Prediction of on-target and off-target activity of CRISPR-Cas13d guide RNAs using deep learning

A systematic method for solving data imbalance in CRISPR off-target prediction tasks

Accurately identifying potential off-target sites in the CRISPR/Cas9 system is crucial for improving the efficiency and safety of editing. However, the imbalance of available off-target datasets has posed a major obstacle in enhancing prediction performance. Despite several prediction models have been developed to address this issue, there remains a lack of systematic research on handling data imbalance in off-target prediction. This article systematically investigates the data imbalance issue in off-target datasets and explores numerous methods to process data imbalance from a novel perspective. First, we highlight the impact of the imbalance problem on off-target prediction tasks by determining the imbalance ratios present in these datasets. Then, we provide a comprehensive review of various sampling techniques and cost-sensitive methods to mitigate class imbalance in off-target datasets. Finally, systematic experiments are conducted on several state-of-the-art prediction models to illustrate the impact of applying data imbalance solutions. The results show that class imbalance processing methods significantly improve the off-target prediction capabilities of the models across multiple testing datasets. The code and datasets used in this study are available at https://github.com/gzrgzx/CRISPR_Data_Imbalance.

Efficient, specific, and combinatorial control of endogenous exon splicing with dCasRx-RBM25

Efficient targeted control of splicing is a major goal of functional genomics and therapeutic applications. Guide (g)RNA-directed, deactivated (d)Cas CRISPR enzymes fused to splicing effectors represent a promising strategy due to the flexibility of these systems. However, efficient, specific, and generalizable activation of endogenous exons using this approach has not been previously reported. By screening over 300 dCasRx-splicing factor fusion proteins tethered to splicing reporters, we identify dCasRx-RBM25 as a potent activator of exons. Moreover, dCasRx-RBM25 efficiently activates the splicing of ∼90% of targeted endogenous alternative exons and displays high on-target specificity. Using gRNA arrays for combinatorial targeting, we demonstrate that dCasRx-RBM25 enables multiplexed activation and repression of exons. Using this feature, the targeting of neural-regulated exons in Ptpb1 and Puf60 in embryonic stem cells reveals combinatorial effects on downstream alternative splicing events controlled by these factors. Collectively, our results enable versatile, combinatorial exon-resolution functional assays and splicing-directed therapeutic applications.

Precise Gene Knock-In Tools with Minimized Risk of DSBs: A Trend for Gene Manipulation

Gene knock-in refers to the insertion of exogenous functional genes into a target genome to achieve continuous expression. Currently, most knock-in tools are based on site-directed nucleases, which can induce double-strand breaks (DSBs) at the target, following which the designed donors carrying functional genes can be inserted via the endogenous gene repair pathway. The size of donor genes is limited by the characteristics of gene repair, and the DSBs induce risks like genotoxicity. New generation tools, such as prime editing, transposase, and integrase, can insert larger gene fragments while minimizing or eliminating the risk of DSBs, opening new avenues in the development of animal models and gene therapy. However, the elimination of off-target events and the production of delivery carriers with precise requirements remain challenging, restricting the application of the current knock-in treatments to mainly in vitro settings. Here, a comprehensive review of the knock-in tools that do not/minimally rely on DSBs and use other mechanisms is provided. Moreover, the challenges and recent advances of in vivo knock-in treatments in terms of the therapeutic process is discussed. Collectively, the new generation of DSBs-minimizing and large-fragment knock-in tools has revolutionized the field of gene editing, from basic research to clinical treatment.

Tunable translation-level CRISPR interference by dCas13 and engineered gRNA in bacteria

Although CRISPR-dCas13, the RNA-guided RNA-binding protein, was recently exploited as a translation-level gene expression modulator, it has still been difficult to precisely control the level due to the lack of detailed characterization. Here, we develop a synthetic tunable translation-level CRISPR interference (Tl-CRISPRi) system based on the engineered guide RNAs that enable precise and predictable down-regulation of mRNA translation. First, we optimize the Tl-CRISPRi system for specific and multiplexed repression of genes at the translation level. We also show that the Tl-CRISPRi system is more suitable for independently regulating each gene in a polycistronic operon than the transcription-level CRISPRi (Tx-CRISPRi) system. We further engineer the handle structure of guide RNA for tunable and predictable repression of various genes in Escherichia coli and Vibrio natriegens. This tunable Tl-CRISPRi system is applied to increase the production of 3-hydroxypropionic acid (3-HP) by 14.2-fold via redirecting the metabolic flux, indicating the usefulness of this system for the flux optimization in the microbial cell factories based on the RNA-targeting machinery.

Big data and deep learning for RNA biology

The exponential growth of big data in RNA biology (RB) has led to the development of deep learning (DL) models that have driven crucial discoveries. As constantly evidenced by DL studies in other fields, the successful implementation of DL in RB depends heavily on the effective utilization of large-scale datasets from public databases. In achieving this goal, data encoding methods, learning algorithms, and techniques that align well with biological domain knowledge have played pivotal roles. In this review, we provide guiding principles for applying these DL concepts to various problems in RB by demonstrating successful examples and associated methodologies. We also discuss the remaining challenges in developing DL models for RB and suggest strategies to overcome these challenges. Overall, this review aims to illuminate the compelling potential of DL for RB and ways to apply this powerful technology to investigate the intriguing biology of RNA more effectively.

Discovery platforms for RNA therapeutics

RNA therapeutics are emerging as a unique opportunity to drug currently "undruggable" molecules and diseases. While their advantages over conventional, small molecule drugs, their therapeutic implications and the tools for their effective in vivo delivery have been extensively reviewed, little attention has been so far paid to the technological platforms exploited for the discovery of RNA therapeutics. Here, we provide an overview of the existing platforms and ex vivo assays for RNA discovery, their advantages and disadvantages, as well as their main fields of application, with specific focus on RNA therapies that have reached either phase 3 or market approval.

Repair of CRISPR-guided RNA breaks enables site-specific RNA excision in human cells

Genome editing with CRISPR RNA-guided endonucleases generates DNA breaks that are resolved by cellular DNA repair machinery. However, analogous methods to manipulate RNA remain unavailable. We show that site-specific RNA breaks generated with type-III CRISPR complexes are repaired in human cells and that this repair can be used for programmable deletions in human transcripts to restore gene function. Collectively, this work establishes a technology for precise RNA manipulation with potential therapeutic applications.

Low copy CRISPR-Cas13d mitigates collateral RNA cleavage

While CRISPR-Cas13 systems excel in accurately targeting RNA, the potential for collateral RNA degradation poses a concern for therapeutic applications and limits broader adoption for transcriptome perturbations. We evaluate the extent to which collateral RNA cleavage occurs when Rfx Cas13d is delivered via plasmid transfection or lentiviral transduction and find that collateral activity only occurs with high levels of Rfx Cas13d expression. Using transcriptome-scale and combinatorial CRISPR pooled screens in cell lines with low-copy Rfx Cas13d, we find high on-target knockdown, without extensive collateral activity regardless of the expression level of the target gene. In contrast, transfection of Rfx Cas13d, which yields higher nuclease expression, results in collateral RNA degradation. Further, our analysis of a high-fidelity Cas13 variant uncovers a marked decrease in on-target efficiency, suggesting that its reduced collateral activity may be due to an overall diminished nuclease capability.

Repurposing CRISPR-Cas13 systems for robust mRNA trans-splicing

Type VI CRISPR enzymes have been developed as programmable RNA-guided Cas proteins for eukaryotic RNA editing. Notably, Cas13 has been utilized for site-targeted single base edits, demethylation, RNA cleavage or knockdown and alternative splicing. However, the ability to edit large stretches of mRNA transcripts remains a significant challenge. Here, we demonstrate that CRISPR-Cas13 systems can be repurposed to assist trans-splicing of exogenous RNA fragments into an endogenous pre-mRNA transcript, a method termed CRISPR Assisted mRNA Fragment Trans-splicing (CRAFT). Using split reporter-based assays, we evaluate orthogonal Cas13 systems, optimize guide RNA length and screen for optimal trans-splicing site(s) across a range of intronic targets. We achieve markedly improved editing of large 5' and 3' segments in different endogenous mRNAs across various mammalian cell types compared to other spliceosome-mediated trans-splicing methods. CRAFT can serve as a versatile platform for attachment of protein tags, studying the impact of multiple mutations/single nucleotide polymorphisms, modification of untranslated regions (UTRs) or replacing large segments of mRNA transcripts.

A versatile CRISPR-Cas13d platform for multiplexed transcriptomic regulation and metabolic engineering in primary human T cells

CRISPR technologies have begun to revolutionize T cell therapies; however, conventional CRISPR-Cas9 genome-editing tools are limited in their safety, efficacy, and scope. To address these challenges, we developed multiplexed effector guide arrays (MEGA), a platform for programmable and scalable regulation of the T cell transcriptome using the RNA-guided, RNA-targeting activity of CRISPR-Cas13d. MEGA enables quantitative, reversible, and massively multiplexed gene knockdown in primary human T cells without targeting or cutting genomic DNA. Applying MEGA to a model of CAR T cell exhaustion, we robustly suppressed inhibitory receptor upregulation and uncovered paired regulators of T cell function through combinatorial CRISPR screening. We additionally implemented druggable regulation of MEGA to control CAR activation in a receptor-independent manner. Lastly, MEGA enabled multiplexed disruption of immunoregulatory metabolic pathways to enhance CAR T cell fitness and anti-tumor activity in vitro and in vivo. MEGA offers a versatile synthetic toolkit for applications in cancer immunotherapy and beyond.

Past, present, and future of CRISPR genome editing technologies

Genome editing has been a transformative force in the life sciences and human medicine, offering unprecedented opportunities to dissect complex biological processes and treat the underlying causes of many genetic diseases. CRISPR-based technologies, with their remarkable efficiency and easy programmability, stand at the forefront of this revolution. In this Review, we discuss the current state of CRISPR gene editing technologies in both research and therapy, highlighting limitations that constrain them and the technological innovations that have been developed in recent years to address them. Additionally, we examine and summarize the current landscape of gene editing applications in the context of human health and therapeutics. Finally, we outline potential future developments that could shape gene editing technologies and their applications in the coming years.

Next-generation forward genetic screens: uniting high-throughput perturbations with single-cell analysis

Programmable genome-engineering technologies, such as CRISPR (clustered regularly interspaced short palindromic repeats) nucleases and massively parallel CRISPR screens that capitalize on this programmability, have transformed biomedical science. These screens connect genes and noncoding genome elements to disease-relevant phenotypes, but until recently have been limited to individual phenotypes such as growth or fluorescent reporters of gene expression. By pairing massively parallel screens with high-dimensional profiling of single-cell types/states, we can now measure how individual genetic perturbations or combinations of perturbations impact the cellular transcriptome, proteome, and epigenome. We review technologies that pair CRISPR screens with single-cell multiomics and the unique opportunities afforded by extending pooled screens using deep multimodal phenotyping.

Artificial Intelligence and Anticancer Drug Development-Keep a Cool Head

Artificial intelligence (AI) is progressively spreading through the world of health, particularly in the field of oncology. AI offers new, exciting perspectives in drug development as toxicity and efficacy can be predicted from computer-designed active molecular structures. AI-based in silico clinical trials are still at their inception in oncology but their wider use is eagerly awaited as they should markedly reduce durations and costs. Health authorities cannot neglect this new paradigm in drug development and should take the requisite measures to include AI as a new pillar in conducting clinical research in oncology.

New design strategies for ultra-specific CRISPR-Cas13a-based RNA detection with single-nucleotide mismatch sensitivity

An increasingly pressing need for clinical diagnostics has required the development of novel nucleic acid-based detection technologies that are sensitive, fast, and inexpensive, and that can be deployed at point-of-care. Recently, the RNA-guided ribonuclease CRISPR-Cas13 has been successfully harnessed for such purposes. However, developing assays for detection of genetic variability, for example single-nucleotide polymorphisms, is still challenging and previously described design strategies are not always generalizable. Here, we expanded our characterization of LbuCas13a RNA-detection specificity by performing a combination of experimental RNA mismatch tolerance profiling, molecular dynamics simulations, protein, and crRNA engineering. We found certain positions in the crRNA-target-RNA duplex that are particularly sensitive to mismatches and establish the effect of RNA concentration in mismatch tolerance. Additionally, we determined that shortening the crRNA spacer or modifying the direct repeat of the crRNA leads to stricter specificities. Furthermore, we harnessed our understanding of LbuCas13a allosteric activation pathways through molecular dynamics and structure-guided engineering to develop novel Cas13a variants that display increased sensitivities to single-nucleotide mismatches. We deployed these Cas13a variants and crRNA design strategies to achieve superior discrimination of SARS-CoV-2 strains compared to wild-type LbuCas13a. Together, our work provides new design criteria and Cas13a variants to use in future easier-to-implement Cas13-based RNA detection applications.

Massively parallel profiling of RNA-targeting CRISPR-Cas13d

CRISPR-Cas13d cleaves RNA and is used in vivo and for diagnostics. However, a systematic understanding of its RNA binding and cleavage specificity is lacking. Here, we describe an RNA Chip-Hybridized Association-Mapping Platform (RNA-CHAMP) for measuring the binding affinity for > 10,000 RNAs containing structural perturbations and other alterations relative to the CRISPR RNA (crRNA). Deep profiling of Cas13d reveals that it does not require a protospacer flanking sequence but is exquisitely sensitive to secondary structure within the target RNA. Cas13d binding is penalized by mismatches in the distal crRNA-target RNA region, while alterations in the proximal region inhibit nuclease activity. A biophysical model built from these data reveals that target recognition initiates in the distal end of the target RNA. Using this model, we design crRNAs that can differentiate between SARS-CoV-2 variants by modulating nuclease activation. This work describes the key determinants of RNA targeting by a type VI CRISPR enzyme.

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.

Programmable RNA targeting with CRISPR-Cas13

The RNA-targeting CRISPR-Cas13 system has enabled precise engineering of endogenous RNAs, significantly advancing our understanding of RNA regulation and the development of RNA-based diagnostic and therapeutic applications. This review aims to provide a summary of Cas13-based RNA targeting tools and applications, discuss limitations and challenges of existing tools and suggest potential directions for further development of the RNA targeting system.

Deep learning and CRISPR-Cas13d ortholog discovery for optimized RNA targeting

Effective and precise mammalian transcriptome engineering technologies are needed to accelerate biological discovery and RNA therapeutics. Despite the promise of programmable CRISPR-Cas13 ribonucleases, their utility has been hampered by an incomplete understanding of guide RNA design rules and cellular toxicity resulting from off-target or collateral RNA cleavage. Here, we quantified the performance of over 127,000 RfxCas13d (CasRx) guide RNAs and systematically evaluated seven machine learning models to build a guide efficiency prediction algorithm orthogonally validated across multiple human cell types. Deep learning model interpretation revealed preferred sequence motifs and secondary features for highly efficient guides. We next identified and screened 46 novel Cas13d orthologs, finding that DjCas13d achieves low cellular toxicity and high specificity-even when targeting abundant transcripts in sensitive cell types, including stem cells and neurons. Our Cas13d guide efficiency model was successfully generalized to DjCas13d, illustrating the power of combining machine learning with ortholog discovery to advance RNA targeting in human cells.

Decoding Heterogenous Single-cell Perturbation Responses

Understanding diverse responses of individual cells to the same perturbation is central to many biological and biomedical problems. Current methods, however, do not precisely quantify the strength of perturbation responses and, more importantly, reveal new biological insights from heterogeneity in responses. Here we introduce the perturbation-response score (PS), based on constrained quadratic optimization, to quantify diverse perturbation responses at a single-cell level. Applied to single-cell transcriptomes of large-scale genetic perturbation datasets (e.g., Perturb-seq), PS outperforms existing methods for quantifying partial gene perturbation responses. In addition, PS presents two major advances. First, PS enables large-scale, single-cell-resolution dosage analysis of perturbation, without the need to titrate perturbation strength. By analyzing the dose-response patterns of over 2,000 essential genes in Perturb-seq, we identify two distinct patterns, depending on whether a moderate reduction in their expression induces strong downstream expression alterations. Second, PS identifies intrinsic and extrinsic biological determinants of perturbation responses. We demonstrate the application of PS in contexts such as T cell stimulation, latent HIV-1 expression, and pancreatic cell differentiation. Notably, PS unveiled a previously unrecognized, cell-type-specific role of coiled-coil domain containing 6 (CCDC6) in guiding liver and pancreatic lineage decisions, where CCDC6 knockouts drive the endoderm cell differentiation towards liver lineage, rather than pancreatic lineage. The PS approach provides an innovative method for dose-to-function analysis and will enable new biological discoveries from single-cell perturbation datasets.

One sentence summary

We present a method to quantify diverse perturbation responses and discover novel biological insights in single-cell perturbation datasets.

RNA structure modulates Cas13 activity and enables mismatch detection

The RNA-targeting CRISPR nuclease Cas13 has emerged as a powerful tool for applications ranging from nucleic acid detection to transcriptome engineering and RNA imaging1-6. Cas13 is activated by the hybridization of a CRISPR RNA (crRNA) to a complementary single-stranded RNA (ssRNA) protospacer in a target RNA1,7. Though Cas13 is not activated by double-stranded RNA (dsRNA) in vitro, it paradoxically demonstrates robust RNA targeting in environments where the vast majority of RNAs are highly structured2,8. Understanding Cas13's mechanism of binding and activation will be key to improving its ability to detect and perturb RNA; however, the mechanism by which Cas13 binds structured RNAs remains unknown9. Here, we systematically probe the mechanism of LwaCas13a activation in response to RNA structure perturbations using a massively multiplexed screen. We find that there are two distinct sequence-independent modes by which secondary structure affects Cas13 activity: structure in the protospacer region competes with the crRNA and can be disrupted via a strand-displacement mechanism, while structure in the region 3' to the protospacer has an allosteric inhibitory effect. We leverage the kinetic nature of the strand displacement process to improve Cas13-based RNA detection, enhancing mismatch discrimination by up to 50-fold and enabling sequence-agnostic mutation identification at low (<1%) allele frequencies. Our work sets a new standard for CRISPR-based nucleic acid detection and will enable intelligent and secondary-structure-guided target selection while also expanding the range of RNAs available for targeting with Cas13.

Cas13d-mediated isoform-specific RNA knockdown with a unified computational and experimental toolbox

Alternative splicing is an essential mechanism for diversifying proteins, in which mature RNA isoforms produce proteins with potentially distinct functions. Two major challenges in characterizing the cellular function of isoforms are the lack of experimental methods to specifically and efficiently modulate isoform expression and computational tools for complex experimental design. To address these gaps, we developed and methodically tested a strategy which pairs the RNA-targeting CRISPR/Cas13d system with guide RNAs that span exon-exon junctions in the mature RNA. We performed a high-throughput essentiality screen, quantitative RT-PCR assays, and PacBio long read sequencing to affirm our ability to specifically target and robustly knockdown individual RNA isoforms. In parallel, we provide computational tools for experimental design and screen analysis. Considering all possible splice junctions annotated in GENCODE for multi-isoform genes and our gRNA efficacy predictions, we estimate that our junction-centric strategy can uniquely target up to 89% of human RNA isoforms, including 50,066 protein-coding and 11,415 lncRNA isoforms. Importantly, this specificity spans all splicing and transcriptional events, including exon skipping and inclusion, alternative 5' and 3' splice sites, and alternative starts and ends.

Repair of CRISPR-guided RNA breaks enables site-specific RNA editing in human cells

Genome editing with CRISPR RNA-guided endonucleases generates DNA breaks that are resolved by cellular DNA repair machinery. However, analogous methods to manipulate RNA remain unavailable. Here, we show that site-specific RNA breaks generated with RNA-targeting CRISPR complexes are repaired in human cells, and this repair can be used for programmable deletions in human transcripts that restore gene function. Collectively, this work establishes a technology for precise RNA manipulation with potential therapeutic applications.

New design strategies for ultra-specific CRISPR-Cas13a-based RNA-diagnostic tools with single-nucleotide mismatch sensitivity

The pressing need for clinical diagnostics has required the development of novel nucleic acid-based detection technologies that are sensitive, fast, and inexpensive, and that can be deployed at point-of-care. Recently, the RNA-guided ribonuclease CRISPR-Cas13 has been successfully harnessed for such purposes. However, developing assays for detection of genetic variability, for example single-nucleotide polymorphisms, is still challenging and previously described design strategies are not always generalizable. Here, we expanded our characterization of LbuCas13a RNA-detection specificity by performing a combination of experimental RNA mismatch tolerance profiling, molecular dynamics simulations, protein, and crRNA engineering. We found certain positions in the crRNA-target-RNA duplex that are particularly sensitive to mismatches and establish the effect of RNA concentration in mismatch tolerance. Additionally, we determined that shortening the crRNA spacer or modifying the direct repeat of the crRNA leads to stricter specificities. Furthermore, we harnessed our understanding of LbuCas13a allosteric activation pathways through molecular dynamics and structure-guided engineering to develop novel Cas13a variants that display increased sensitivities to single-nucleotide mismatches. We deployed these Cas13a variants and crRNA design strategies to achieve superior discrimination of SARS-CoV-2 strains compared to wild-type LbuCas13a. Together, our work provides new design criteria and new Cas13a variants for easier-to-implement Cas13-based diagnostics.

KEY POINTS

Certain positions in the Cas13a crRNA-target-RNA duplex are particularly sensitive to mismatches.Understanding Cas13a's allosteric activation pathway allowed us to develop novel high-fidelity Cas13a variants.These Cas13a variants and crRNA design strategies achieve superior discrimination of SARS-CoV-2 strains.

GRAPHICAL ABSTRACT