Sequence-specific prediction of the efficiencies of adenine and cytosine base editors

Programmed RNA editing with an evolved bacterial adenosine deaminase

Programmed RNA editing presents an attractive therapeutic strategy for genetic disease. In this study, we developed bacterial deaminase-enabled recoding of RNA (DECOR), which employs an evolved Escherichia coli transfer RNA adenosine deaminase, TadA8e, to deposit adenosine-to-inosine editing to CRISPR-specified sites in the human transcriptome. DECOR functions in a variety of cell types, including human lung fibroblasts, and delivers on-target activity similar to ADAR-overexpressing RNA-editing platforms with 88% lower off-target effects. High-fidelity DECOR further reduces off-target effects to basal level. We demonstrate the clinical potential of DECOR by targeting Van der Woude syndrome-causing interferon regulatory factor 6 (IRF6) insufficiency. DECOR-mediated RNA editing removes a pathogenic upstream open reading frame (uORF) from the 5' untranslated region of IRF6 and rescues primary ORF expression from 12.3% to 36.5%, relative to healthy transcripts. DECOR expands the current portfolio of effector proteins and opens new territory in programmed RNA editing.

igRNA Prediction and Selection AI Models (igRNA-PS) for Bystander-less ABE Base Editing

CRISPR derived base editing techniques tend to edit multiple bases in the targeted region, which impedes precise reversion of disease-associated single nucleotide variations (SNVs). We designed an imperfect gRNA (igRNA) editing strategy to achieve bystander-less single-base editing. To predict the performance and provide ready-to-use igRNAs, we employed a high-throughput method to edit 5000 loci, each with approximate 19 systematically designed ABE igRNAs. Through deep learning of the relationship of editing efficiency, original gRNA sequence and igRNA sequence, AI models were constructed and tested, designated igRNA Prediction and Selection AI models (igRNA-PS). The models have three functions, First, they can identify the major editing site from the bystanders on a gRNA protospacer with a near 90% accuracy. second, a modified single-base editing efficiency (SBE), considering both single-base editing efficiency and product purity, can be predicted for any given igRNAs. Third, for an editing locus, a set of 64 igRNAs derived from a gRNA can be generated, evaluated through igRNA-PS to select for the best performer, and provided to the user. In this work, we overcome one of the most significant obstacles of base editors, and provide a convenient and efficient approach for single-base bystander-less ABE base editing.

An adenine base editor variant expands context compatibility

Adenine base editors (ABEs) are precise gene-editing agents that convert A:T pairs into G:C through a deoxyinosine intermediate. Existing ABEs function most effectively when the target A is in a TA context. Here we evolve the Escherichia coli transfer RNA-specific adenosine deaminase (TadA) to generate TadA8r, which extends potent deoxyadenosine deamination to RA (R = A or G) and is faster in processing GA than TadA8.20 and TadA8e, the two most active TadA variants reported so far. ABE8r, comprising TadA8r and a Streptococcus pyogenes Cas9 nickase, expands the editing window at the protospacer adjacent motif-distal end and outperforms ABE7.10, ABE8.20 and ABE8e in correcting disease-associated G:C-to-A:T transitions in the human genome, with a controlled off-target profile. We show ABE8r-mediated editing of clinically relevant sites that are poorly accessed by existing editors, including sites in PCSK9, whose disruption reduces low-density lipoprotein cholesterol, and ABCA4-p.Gly1961Glu, the most frequent mutation in Stargardt disease.

Direct measurement of engineered cancer mutations and their transcriptional phenotypes in single cells

Genome sequencing studies have identified numerous cancer mutations across a wide spectrum of tumor types, but determining the phenotypic consequence of these mutations remains a challenge. Here, we developed a high-throughput, multiplexed single-cell technology called TISCC-seq to engineer predesignated mutations in cells using CRISPR base editors, directly delineate their genotype among individual cells and determine each mutation's transcriptional phenotype. Long-read sequencing of the target gene's transcript identifies the engineered mutations, and the transcriptome profile from the same set of cells is simultaneously analyzed by short-read sequencing. Through integration, we determine the mutations' genotype and expression phenotype at single-cell resolution. Using cell lines, we engineer and evaluate the impact of >100 TP53 mutations on gene expression. Based on the single-cell gene expression, we classify the mutations as having a functionally significant phenotype.

In vivo adenine base editing rescues adrenoleukodystrophy in a humanized mouse model

X-linked adrenoleukodystrophy (ALD), an inherited neurometabolic disorder caused by mutations in ABCD1, which encodes the peroxisomal ABC transporter, mainly affects the brain, spinal cord, adrenal glands, and testes. In ALD patients, very-long-chain fatty acids (VLCFAs) fail to enter the peroxisome and undergo subsequent β-oxidation, resulting in their accumulation in the body. It has not been tested whether in vivo base editing or prime editing can be harnessed to ameliorate ALD. We developed a humanized mouse model of ALD by inserting a human cDNA containing the pathogenic variant into the mouse Abcd1 locus. The humanized ALD model showed increased levels of VLCFAs. To correct the mutation, we tested both base editing and prime editing and found that base editing using ABE8e(V106W) could correct the mutation in patient-derived fibroblasts at an efficiency of 7.4%. Adeno-associated virus (AAV)-mediated systemic delivery of NG-ABE8e(V106W) enabled robust correction of the pathogenic variant in the mouse brain (correction efficiency: ∼5.5%), spinal cord (∼5.1%), and adrenal gland (∼2%), leading to a significant reduction in the plasma levels of C26:0/C22:0. This established humanized mouse model and the successful correction of the pathogenic variant using a base editor serve as a significant step toward treating human ALD disease.

Structure-guided discovery of highly efficient cytidine deaminases with sequence-context independence

The applicability of cytosine base editors is hindered by their dependence on sequence context and by off-target effects. Here, by using AlphaFold2 to predict the three-dimensional structure of 1,483 cytidine deaminases and by experimentally characterizing representative deaminases (selected from each structural cluster after categorizing them via partitional clustering), we report the discovery of a few deaminases with high editing efficiencies, diverse editing windows and increased ratios of on-target to off-target effects. Specifically, several deaminases induced C-to-T conversions with comparable efficiency at AC/TC/CC/GC sites, the deaminases could introduce stop codons in single-copy and multi-copy genes in mammalian cells without double-strand breaks, and some residue conversions at predicted DNA-interacting sites reduced off-target effects. Structure-based generative machine learning could be further leveraged to expand the applicability of base editors in gene therapies.

Multiplex CRISPR-Cas Genome Editing: Next-Generation Microbial Strain Engineering

Genome editing is a crucial technology for obtaining desired phenotypes in a variety of species, ranging from microbes to plants, animals, and humans. With the advent of CRISPR-Cas technology, it has become possible to edit the intended sequence by modifying the target recognition sequence in guide RNA (gRNA). By expressing multiple gRNAs simultaneously, it is possible to edit multiple targets at the same time, allowing for the simultaneous introduction of various functions into the cell. This can significantly reduce the time and cost of obtaining engineered microbial strains for specific traits. In this review, we investigate the resolution of multiplex genome editing and its application in engineering microorganisms, including bacteria and yeast. Furthermore, we examine how recent advancements in artificial intelligence technology could assist in microbial genome editing and engineering. Based on these insights, we present our perspectives on the future evolution and potential impact of multiplex genome editing technologies in the agriculture and food industry.

Expanding the flexibility of base editing for high-throughput genetic screens in bacteria

Genome-wide screens have become powerful tools for elucidating genotype-to-phenotype relationships in bacteria. Of the varying techniques to achieve knockout and knockdown, CRISPR base editors are emerging as promising options. However, the limited number of available, efficient target sites hampers their use for high-throughput screening. Here, we make multiple advances to enable flexible base editing as part of high-throughput genetic screening in bacteria. We first co-opt the Streptococcus canis Cas9 that exhibits more flexible protospacer-adjacent motif recognition than the traditional Streptococcus pyogenes Cas9. We then expand beyond introducing premature stop codons by mutating start codons. Next, we derive guide design rules by applying machine learning to an essentiality screen conducted in Escherichia coli. Finally, we rescue poorly edited sites by combining base editing with Cas9-induced cleavage of unedited cells, thereby enriching for intended edits. The efficiency of this dual system was validated through a conditional essentiality screen based on growth in minimal media. Overall, expanding the scope of genome-wide knockout screens with base editors could further facilitate the investigation of new gene functions and interactions in bacteria.

Recent advances in CRISPR-based functional genomics for the study of disease-associated genetic variants

Advances in sequencing technology have greatly increased our ability to gather genomic data, yet understanding the impact of genetic mutations, particularly variants of uncertain significance (VUSs), remains a challenge in precision medicine. The CRISPR‒Cas system has emerged as a pivotal tool for genome engineering, enabling the precise incorporation of specific genetic variations, including VUSs, into DNA to facilitate their functional characterization. Additionally, the integration of CRISPR‒Cas technology with sequencing tools allows the high-throughput evaluation of mutations, transforming uncertain genetic data into actionable insights. This allows researchers to comprehensively study the functional consequences of point mutations, paving the way for enhanced understanding and increasing application to precision medicine. This review summarizes the current genome editing tools utilizing CRISPR‒Cas systems and their combination with sequencing tools for functional genomics, with a focus on point mutations.

Deep learning models to predict the editing efficiencies and outcomes of diverse base editors

Applications of base editing are frequently restricted by the requirement for a protospacer adjacent motif (PAM), and selecting the optimal base editor (BE) and single-guide RNA pair (sgRNA) for a given target can be difficult. To select for BEs and sgRNAs without extensive experimental work, we systematically compared the editing windows, outcomes and preferred motifs for seven BEs, including two cytosine BEs, two adenine BEs and three C•G to G•C BEs at thousands of target sequences. We also evaluated nine Cas9 variants that recognize different PAM sequences and developed a deep learning model, DeepCas9variants, for predicting which variants function most efficiently at sites with a given target sequence. We then develop a computational model, DeepBE, that predicts editing efficiencies and outcomes of 63 BEs that were generated by incorporating nine Cas9 variants as nickase domains into the seven BE variants. The predicted median efficiencies of BEs with DeepBE-based design were 2.9- to 20-fold higher than those of rationally designed SpCas9-containing BEs.

Deep learning models incorporating endogenous factors beyond DNA sequences improve the prediction accuracy of base editing outcomes

Adenine base editors (ABEs) and cytosine base editors (CBEs) enable the single nucleotide editing of targeted DNA sites avoiding generation of double strand breaks, however, the genomic features that influence the outcomes of base editing in vivo still remain to be characterized. High-throughput datasets from lentiviral integrated libraries were used to investigate the sequence features affecting base editing outcomes, but the effects of endogenous factors beyond the DNA sequences are still largely unknown. Here the base editing outcomes of ABE and CBE were evaluated in mammalian cells for 5012 endogenous genomic sites and 11,868 genome-integrated target sequences, with 4654 genomic sites sharing the same target sequences. The comparative analyses revealed that the editing outcomes of ABE and CBE at endogenous sites were substantially different from those obtained using genome-integrated sequences. We found that the base editing efficiency at endogenous target sites of both ABE and CBE was influenced by endogenous factors, including epigenetic modifications and transcriptional activity. A deep-learning algorithm referred as BE_Endo, was developed based on the endogenous factors and sequence information from our genomic datasets, and it yielded unprecedented accuracy in predicting the base editing outcomes. These findings along with the developed computational algorithms may facilitate future application of BEs for scientific research and clinical gene therapy.

Adenine base editor-based correction of the cardiac pathogenic Lmna c.1621C > T mutation in murine hearts

Base editors are emerging as powerful tools to correct single-nucleotide variants and treat genetic diseases. In particular, the adenine base editors (ABEs) exhibit robust and accurate adenine-to-guanidine editing capacity and have entered the clinical stage for cardiovascular therapy. Despite the tremendous progress using ABEs to treat heart diseases, a standard technical route toward successful ABE-based therapy remains to be fully established. In this study, we harnessed adeno-associated virus (AAV) and a mouse model carrying the cardiomyopathy-causing Lmna c.1621C > T mutation to demonstrate key steps and concerns in designing a cardiac ABE experiment in vivo. We found DeepABE as a reliable deep-learning-based model to predict ABE editing outcomes in the heart. Screening of sgRNAs for a Cas9 mutant with relieved protospacer adjacent motif (PAM) allowed the reduction of bystander editing. The ABE editing efficiency can be significantly enhanced by modifying the TadA and Cas9 variants, which are core components of ABEs. The ABE systems can be delivered into the heart via either dual AAV or all-in-one AAV vectors. Together, this study showcased crucial technical considerations in designing an ABE system for the heart and pointed out major challenges in further improvement of this new technology for gene therapy.

Advancing genome editing with artificial intelligence: opportunities, challenges, and future directions

Clustered regularly interspaced short palindromic repeat (CRISPR)-based genome editing (GED) technologies have unlocked exciting possibilities for understanding genes and improving medical treatments. On the other hand, Artificial intelligence (AI) helps genome editing achieve more precision, efficiency, and affordability in tackling various diseases, like Sickle cell anemia or Thalassemia. AI models have been in use for designing guide RNAs (gRNAs) for CRISPR-Cas systems. Tools like DeepCRISPR, CRISTA, and DeepHF have the capability to predict optimal guide RNAs (gRNAs) for a specified target sequence. These predictions take into account multiple factors, including genomic context, Cas protein type, desired mutation type, on-target/off-target scores, potential off-target sites, and the potential impacts of genome editing on gene function and cell phenotype. These models aid in optimizing different genome editing technologies, such as base, prime, and epigenome editing, which are advanced techniques to introduce precise and programmable changes to DNA sequences without relying on the homology-directed repair pathway or donor DNA templates. Furthermore, AI, in collaboration with genome editing and precision medicine, enables personalized treatments based on genetic profiles. AI analyzes patients' genomic data to identify mutations, variations, and biomarkers associated with different diseases like Cancer, Diabetes, Alzheimer's, etc. However, several challenges persist, including high costs, off-target editing, suitable delivery methods for CRISPR cargoes, improving editing efficiency, and ensuring safety in clinical applications. This review explores AI's contribution to improving CRISPR-based genome editing technologies and addresses existing challenges. It also discusses potential areas for future research in AI-driven CRISPR-based genome editing technologies. The integration of AI and genome editing opens up new possibilities for genetics, biomedicine, and healthcare, with significant implications for human health.

CRISPR/Cas9-gene editing approaches in plant breeding

CRISPR/Cas9 gene editing system is recently developed robust genome editing technology for accelerating plant breeding. Various modifications of this editing system have been established for adaptability in plant varieties as well as for its improved efficiency and portability. This review provides an in-depth look at the various strategies for synthesizing gRNAs for efficient delivery in plant cells, including chemical synthesis and in vitro transcription. It also covers traditional analytical tools and emerging developments in detection methods to analyze CRISPR/Cas9 mediated mutation in plant breeding. Additionally, the review outlines the various analytical tools which are used to detect and analyze CRISPR/Cas9 mediated mutations, such as next-generation sequencing, restriction enzyme analysis, and southern blotting. Finally, the review discusses emerging detection methods, including digital PCR and qPCR. Hence, CRISPR/Cas9 has great potential for transforming agriculture and opening avenues for new advancements in the system for gene editing in plants.

Systematic Exploration of Optimized Base Editing gRNA Design and Pleiotropic Effects with BExplorer

Base editing technology is being increasingly applied in genome engineering, but the current strategy for designing guide RNAs (gRNAs) relies substantially on empirical experience rather than a dependable and efficient in silico design. Furthermore, the pleiotropic effect of base editing on disease treatment remains unexplored, which prevents its further clinical usage. Here, we presented BExplorer, an integrated and comprehensive computational pipeline to optimize the design of gRNAs for 26 existing types of base editors in silico. Using BExplorer, we described its results for two types of mainstream base editors, BE3 and ABE7.10, and evaluated the pleiotropic effects of the corresponding base editing loci. BExplorer revealed 524 and 900 editable pathogenic single nucleotide polymorphism (SNP) loci in the human genome together with the selected optimized gRNAs for BE3 and ABE7.10, respectively. In addition, the impact of 707 edited pathogenic SNP loci following base editing on 131 diseases was systematically explored by revealing their pleiotropic effects, indicating that base editing should be carefully utilized given the potential pleiotropic effects. Collectively, the systematic exploration of optimized base editing gRNA design and the corresponding pleiotropic effects with BExplorer provides a computational basis for applying base editing in disease treatment.

Engineering of cytosine base editors with DNA damage minimization and editing scope diversification

Cytosine base editors (CBEs), which enable precise C-to-T substitutions, have been restricted by potential safety risks, including DNA off-target edits, RNA off-target edits and additional genotoxicity such as DNA damages induced by double-strand breaks (DSBs). Though DNA and RNA off-target edits have been ameliorated via various strategies, evaluation and minimization of DSB-associated DNA damage risks for most CBEs remain to be resolved. Here we demonstrate that YE1, an engineered CBE variant with minimized DNA and RNA off-target edits, could induce prominent DSB-associated DNA damage risks, manifested as γH2AX accumulation in human cells. We then perform deaminase engineering for two deaminases lamprey LjCDA1 and human APOBEC3A, and generate divergent CBE variants with eliminated DSB-associated DNA damage risks, in addition to minimized DNA/RNA off-target edits. Furthermore, the editing scopes and sequence preferences of APOBEC3A-derived CBEs could be further diversified by internal fusion strategy. Taken together, this study provides updated evaluation platform for DSB-associated DNA damage risks of CBEs and further generates a series of safer toolkits with diversified editing signatures to expand their applications.

Prediction of prime editing insertion efficiencies using sequence features and DNA repair determinants

Most short sequences can be precisely written into a selected genomic target using prime editing; however, it remains unclear what factors govern insertion. We design a library of 3,604 sequences of various lengths and measure the frequency of their insertion into four genomic sites in three human cell lines, using different prime editor systems in varying DNA repair contexts. We find that length, nucleotide composition and secondary structure of the insertion sequence all affect insertion rates. We also discover that the 3' flap nucleases TREX1 and TREX2 suppress the insertion of longer sequences. Combining the sequence and repair features into a machine learning model, we can predict relative frequency of insertions into a site with R = 0.70. Finally, we demonstrate how our accurate prediction and user-friendly software help choose codon variants of common fusion tags that insert at high efficiency, and provide a catalog of empirically determined insertion rates for over a hundred useful sequences.

Prediction of base editor off-targets by deep learning

Due to the tolerance of mismatches between gRNA and targeting sequence, base editors frequently induce unwanted Cas9-dependent off-target mutations. Here, to develop models to predict such off-targets, we design gRNA-off- target pairs for adenine base editors (ABEs) and cytosine base editors (CBEs) and stably integrate them into the human cells. After five days of editing, we obtain valid efficiency datasets of 54,663 and 55,727 off-targets for ABEs and CBEs, respectively. We use the datasets to train deep learning models, resulting in ABEdeepoff and CBEdeepoff, which can predict off-target sites. We use these tools to predict off-targets for a panel of endogenous loci and achieve Spearman correlation values varying from 0.710 to 0.859. Finally, we develop an integrated tool that is freely accessible via an online web server http://www.deephf.com/#/bedeep/bedeepoff . These tools could facilitate minimizing the off-target effects of base editing.

Predicting Mutations Generated by Cas9, Base Editing, and Prime Editing in Mammalian Cells

The first fruits of the CRISPR-Cas revolution are starting to enter the clinic, with gene editing therapies offering solutions to previously incurable genetic diseases. The success of such applications hinges on control over the mutations that are generated, which are known to vary depending on the targeted locus. In this review, we present the current state of understanding and predicting CRISPR-Cas cutting, base editing, and prime editing outcomes in mammalian cells. We first provide an introduction to the basics of DNA repair and machine learning that the models rely on. We then overview the datasets and methods created for characterizing edits at scale, as well as the insights that have been derived from them. The predictions generated from these models serve as a foundation for designing efficient experiments across the broad contexts where these tools are applied.

Base editor screens for in situ mutational scanning at scale

A fundamental challenge in biology is understanding the molecular details of protein function. How mutations alter protein activity, regulation, and response to drugs is of critical importance to human health. Recent years have seen the emergence of pooled base editor screens for in situ mutational scanning: the interrogation of protein sequence-function relationships by directly perturbing endogenous proteins in live cells. These studies have revealed the effects of disease-associated mutations, discovered novel drug resistance mechanisms, and generated biochemical insights into protein function. Here, we discuss how this "base editor scanning" approach has been applied to diverse biological questions, compare it with alternative techniques, and describe the emerging challenges that must be addressed to maximize its utility. Given its broad applicability toward profiling mutations across the proteome, base editor scanning promises to revolutionize the investigation of proteins in their native contexts.

Cytosine base editors (CBEs) for inducing targeted DNA base editing in Nicotiana benthamiana

BACKGROUND

The base editors can introduce point mutations accurately without causing double-stranded DNA breaks or requiring donor DNA templates. Previously, cytosine base editors (CBEs) containing different deaminases are reported for precise and accurate base editing in plants. However, the knowledge of CBEs in polyploid plants is inadequate and needs further exploration.

RESULTS

In the present study, we constructed three polycistronic tRNA-gRNA expression cassettes CBEs containing A3A, A3A (Y130F), and rAPOBEC1(R33A) to compare their base editing efficiency in allotetraploid N. benthamiana (n = 4x). We used 14 target sites to compare their editing efficiency using transient transformation in tobacco plants. The sanger sequencing and deep sequencing results showed that A3A-CBE was the most efficient base editor. In addition, the results showed that A3A-CBE provided most comprehensive editing window (C₁ ~ C₁₇ could be edited) and had a better editing efficiency under the base background of TC. The target sites (T2 and T6) analysis in transformed N. benthamiana showed that only A3A-CBE can have C-to-T editing events and the editing efficiency of T2 was higher than T6. Additionally, no off-target events were found in transformed N. benthamiana.

CONCLUSIONS

All in all, we conclude that A3A-CBE is the most suitable vector for specific C to T conversion in N. benthamiana. Current findings will provide valuable insights into selecting an appropriate base editor for breeding polyploid plants.

Massively parallel evaluation and computational prediction of the activities and specificities of 17 small Cas9s

Recently, various small Cas9 orthologs and variants have been reported for use in in vivo delivery applications. Although small Cas9s are particularly suited for this purpose, selecting the most optimal small Cas9 for use at a specific target sequence continues to be challenging. Here, to this end, we have systematically compared the activities of 17 small Cas9s for thousands of target sequences. For each small Cas9, we have characterized the protospacer adjacent motif and determined optimal single guide RNA expression formats and scaffold sequence. High-throughput comparative analyses revealed distinct high- and low-activity groups of small Cas9s. We also developed DeepSmallCas9, a set of computational models predicting the activities of the small Cas9s at matched and mismatched target sequences. Together, this analysis and these computational models provide a useful guide for researchers to select the most suitable small Cas9 for specific applications.

Vision rescue via unconstrained in vivo prime editing in degenerating neural retinas

Retinitis pigmentosa (RP) is an inherited retinal dystrophy causing progressive and irreversible loss of retinal photoreceptors. Here, we developed a genome-editing tool characterized by the versatility of prime editors (PEs) and unconstrained PAM requirement of a SpCas9 variant (SpRY), referred to as PESpRY. The diseased retinas of Pde6b-associated RP mouse model were transduced via a dual AAV system packaging PESpRY for the in vivo genome editing through a non-NGG PAM (GTG). The progressing cell loss was reversed once the mutation was corrected, leading to substantial rescue of photoreceptors and production of functional PDE6β. The treated mice exhibited significant responses in electroretinogram and displayed good performance in both passive and active avoidance tests. Moreover, they presented an apparent improvement in visual stimuli-driven optomotor responses and efficiently completed visually guided water-maze tasks. Together, our study provides convincing evidence for the prevention of vision loss caused by RP-associated gene mutations via unconstrained in vivo prime editing in the degenerating retinas.

HMGN1 enhances CRISPR-directed dual-function A-to-G and C-to-G base editing

C-to-G base editors have been successfully constructed recently, but limited work has been done on concurrent C-to-G and A-to-G base editing. In addition, there is also limited data on how chromatin-associated factors affect the base editing. Here, we test a series of chromatin-associated factors, and chromosomal protein HMGN1 was found to enhance the efficiency of both C-to-G and A-to-G base editing. By fusing HMGN1, GBE and ABE to Cas9, we develop a CRISPR-based dual-function A-to-G and C-to-G base editor (GGBE) which is capable of converting simultaneous A and C to G conversion with substantial editing efficiency. Accordingly, the HMGN1 role shown in this work and the resulting GGBE tool further broaden the genome manipulation capacity of CRISPR-directed base editors.

Prediction of efficiencies for diverse prime editing systems in multiple cell types

Applications of prime editing are often limited due to insufficient efficiencies, and it can require substantial time and resources to determine the most efficient pegRNAs and prime editors (PEs) to generate a desired edit under various experimental conditions. Here, we evaluated prime editing efficiencies for a total of 338,996 pairs of pegRNAs including 3,979 epegRNAs and target sequences in an error-free manner. These datasets enabled a systematic determination of factors affecting prime editing efficiencies. Then, we developed computational models, named DeepPrime and DeepPrime-FT, that can predict prime editing efficiencies for eight prime editing systems in seven cell types for all possible types of editing of up to 3 base pairs. We also extensively profiled the prime editing efficiencies at mismatched targets and developed a computational model predicting editing efficiencies at such targets. These computational models, together with our improved knowledge about prime editing efficiency determinants, will greatly facilitate prime editing applications.

PAM-flexible Cas9-mediated base editing of a hemophilia B mutation in induced pluripotent stem cells

BACKGROUND

Base editing via CRISPR-Cas9 has garnered attention as a method for correcting disease-specific mutations without causing double-strand breaks, thereby avoiding large deletions and translocations in the host chromosome. However, its reliance on the protospacer adjacent motif (PAM) can limit its use. We aimed to restore a disease mutation in a patient with severe hemophilia B using base editing with SpCas9-NG, a modified Cas9 with the board PAM flexibility.

METHODS

We generated induced pluripotent stem cells (iPSCs) from a patient with hemophilia B (c.947T>C; I316T) and established HEK293 cells and knock-in mice expressing the patient's F9 cDNA. We transduced the cytidine base editor (C>T), including the nickase version of Cas9 (wild-type SpCas9 or SpCas9-NG), into the HEK293 cells and knock-in mice through plasmid transfection and an adeno-associated virus vector, respectively.

RESULTS

Here we demonstrate the broad PAM flexibility of SpCas9-NG near the mutation site. The base-editing approach using SpCas9-NG but not wild-type SpCas9 successfully converts C to T at the mutation in the iPSCs. Gene-corrected iPSCs differentiate into hepatocyte-like cells in vitro and express substantial levels of F9 mRNA after subrenal capsule transplantation into immunodeficient mice. Additionally, SpCas9-NG-mediated base editing corrects the mutation in both HEK293 cells and knock-in mice, thereby restoring the production of the coagulation factor.

CONCLUSION

A base-editing approach utilizing the broad PAM flexibility of SpCas9-NG can provide a solution for the treatment of genetic diseases, including hemophilia B.

Highly Efficient A-to-G Editing in PFFs via Multiple ABEs

Cytosine base editors (CBEs) and adenine base editors (ABEs) are recently developed CRISPR-mediated genome-editing tools that do not introduce double-strand breaks. In this study, five ABEs, ABE7.10, ABEmax, NG-ABEmax, ABE8e and NG-ABE8e, were used to generate A-to-G (T-to-C) conversions in five genome loci in porcine fetal fibroblasts (PFFs). Variable yet appreciable editing efficiencies and variable activity windows were observed in these targeting regions via these five editors. The strategy of two sgRNAs in one vector exhibited superior editing efficiency to that of using two separate sgRNA expression vectors. ABE-mediated start-codon mutation in APOE silenced its expression of protein and, unexpectedly, eliminated the vast majority of its mRNA. No off-target DNA site was detected for these editors. Substantial off-target RNA events were present in the ABE-edited cells, but no KEGG pathway was found to be significantly enriched. Our study supports that ABEs are powerful tools for A-to-G (T-to-C) point-mutation modification in porcine cells.

Obtaining the best igRNAs for bystander-less correction of all ABE-reversible pathogenic SNVs using high-throughput screening

Imperfect -gRNA (igRNA) provides a simple strategy for single-base editing of a base editor. However, a significant number of igRNAs need to be generated and tested for each target locus to achieve efficient single-base reversion of pathogenic single nucleotide variations (SNVs), which hinders the direct application of this technology. To provide ready-to-use igRNAs for single-base and bystander-less correction of all the adenine base editor (ABE)-reversible pathogenic SNVs, we employed a high-throughput method to edit all 5,253 known ABE-reversible pathogenic SNVs, each with multiple systematically designed igRNAs, and two libraries of 96,000 igRNAs were tested. A total of 1,988 SNV loci could be single-base reversed by igRNA with a >30% efficiency. Among these 1,988 loci, 378 SNV loci exhibited an efficiency of more than 90%. At the same time, the bystander editing efficiency of 76.62% of the SNV loci was reduced to 0%, while remaining below 1% for another 18.93% of the loci. These ready-to-use igRNAs provided the best solutions for a substantial portion of the 4,657 pathogenic/likely pathogenic SNVs. In this work, we overcame one of the most significant obstacles of base editors and provide a ready-to-use platform for the genetic treatment of diseases caused by ABE-reversible SNVs.

Targeted dual base editing with Campylobacter jejuni Cas9 by single AAV-mediated delivery

Various CRISPR‒Cas9 orthologs are used in genome engineering. One of the smallest Cas9 orthologs is cjCas9 derived from Campylobacter jejuni, which is a highly specific genome editing tool. Here, we developed cjCas9-based base editors including a cytosine base editor (cjCBEmax) and an adenine base editor (cjABE8e) that can successfully induce endogenous base substitutions by up to 91.2% at the HPD gene in HEK293T cells. Analysis of the base editing efficiency of 13 endogenous target sites showed that the active windows of cjCBEmax and cjABE8e are wider than those of spCas9-based base editors and that their specificities are slightly lower than that of cjCas9. Importantly, engineered cjCas9 and gRNA scaffolds can improve the base editing efficiency of cjABE8e by up to 6.4-fold at the HIF1A gene in HEK293T cells. Due to its small size, cjABE8e can be packaged in a single adeno-associated virus vector with two tandem arrays of gRNAs, and the delivery of the resulting AAV could introduce base substitutions at endogenous ANGPT2 and HPD target sites. Overall, our findings have expanded the potential of the use of base editors for in vivo or ex vivo therapeutic approaches.

TadA orthologs enable both cytosine and adenine editing of base editors

Cytidine and adenosine deaminases are required for cytosine and adenine editing of base editors respectively, and no single deaminase could enable concurrent and comparable cytosine and adenine editing. Additionally, distinct properties of cytidine and adenosine deaminases lead to various types of off-target effects, including Cas9-indendepent DNA off-target effects for cytosine base editors (CBEs) and RNA off-target effects particularly severe for adenine base editors (ABEs). Here we demonstrate that 25 TadA orthologs could be engineered to generate functional ABEs, CBEs or ACBEs via single or double mutations, which display minimized Cas9-independent DNA off-target effects and genotoxicity, with orthologs B5ZCW4, Q57LE3, E8WVH3, Q13XZ4 and B3PCY2 as promising candidates for further engineering. Furthermore, RNA off-target effects of TadA ortholog-derived base editors could be further reduced or even eliminated by additional single mutation. Taken together, our work expands the base editing toolkits, and also provides important clues for the potential evolutionary process of deaminases.

Prediction of Base Editing Efficiencies and Outcomes Using DeepABE and DeepCBE

Adenine base editors (ABEs) and cytosine base editors (CBEs) have been widely used to introduce disease-relevant point mutations at target DNA sites of interest. However, the introduction of point mutations using base editors can be difficult due to low editing efficiencies and/or the existence of multiple target nucleotides within the base editing window at the target site. Thus, previous works have relied heavily on experimentally evaluating the base editing efficiencies and outcomes using time-consuming and labor-intensive multi-step experimental processes. DeepABE and DeepCBE are deep learning-based computational models to predict the efficiencies and outcome frequencies of ABE and CBE at given target DNA sites, in silico. Here, we describe the step-by-step procedure for the accurate determination of specific target nucleotides for ABE or CBE editing on the online available web tool, (DeepBaseEditor, https://deepcrispr.info/DeepBaseEditor ).

Web-Based Computational Tools for Base Editors

CRISPR-based base editors are efficient genome editing tools for use in base correction. Currently, there are various versions and types of base editors with different substitution patterns, editing windows, and protospacer adjacent motif (PAM) sequences. For the design of target sequences, consideration of off-target sequences is required. In addition, for assessment of base editing outcomes in bulk populations, the analysis of high-throughput sequencing data is required. Several web browser-based computation programs have been developed for the purpose of target design and NGS data analysis, especially for users with less computational knowledge. In this manuscript, depending on the purpose of each program, we provide an explanation of useful tools including BE-Designer for design of targets and BE-Analyzer for analysis of NGS data that were developed by our group, CRISPResso2 for analysis of NGS data developed by Luca Pinello group, DeepBaseEditor for prediction of target efficiency developed by Hyongbum Henry Kim group, and BE-Hive for prediction of target outcome developed by David Liu group.

Designing Guide-RNA for Generating Premature Stop Codons for Gene Knockout Using CRISPR-BETS

Cytosine base editors (CBEs) accurately modify target sites by mediating a C to T change (or a G to A change on the opposite strand). This allows us to install premature stop codons for gene knockout. However, highly specific sgRNAs (single-guide RNAs) are necessary for the CRISPR-Cas nuclease to work efficiently. In this study, we introduce a method of designing highly specific gRNA to generate premature stop codons and knock out a gene using CRISPR-BETS software.

Automated high-throughput genome editing platform with an AI learning in situ prediction model

A great number of cell disease models with pathogenic SNVs are needed for the development of genome editing based therapeutics or broadly basic scientific research. However, the generation of traditional cell disease models is heavily dependent on large-scale manual operations, which is not only time-consuming, but also costly and error-prone. In this study, we devise an automated high-throughput platform, through which thousands of samples are automatically edited within a week, providing edited cells with high efficiency. Based on the large in situ genome editing data obtained by the automatic high-throughput platform, we develop a Chromatin Accessibility Enabled Learning Model (CAELM) to predict the performance of cytosine base editors (CBEs), both chromatin accessibility and the context-sequence are utilized to build the model, which accurately predicts the result of in situ base editing. This work is expected to accelerate the development of BE-based genetic therapies.

ABE8e adenine base editor precisely and efficiently corrects a recurrent COL7A1 nonsense mutation

Base editing introduces precise single-nucleotide edits in genomic DNA and has the potential to treat genetic diseases such as the blistering skin disease recessive dystrophic epidermolysis bullosa (RDEB), which is characterized by mutations in the COL7A1 gene and type VII collagen (C7) deficiency. Adenine base editors (ABEs) convert A-T base pairs to G-C base pairs without requiring double-stranded DNA breaks or donor DNA templates. Here, we use ABE8e, a recently evolved ABE, to correct primary RDEB patient fibroblasts harboring the recurrent RDEB nonsense mutation c.5047 C > T (p.Arg1683Ter) in exon 54 of COL7A1 and use a next generation sequencing workflow to interrogate post-treatment outcomes. Electroporation of ABE8e mRNA into a bulk population of RDEB patient fibroblasts resulted in remarkably efficient (94.6%) correction of the pathogenic allele, restoring COL7A1 mRNA and expression of C7 protein in western blots and in 3D skin constructs. Off-target DNA analysis did not detect off-target editing in treated patient-derived fibroblasts and there was no detectable increase in A-to-I changes in the RNA. Taken together, we have established a highly efficient pipeline for gene correction in primary fibroblasts with a favorable safety profile. This work lays a foundation for developing therapies for RDEB patients using ex vivo or in vivo base editing strategies.

Annotation and evaluation of base editing outcomes in multiple cell types using CRISPRbase

CRISPR-Cas base editing (BE) system is a powerful tool to expand the scope and efficiency of genome editing with single-nucleotide resolution. The editing efficiency, product purity, and off-target effect differ among various BE systems. Herein, we developed CRISPRbase (http://crisprbase.maolab.org), by integrating 1 252 935 records of base editing outcomes in more than 50 cell types from 17 species. CRISPRbase helps to evaluate the putative editing precision of different BE systems by integrating multiple annotations, functional predictions and a blasting system for single-guide RNA sequences. We systematically assessed the editing window, editing efficiency and product purity of various BE systems. Intensive efforts were focused on increasing the editing efficiency and product purity of base editors since the byproduct could be detrimental in certain applications. Remarkably, more than half of cancer-related off-target mutations were non-synonymous and extremely damaging to protein functions in most common tumor types. Luckily, most of these cancer-related mutations were passenger mutations (4840/5703, 84.87%) rather than cancer driver mutations (863/5703, 15.13%), indicating a weak effect of off-target mutations on carcinogenesis. In summary, CRISPRbase is a powerful and convenient tool to study the outcomes of different base editors and help researchers choose appropriate BE designs for functional studies.

Genome-wide base editor screen identifies regulators of protein abundance in yeast

Proteins are key molecular players in a cell, and their abundance is extensively regulated not just at the level of gene expression but also post-transcriptionally. Here, we describe a genetic screen in yeast that enables systematic characterization of how protein abundance regulation is encoded in the genome. The screen combines a CRISPR/Cas9 base editor to introduce point mutations with fluorescent tagging of endogenous proteins to facilitate a flow-cytometric readout. We first benchmarked base editor performance in yeast with individual gRNAs as well as in positive and negative selection screens. We then examined the effects of 16,452 genetic perturbations on the abundance of eleven proteins representing a variety of cellular functions. We uncovered hundreds of regulatory relationships, including a novel link between the GAPDH isoenzymes Tdh1/2/3 and the Ras/PKA pathway. Many of the identified regulators are specific to one of the eleven proteins, but we also found genes that, upon perturbation, affected the abundance of most of the tested proteins. While the more specific regulators usually act transcriptionally, broad regulators often have roles in protein translation. Overall, our novel screening approach provides unprecedented insights into the components, scale and connectedness of the protein regulatory network.

In vivo application of base and prime editing to treat inherited retinal diseases

Inherited retinal diseases (IRDs) are vision-threatening retinal disorders caused by pathogenic variants of genes related to visual functions. Genomic analyses in patients with IRDs have revealed pathogenic variants which affect vision. However, treatment options for IRDs are limited to nutritional supplements regardless of genetic variants or gene-targeting approaches based on antisense oligonucleotides and adeno-associated virus vectors limited to targeting few genes. Genome editing, particularly that involving clustered regularly interspaced short palindromic repeat (CRISPR)-Cas9 technologies, can correct pathogenic variants and provide additional treatment opportunities. Recently developed base and prime editing platforms based on CRISPR-Cas9 technologies are promising for therapeutic genome editing because they do not employ double-stranded breaks (DSBs), which are associated with P53 activation, large deletions, and chromosomal translocations. Instead, using attached deaminases and reverse transcriptases, base and prime editing efficiently induces specific base substitutions and intended genetic changes (substitutions, deletions, or insertions), respectively, without DSBs. In this review, we will discuss the recent in vivo application of CRISPR-Cas9 technologies, focusing on base and prime editing, in animal models of IRDs.

BEtarget: a versatile web-based tool to design guide RNAs for base editing in plants

•

BEtarget supports the gRNA design of base editing with different types of PAM.

•

BEtarget provides an interactive and customized visualization interface.

•

BEtarget can automatically detect the coordinates of coding regions (exons) in the genomic sequence of the target gene.

CRISPR-dependent base editors enable direct nucleotide conversion without the introduction of double-strand DNA break or donor DNA template, thus expanding the CRISPR toolbox for genetic manipulation. However, designing guide RNAs (gRNAs) for base editors to enable gene correction or inactivation is more complicated than using the CRISPR system for gene disruption. Here, we present a user-friendly web tool named BEtarget dedicated to the design of gRNA for base editing. It is currently supported by 46 plant reference genomes and 5 genomes of non-plant model organisms. BEtarget supports the design of gRNAs with different types of protospacer adjacent motifs (PAM) and integrates various functions, including automatic identification of open reading frame, prediction of potential off-target sites, annotation of codon change, and assessment of gRNA quality. Moreover, the program provides an interactive interface for users to selectively display information about the desired target sites. In brief, we have developed a flexible and versatile web-based tool to simplify complications associated with the design of base editing technology. BEtarget is freely accessible at https://skl.scau.edu.cn/betarget/.

Cas-Based Systems for RNA Editing in Gene Therapy of Monogenic Diseases: In Vitro and in Vivo Application and Translational Potential

Rare genetic diseases reduce quality of life and can significantly shorten the lifespan. There are few effective treatment options for these diseases, and existing therapeutic strategies often represent only supportive or palliative care. Therefore, designing genetic-engineering technologies for the treatment of genetic diseases is urgently needed. Rapid advances in genetic editing technologies based on programmable nucleases and in the engineering of gene delivery systems have made it possible to conduct several dozen successful clinical trials; however, the risk of numerous side effects caused by off-target double-strand breaks limits the use of these technologies in the clinic. Development of adenine-to-inosine (A-to-I) and cytosine-to-uracil (C-to-U) RNA-editing systems based on dCas13 enables editing at the transcriptional level without double-strand breaks in DNA. In this review, we discuss recent progress in the application of these technologies in in vitro and in vivo experiments. The main strategies for improving RNA-editing tools by increasing their efficiency and specificity are described as well. These data allow us to outline the prospects of base-editing systems for clinical application.

High-throughput functional evaluation of human cancer-associated mutations using base editors

Comprehensive phenotypic characterization of the many mutations found in cancer tissues is one of the biggest challenges in cancer genomics. In this study, we evaluated the functional effects of 29,060 cancer-related transition mutations that result in protein variants on the survival and proliferation of non-tumorigenic lung cells using cytosine and adenine base editors and single guide RNA (sgRNA) libraries. By monitoring base editing efficiencies and outcomes using surrogate target sequences paired with sgRNA-encoding sequences on the lentiviral delivery construct, we identified sgRNAs that induced a single primary protein variant per sgRNA, enabling linking those mutations to the cellular phenotypes caused by base editing. The functions of the vast majority of the protein variants (28,458 variants, 98%) were classified as neutral or likely neutral; only 18 (0.06%) and 157 (0.5%) variants caused outgrowing and likely outgrowing phenotypes, respectively. We expect that our approach can be extended to more variants of unknown significance and other tumor types.

Base editing sensor libraries for high-throughput engineering and functional analysis of cancer-associated single nucleotide variants

Base editing can be applied to characterize single nucleotide variants of unknown function, yet defining effective combinations of single guide RNAs (sgRNAs) and base editors remains challenging. Here, we describe modular base-editing-activity 'sensors' that link sgRNAs and cognate target sites in cis and use them to systematically measure the editing efficiency and precision of thousands of sgRNAs paired with functionally distinct base editors. By quantifying sensor editing across >200,000 editor-sgRNA combinations, we provide a comprehensive resource of sgRNAs for introducing and interrogating cancer-associated single nucleotide variants in multiple model systems. We demonstrate that sensor-validated tools streamline production of in vivo cancer models and that integrating sensor modules in pooled sgRNA libraries can aid interpretation of high-throughput base editing screens. Using this approach, we identify several previously uncharacterized mutant TP53 alleles as drivers of cancer cell proliferation and in vivo tumor development. We anticipate that the framework described here will facilitate the functional interrogation of cancer variants in cell and animal models.

Current progress and open challenges for applying deep learning across the biosciences

Deep Learning (DL) has recently enabled unprecedented advances in one of the grand challenges in computational biology: the half-century-old problem of protein structure prediction. In this paper we discuss recent advances, limitations, and future perspectives of DL on five broad areas: protein structure prediction, protein function prediction, genome engineering, systems biology and data integration, and phylogenetic inference. We discuss each application area and cover the main bottlenecks of DL approaches, such as training data, problem scope, and the ability to leverage existing DL architectures in new contexts. To conclude, we provide a summary of the subject-specific and general challenges for DL across the biosciences.

Improvement of base editors and prime editors advances precision genome engineering in plants

Clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR associated protein (Cas)-mediated gene disruption has revolutionized biomedical research as well as plant and animal breeding. However, most disease-causing mutations and agronomically important genetic variations are single base polymorphisms (single-nucleotide polymorphisms) that require precision genome editing tools for correction of the sequences. Although homology-directed repair of double-stranded breaks (DSBs) can introduce precise changes, such repairs are inefficient in differentiated animal and plant cells. Base editing and prime editing are two recently developed genome engineering approaches that can efficiently introduce precise edits into target sites without requirement of DSB formation or donor DNA templates. They have been applied in several plant species with promising results. Here, we review the extensive literature on improving the efficiency, target scope, and specificity of base editors and prime editors in plants. We also highlight recent progress on base editing in plant organellar genomes and discuss how these precision genome editing tools are advancing basic plant research and crop breeding.

CRISPR-BETS: a base-editing design tool for generating stop codons

Cytosine base editors (CBEs) can install a predefined stop codon at the target site, representing a more predictable and neater method for creating genetic knockouts without altering the genome size. Due to the enhanced predictability of the editing outcomes, it is also more efficient to obtain homozygous mutants in the first generation. With the recent advancement of CBEs on improved editing activity, purify and specificity in plants and animals, base editing has become a more appealing technology for generating knockouts. However, there is a lack of design tools that can aid the adoption of CBEs for achieving such a purpose, especially in plants. Here, we developed a user-friendly design tool named CRISPR-BETS (base editing to stop), which helps with guide RNA (gRNA) design for introducing stop codons in the protein-coding genes of interest. We demonstrated in rice and tomato that CRISPR-BETS is easy-to-use, and its generated gRNAs are highly specific and efficient for generating stop codons and obtaining homozygous knockout lines. While we tailored the tool for the plant research community, CRISPR-BETS can also serve non-plant species.

The Use of Base Editing Technology to Characterize Single Nucleotide Variants

Single nucleotide variants (SNVs) represent the most common type of polymorphism in the human genome. However, in many cases the phenotypic impacts of such variants are not well understood. Intriguingly, while some SNVs cause debilitating diseases, other variants in the same gene may have no, or limited, impact. The mechanisms underlying these complex patterns are difficult to study at scale. Additionally, current data and research is mainly focused on European populations, and the mechanisms underlying genetic traits in other populations are poorly studied. Novel technologies may be able to mitigate this disparity and improve the applicability of personalized healthcare to underserved populations. In this review we discuss base editing technologies and their potential to accelerate progress in this field, particularly in combination with single-cell RNA sequencing. We further explore how base editing screens can help link SNVs to distinct disease phenotypes. We then highlight several studies that take advantage of single-cell RNA sequencing and CRISPR screens to emphasize the current limitations and future potential of this technique. Lastly, we consider the use of such approaches to potentially accelerate the study of genetic mechanisms in non-European populations.

Application of prime editing to the correction of mutations and phenotypes in adult mice with liver and eye diseases

The use of prime editing-a gene-editing technique that induces small genetic changes without the need for donor DNA and without causing double strand breaks-to correct pathogenic mutations and phenotypes needs to be tested in animal models of human genetic diseases. Here we report the use of prime editors 2 and 3, delivered by hydrodynamic injection, in mice with the genetic liver disease hereditary tyrosinemia, and of prime editor 2, delivered by an adeno-associated virus vector, in mice with the genetic eye disease Leber congenital amaurosis. For each pathogenic mutation, we identified an optimal prime-editing guide RNA by using cells transduced with lentiviral libraries of guide-RNA-encoding sequences paired with the corresponding target sequences. The prime editors precisely corrected the disease-causing mutations and led to the amelioration of the disease phenotypes in the mice, without detectable off-target edits. Prime editing should be tested further in more animal models of genetic diseases.

A systematic mapping study on machine learning techniques for the prediction of CRISPR/Cas9 sgRNA target cleavage

CRISPR/Cas9 technology has greatly accelerated genome engineering research. The CRISPR/Cas9 complex, a bacterial immune response system, is widely adopted for RNA-driven targeted genome editing. The systematic mapping study presented in this paper examines the literature on machine learning (ML) techniques employed in the prediction of CRISPR/Cas9 sgRNA on/off-target cleavage, focusing on improving support in sgRNA design activities and identifying areas currently being researched.

This area of research has greatly expanded recently, and we found it appropriate to work on a Systematic Mapping Study (SMS), an investigation that has proven to be an effective secondary study method. Unlike a classic review, in an SMS, no comparison of methods or results is made, while this task can instead be the subject of a systematic literature review that chooses one theme among those highlighted in this SMS. The study is illustrated in this paper. To the best of the authors' knowledge, no other SMS studies have been published on this topic.

Fifty-seven papers published in the period 2017–2022 (April, 30) were analyzed. This study reveals that the most widely used ML model is the convolutional neural network (CNN), followed by the feedforward neural network (FNN), while the use of other models is marginal. Other interesting information has emerged, such as the wide availability of both open code and platforms dedicated to supporting the activity of researchers or the fact that there is a clear prevalence of public funds that finance research on this topic.

OUP accepted manuscript

CRISPR/Cas base editors promise nucleotide-level control over DNA sequences, but the determinants of their activity remain incompletely understood. We measured base editing frequencies in two human cell lines for two cytosine and two adenine base editors at ∼14 000 target sequences and find that base editing activity is sequence-biased, with largest effects from nucleotides flanking the target base. Whether a base is edited depends strongly on the combination of its position in the target and the preceding base, acting to widen or narrow the effective editing window. The impact of features on editing rate depends on the position, with sequence bias efficacy mainly influencing bases away from the center of the window. We use these observations to train a machine learning model to predict editing activity per position, with accuracy ranging from 0.49 to 0.72 between editors, and with better generalization across datasets than existing tools. We demonstrate the usefulness of our model by predicting the efficacy of disease mutation correcting guides, and find that most of them suffer from more unwanted editing than pure outcomes. This work unravels the position-specificity of base editing biases and allows more efficient planning of editing campaigns in experimental and therapeutic contexts.