ostendorflab.org↗. This website is free and open to all users and there is no login requirement. A command line tool for local installation is available at github.com/ostendorflab/bescreen↗ and at doi.org/10.6084/m9.figshare.28616528↗.