Off-target Issues Research Articles

On the shortcut learning in multilingual neural machine translation

In this study, we revisit the commonly-cited off-target issue in multilingual neural machine translation (MNMT). By carefully designing experiments on different MNMT scenarios and models, we attribute the off-target issue to the overfitting of the shortcuts of (non-centric, centric) language mappings. Specifically, the learned shortcuts biases MNMT to mistakenly translate non-centric languages into the centric language instead of the expected non-centric language for zero-shot translation. Analyses on learning dynamics show that the shortcut learning generally occurs in the later stage of model training, and multilingual pretraining accelerates and aggravates the shortcut learning. Based on these observations, we propose a simple and effective training strategy to eliminate the shortcuts in MNMT models by leveraging the forgetting nature of model training. The only difference from the standard training is that we remove the training instances that may induce the shortcut learning in the later stage of model training. Without introducing any additional data and computational costs, our approach can consistently and significantly improve the zero-shot translation performance by alleviating the shortcut learning for different MNMT models and benchmarks.

CRISPR/Cas System in Human Genetic Diseases

Gene editing involves altering a particular sequence within an organism’s genome using gene-editing techniques. Efficiently and accurately insert, delete or replace genes to alter their genetic information and phenotypic characteristics. DNA nuclease-based gene editing technology has advanced rapidly, from the first-generation editing system ZFNs, the second-generation TALENs to the third-generation CRISPR/Cas9 system, the efficiency of gene editing has been continuously improved, the cost has been gradually reduced, and the application scope has been expanding. The new classification system categorizes CRISPR-Cas proteins into Class I and II as the two main classes. Class I encompasses several subtypes, including Type I, III, and IV, all form complexes through the coordinated function of multiple Cas proteins. While Class II includes Type II, Type V, and Type VI, which all rely on single large Cas protein to perform all response. By designing single guide RNA (sgRNA), CRISPR can be used to target any gene sequence intended to be edited, achieving the wanted therapeutic outcome in treatment of inherited genetic disease. Although potential for off-target leads to undesired genetic alterations even result in oncogenesis, CRISPR-Cas still works as powerful therapy and gene editing tool in all aspects. Future research requires us to solve off-target issue. This review systematically introduces CRISPR-Cas systems as promising therapeutic strategy towards various genetic disease, explaining molecular mechanism and classification of CRISPR and differences among them.

Abstract 724: MPSA-AB5000: A high-throughput membrane proteins screening array for therapeutic biologics specificity profiling

Abstract The development of novel antibody drugs is a lengthy journey. Many factors can lead to failures during clinical development. According to reports, off-target binding of antibodies could also be a significant cause. It has been stated that about 25% of preclinical candidates might have off-target issues. With the emergence of new modalities like CD3 bispecific antibodies, ADCs, and CAR-T therapies, early-stage off-target screening of antibodies has become even more crucial. Although techniques such as immunohistochemistry, ELISA, and flow cytometry (FACS) have been used for off-target testing of antibodies, these methods still lack sensitivity and specificity, and their throughput is limited. To address the need for antibody target deconvolution and receptor identification, we developed a novel membrane protein screening array-AB5000 (MPSA-AB5000). MPSA-AB5000 is a high-throughput cell-based platform for identifying the targets of antibodies and other ligands that bind to membrane proteins. MPSA-AB5000 is an excellent solution for antibody specificity screening, which enables early off-target screening and can de-risk the antibody drug development program. The MPSA-AB5000 makes use of the full coverage of unique human membrane proteins, including receptors, transporters, enzymes, miscellaneous, and others. Each membrane protein is expressed in live cells to ensure native conformation. In addition, the same membrane protein clones with a C-terminal epitope tag were constructed, allowing confirmation of protein expression and cellular location. To provide the highest level of sensitivity, MPSA-AB5000 uses luciferase reporter cell line detection. The wash-free strategy reduces false negatives, and signal amplification makes remarkable discrimination. Compared to similar technologies, we possess significant advantages in the following aspects: a. High sensitivity and sigh throughput: our technology platform offers not only high sensitivity, capable of detecting subtle interactions but also supports high-throughput screening, allowing for the evaluation of a large number of samples in a short timeframe. b. Avoidance of false negative signals: by eliminating washing steps, our method reduces the risk of false negative signals, ensuring reliable detection of non-specific binding. c. Comprehensive validation approaches: We employ various validation methods such as FACS and ELISA to eliminate false positive signals and enhance the accuracy and credibility of our results. All the advantages ensure that MPSA-AB5000 is more rapid, simple, and highly sensitive than traditional assays. Citation Format: dan Li, Junyi Xiong, Jinying Ning, Feng Hao. MPSA-AB5000: A high-throughput membrane proteins screening array for therapeutic biologics specificity profiling [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 724.

A high-fidelity DNAzyme-assisted CRISPR/Cas13a system with single-nucleotide resolved specificity.

A CRISPR/Cas system represents an innovative tool for developing a new-generation biosensing and diagnostic strategy. However, the off-target issue (i.e., mistaken cleavage of nucleic acid targets and reporters) remains a great challenge for its practical applications. We hypothesize that this issue can be overcome by taking advantage of the site-specific cleavage ability of RNA-cleaving DNAzymes. To test this idea, we propose a DNAzyme Operation Enhances the Specificity of CRISPR/Cas13a strategy (termed DOES-CRISPR) to overcome the problem of relatively poor specificity that is typical of the traditional CRISPR/Cas13a system. The key to the design is that the partial hybridization of the CRISPR RNA (crRNA) with the cleavage fragment of off-target RNA was not able to activate the collateral cleavage activity of Cas13a. We showed that DOES-CRISPR can significantly improve the specificity of traditional CRISPR/Cas13a-based molecular detection by up to ∼43-fold. The broad utility of the strategy is illustrated through engineering three different systems for the detection of microRNAs (miR-17 and let-7e), CYP2C19*17 gene, SARS-Cov-2 variants (Gamma, Delta, and Omicron) and Omicron subtypes (BQ.1 and XBB.1) with single-nucleotide resolved specificity. Finally, clinical evaluation of this assay using 10 patient blood samples demonstrated a clinical sensitivity of 100% and specificity of 100% for genotyping CYP2C19*17, and analyzing 20 throat swab samples provided a diagnostic sensitivity of 95% and specificity of 100% for Omicron detection, and a clinical sensitivity of 92% and specificity of 100% for XBB.1 detection.

Natural Therapeutics in Aid of Treating Alzheimer's Disease: A Green Gateway Toward Ending Quest for Treating Neurological Disorders.

The current scientific community is facing a daunting challenge to unravel reliable natural compounds with realistic potential to treat neurological disorders such as Alzheimer’s disease (AD). The reported compounds/drugs mostly synthetic deemed the reliability and therapeutic potential largely due to their complexity and off-target issues. The natural products from nutraceutical compounds emerge as viable preventive therapeutics to fill the huge gap in treating neurological disorders. Considering that Alzheimer’s disease is a multifactorial disease, natural compounds offer the advantage of a multitarget approach, tagging different molecular sites in the human brain, as compared with the single-target activity of most of the drugs so far used to treat Alzheimer’s disease. A wide range of plant extracts and phytochemicals reported to possess the therapeutic potential to Alzheimer’s disease includes curcumin, resveratrol, epigallocatechin-3-gallate, morin, delphinidins, quercetin, luteolin, oleocanthal, and other phytochemicals such as huperzine A, limonoids, and azaphilones. Reported targets of these natural compounds include inhibition of acetylcholinesterase, amyloid senile plaques, oxidation products, inflammatory pathways, specific brain receptors, etc. We tenaciously aimed to review the in-depth potential of natural products and their therapeutic applications against Alzheimer’s disease, with a special focus on a diversity of medicinal plants and phytocompounds and their mechanism of action against Alzheimer’s disease pathologies. We strongly believe that the medicinal plants and phytoconstituents alone or in combination with other compounds would be effective treatments against Alzheimer’s disease with lesser side effects as compared to currently available treatments.

An unbiased method for evaluating the genome-wide specificity of base editors in rice.

Base editors can achieve targeted genomic base conversion. However, the off-target issue is one of the major concerns in their application. Whole-genome sequencing (WGS) at the individual level can provide direct information on genome-wide specificity, but it is difficult to distinguish true off-target single-nucleotide variants (SNVs) induced by base editors from background variation. Here we describe an unbiased WGS method for evaluating the specificity of base editors in rice. In this protocol, we describe the experimental design and provide details of vector construction, rice transformation and tissue culture, as well as a comprehensive WGS data analysis pipeline for overcoming two related core problems in various plant species: high background mutation rates and the heterogeneity of examined populations. Using this protocol, researchers can straightforwardly and accurately assess the genome-wide specificity of base editors and other genome editing tools in 12-15 weeks.

Localized delivery of CRISPR/dCas9 via layer-by-layer self-assembling peptide coating on nanofibers for neural tissue engineering

The clustered regularly interspaced short palindromic repeat (CRISPR) systems have a wide variety of applications besides precise genome editing. In particular, the CRISPR/dCas9 system can be used to control specific gene expression by CRISPR activation (CRISPRa) or interference (CRISPRi). However, the safety concerns associated with viral vectors and the possible off-target issues of systemic administration remain huge concerns to be safe delivery methods for CRISPR/Cas9 systems. In this study, a layer-by-layer (LbL) self-assembling peptide (SAP) coating on nanofibers is developed to mediate localized delivery of CRISPR/dCas9 systems. Specifically, an amphiphilic negatively charged SAP− is first coated onto PCL nanofibers through strong hydrophobic interactions, and the pDNA complexes and positively charged SAP+-RGD are then absorbed via electrostatic interactions. The SAPcoated scaffolds facilitate efficient loading and sustained release of the pDNA complexes, while enhancing cell adhesion and proliferation. As a proof of concept, the scaffolds are used to activate GDNF expression in mammalian cells, and the secreted GDNF subsequently promotes neurite outgrowth of rat neurons. These promising results suggest that the LbL self-assembling peptide coated nanofibers can be a new route to establish a bioactive interface, which provides a simple and efficient platform for the delivery of CRISPR/dCas9 systems for regenerative medicine.

Strategies for Targeting Cancer Immunotherapy Through Modulation of the Tumor Microenvironment

Cancer immunotherapeutic strategies have shifted the focus of cancer treatment from eradicating the tumor cell by conventional cytotoxic chemotherapy, to educating the immune system to eliminate tumor, thereby preventing the recurrence of cancer. The understanding of tumor microenvironment and its components which generate an immunosuppressive environment is critical in further developing efficient immunotherapies. In this review, we have classified the current immunotherapies based on their effect in modulating the tumor microenvironment. Additionally, we propose the inclusion of nanotechnology and tissue engineering approaches, which provide unique strategies to enhance the therapeutic efficacy and could lead to developing nano/engineered immunotherapies for improved clinical outcomes. Specifically, we focus on criteria for designing nano/engineered immunotherapies and discuss targeted delivery strategies that can optimize the bioavailability of immunotherapies and, in turn, improve the therapeutic outcomes in the treatment of cancer. Several strategies aimed to exploit the therapeutic benefits of immunotherapy are based on alterations of the complex immunosuppressive tumor microenvironment. Such targeted approaches have also been significantly improved by various design criteria based on the concepts of nanotechnology and tissue engineering. The properties of specific targeting, controlled release, and ability to attain enhanced therapeutic effects with low doses conferred by these approaches have immensely helped to surmount the side effects and off-target issues of existing methods. Incorporation of these design criteria while developing various carrier systems for targeted immunotherapy would certainly enhance their clinical translation potential, eventually augmenting anti-tumor responses. Modulation of tumor microenvironment with strategies discussed in this review will provide additional opportunities to improve cancer immunotherapy, especially in challenging diseases such as pancreatic and brain tumors. Various design attributes with targeted systems would provide numerous advantages, widening the scope of clinical translation and benefits to cancer patients.

What´s new in Gene Therapy of Hemophilia.

Several methods have been investigated to effectively and safely transmit genes that stimulate cells to release therapeutic factor VIII (FVIII) and factor IX (FIX) into the circulation of people with hemophilia (PWH). To review the role of gene therapy (GT) in PWH. A Cochrane Library and PubMed (MEDLINE) search related to the role of GT in hemophilia was analyzed. The most promising vectors for hemophilia GT are adeno-associated virus (AAV) and lentivirus. Several gene methods are available to lessen risks related to random vector integration and insertional mutagenesis, based on designer nucleases or CRISPR/Cas9 (clustered regularly interspaced short palindromic repeats/CRISPR associated system). However, off-target issues need to be more meticulously and widely evaluated. Some clinical studies on hemophilia B based on AAV have obtained transitory or subtherapeutic levels of FIX expression. Another problem is possible transitory liver toxicity. Therefore, to reduce unintentional immune responses, transitory immunosuppression must be used, particularly when administering high-vector doses. Codon-optimized FVIII or FIX transgenes are able to promote clotting factor expression levels. The inclusion of a hyper-active gain-of-function R338L mutation in the FIX gene (FIX-R338L [FIX Padua]) makes the procedure more effective. Achieving a safe and efficient remedy for hemophilia A and B by means of GT vector engineering needs further improvement. No randomized or quasi-randomized clinical trials of GT for hemophilia have been found. Given it is in its incipient period, there is need for well-designed clinical trials to evaluate the long-term practicability, efficacy and risks of GT for PWH.

Targeted genome engineering in human induced pluripotent stem cells from patients with hemophilia B using the CRISPR-Cas9 system

BackgroundReplacement therapy for hemophilia remains a lifelong treatment. Only gene therapy can cure hemophilia at a fundamental level. The clustered regularly interspaced short palindromic repeats–CRISPR associated nuclease 9 (CRISPR-Cas9) system is a versatile and convenient genome editing tool which can be applied to gene therapy for hemophilia.MethodsA patient’s induced pluripotent stem cells (iPSCs) were generated from their peripheral blood mononuclear cells (PBMNCs) using episomal vectors. The AAVS1-Cas9-sgRNA plasmid which targets the AAVS1 locus and the AAVS1-EF1α-F9 cDNA-puromycin donor plasmid were constructed, and they were electroporated into the iPSCs. When insertion of F9 cDNA into the AAVS1 locus was confirmed, whole genome sequencing (WGS) was carried out to detect the off-target issue. The iPSCs were then differentiated into hepatocytes, and human factor IX (hFIX) antigen and activity were measured in the culture supernatant. Finally, the hepatocytes were transplanted into non-obese diabetic/severe combined immunodeficiency disease (NOD/SCID) mice through splenic injection.ResultsThe patient’s iPSCs were generated from PBMNCs. Human full-length F9 cDNA was inserted into the AAVS1 locus of iPSCs of a hemophilia B patient using the CRISPR-Cas9 system. No off-target mutations were detected by WGS. The hepatocytes differentiated from the inserted iPSCs could secrete hFIX stably and had the ability to be transplanted into the NOD/SCID mice in the short term.ConclusionsPBMNCs are good somatic cell choices for generating iPSCs from hemophilia patients. The iPSC technique is a good tool for genetic therapy for human hereditary diseases. CRISPR-Cas9 is versatile, convenient, and safe to be used in iPSCs with low off-target effects. Our research offers new approaches for clinical gene therapy for hemophilia.

CRISPR/Cas9 System: A Breakthrough in Genome Editing

Clustered Regularly Interspaced Short Palindromic Repeats is a new and advance gene editing tool using specific nuclease enzyme for specific cleavage. It is the most efficient technique, commonly used for many purposes like gene therapy, production of desired plants and transgenic animals. But it has some limitations like off-target issue and this issue can be minimized by the production of specific sequence of guide RNA. In the future, it can be used for the treatment of many human genetic diseases

Hemophilia Gene Therapy: Ready for Prime Time?

Hemophilia A and B are congenital, X-linked bleeding disorders caused by mutations in the genes encoding for the blood clotting factor VIII (FVIII) or factor IX (FIX), respectively. Since the beginning of gene therapy, hemophilia has been considered an attractive disease target that served as a trailblazer for the field at large. Different technologies have been explored to efficiently and safely deliver the therapeutic FVIII and FIX genes into the patients' cells. Currently, the most promising vectors for hemophilia gene therapy are adeno-associated viral vectors (AAVs) and lentiviral vectors. More recently, gene editing approaches based on designer nucleases or CRISPR/Cas, have also been considered to minimize risks associated with random vector integration and insertional mutagenesis though off-target issues would have to be carefully and comprehensively assessed. In the past two decades, several phase 1 hemophilia gene therapy clinical trials have been initiated with varying success. In particular, the early gene therapy clinical trials in hemophilia B patients based on AAV showed either transient or subtherapeutic clotting factor expression levels. This could be ascribed, at least in part, to suboptimal vector design and/or inadvertent immune consequences triggering hepatic inflammation. Hence, there was an unmet need to further increase vector safety and efficacy in future trials, preferably by using lower vector doses. It is particularly encouraging that sustained therapeutic FVIII and FIX expression levels have recently been attained after gene therapy in patients with severe hemophilia paving the way towards pivotal trials and commercialization. Nevertheless, transient liver toxicity still occurs and the use of transient immunosuppression was still required to curtail inadvertent immune responses, especially at high vector doses. To further boost clotting factor expression levels, codon-usage optimized synthetic FVIII or FIX transgenes have been employed. Alternatively, we and others have shown that the incorporation of hyperactive gain-of-function R338L mutation in the FIX gene substantially increased the overall efficacy. It is inevitable that the continued improvements in vector engineering and new insights in the vector-patient interactions will further benefit the development of a safe and effective cure for hemophilia A and B.

Method for Identifying Small Molecule Inhibitors of the Protein-protein Interaction Between HCN1 and TRIP8b.

Hyperpolarization-activated cyclic nucleotide-gated (HCN) channels are expressed ubiquitously throughout the brain, where they function to regulate the excitability of neurons. The subcellular distribution of these channels in pyramidal neurons of hippocampal area CA1 is regulated by tetratricopeptide repeat-containing Rab8b interacting protein (TRIP8b), an auxiliary subunit. Genetic knockout of HCN pore forming subunits or TRIP8b, both lead to an increase in antidepressant-like behavior, suggesting that limiting the function of HCN channels may be useful as a treatment for Major Depressive Disorder (MDD). Despite significant therapeutic interest, HCN channels are also expressed in the heart, where they regulate rhythmicity. To circumvent off-target issues associated with blocking cardiac HCN channels, our lab has recently proposed targeting the protein-protein interaction between HCN and TRIP8b in order to specifically disrupt HCN channel function in the brain. TRIP8b binds to HCN pore forming subunits at two distinct interaction sites, although here the focus is on the interaction between the tetratricopeptide repeat (TPR) domains of TRIP8b and the C terminal tail of HCN1. In this protocol, an expanded description of a method for purifying TRIP8b and executing a high throughput screen to identify small molecule inhibitors of the interaction between HCN and TRIP8b, is described. The method for high throughput screening utilizes a Fluorescence Polarization (FP) -based assay to monitor the binding of a large TRIP8b fragment to a fluorophore-tagged eleven amino acid peptide corresponding to the HCN1 C terminal tail. This method allows 'hit' compounds to be identified based on the change in the polarization of emitted light. Validation assays are then performed to ensure that 'hit' compounds are not artifactual.

AJ-4 - Response Prediction for Anti-Receptor Tyrosine Kinase Agents

ABSTRACT Anti-cancer agents against receptor tyrosine kinases (RTKs) have been applied clinically for the last two decades. In parallel with drug development, much effort has gone into response prediction for these agents, some of which have been successfully adopted into clinical practice. The most successful examples include HER2 amplification in breast cancer for trastuzumab or lapatinib, and epidermal growth factor receptor (EGFR) mutations and EML4/ALK rearrangement in non-small-cell lung cancer for EGFR- and ALK-tyrosine kinase inhibitors (TKIs), respectively. These examples have emphasized the necessity of genetic surveillance of target RTKs in drug and biomarker developments. Nevertheless, studies suggest that the identification of additional response-predictive markers may require a broadening of scope beyond the target RTK gene. First, the off-target issue: while many TKIs are known to have multiple kinases as their targets, only a minority of kinases is screened in the preclinical phase of drug development, and the screening is mostly done in cell-free conditions. It should be recognized that TKIs may have off-targets out of the spectrum of in vitro screening, and that cross-activity against blind targets may potentially even compromise anti-tumor activity. Secondly, the downstream alteration: multiple lines of evidence suggest that aberrant activation of downstream signaling molecules might cause resistance to anti-RTK agents. Thirdly, the peri-RTK protein association: recent studies suggest that ligand expression rather than that of the target RTK itself may be predictive for certain anti-RTK agents. In addition, several studies suggest that overexpression of certain adaptor proteins might also be predictive of response to drugs against RTKs associated with the adaptor protein.

Bringing Kinases Into Focus: Efficient Drug Design Through the Use of Chemogenomic Toolkits

The study of protein target families, as opposed to single targets, has become a very powerful tool in chemogenomics-led drug discovery. By integrating comprehensive chemoinformatics and bioinformatics databases with customised analytical tools, a 'Toolkit' approach for the target family is possible, thus allowing predictions of the ligand class, affinity, selectivity and likely off-target issues to be made for the guidance of the medicinal chemist. In this review, we highlight the development and application of the Toolkit approach to the protein kinase superfamily, drawing on examples from lead optimisation studies and the design of focused libraries for lead discovery.