To assess the AI agent\u2019s capabilities to perform gene-editing research, we compiled an evaluation test set, Gene-editing bench, from both public sources and human experts (details in Supplementary Note C<\/a>). This test set covers a variety of gene-editing tasks (Fig. 1<\/a>). By using the test set, we performed extensive evaluation of CRISPR-GPT\u2019s capabilities in major gene-editing research tasks, such as experiment planning, delivery selection, single guide (sg)RNA design and experiment troubleshooting. In addition, we invited human experts to perform a thorough user experience evaluation of CRISPR-GPT and collected valuable human feedback.<\/p>\n Further, we implement CRISPR-GPT in real-world wet labs. Using CRISPR-GPT as an AI co-pilot, we demonstrate a fully AI-guided knockout (KO) of four genes: TGF\u03b2R1, SNAI1, BAX and BCL2L1, using CRISPR-Cas12a in human lung adenocarcinoma cell line, as well as AI-guided CRISPR-dCas9 epigenetic activation of two genes: NCR3LG1 and CEACAM1, in a human melanoma model. All these wet-lab experiments were carried out by junior researchers not familiar with gene editing. They both succeeded on the first attempt, confirmed by not only editing efficiencies, but also biologically relevant phenotypes and protein-level validation, highlighting the potential of LLM-guided biological research.<\/p>\n CRISPR-GPT is a multi-agent, compositional system involving a team of LLM-based agents, including an LLM Planner agent, a User-proxy agent, Task executor agents and Tool provider agents (Fig. 2a<\/a>). These components are powered by LLMs to interact with one another as well as the human user. We also refer to the full system as an \u2018agent\u2019 to encapsulate the overall functionalities.<\/p>\n To automate biological experiment design and analysis, we view the overall problem as sequential decision-making. This perspective frames the interaction between the user and the automated system as a series of decision-making steps, each essential for progressing towards the ultimate goal. Take the Auto mode for example. A user can initiate the process with a meta-request, for example, \u201cI want to knock out the human TGF\u03b2R1 gene in A549 lung cancer cells\u201d. In response, the agent\u2019s LLM Planner will analyse the user\u2019s request, drawing on its extensive internal knowledge base via retrieval techniques. Leveraging the reasoning abilities of the base LLM, the Planner generates a chain-of-thought37<\/a>,38<\/a> reasoning path and chooses an optimal action from a set of plausible ones while following expert-written guidelines. Consequently, the Planner breaks down the user\u2019s request into a sequence of discrete tasks, for example, \u2018CRISPR-Cas system selection\u2019 and \u2018gRNA design for knockout\u2019, while managing interdependencies among these tasks. Each individual task is solved by an LLM-powered state machine, via the Task executor, entailing a sequence of states to progress towards the specific goal. After the meta-task decomposition, the Task executor will chain the state machines of the corresponding tasks together into a larger state machine and begin the execution process, systematically addressing each task in sequence to ensure that the experiment\u2019s objectives are met efficiently and effectively (Fig. 2a<\/a>).<\/p>\n The User-proxy agent is responsible for guiding the user throughout the decision-making process via multiple rounds of textual interactions (typical user interactions required by each task detailed in Supplementary Table 2<\/a>). At each decision point, the internal state machine presents a \u2018state variable\u2019 to the User-proxy agent, which includes the current task instructions, and specifies any necessary input from the user to proceed. The User-proxy agent then interprets this state given the user interactions and makes informed decisions as input to the Task executor on behalf of the user. Subsequently, the User-proxy agent receives feedback from the Task executor, including the task results and the reasoning process that led to those outcomes. Concurrently, the User-proxy agent continues to interact with the user and provides them with instructions, continuously integrating their feedback to ensure alignment with the user\u2019s objectives (detailed in Methods<\/a>; Fig. 2a<\/a> and Supplementary Fig. 1<\/a>).<\/p>\n To enhance the LLM with domain knowledge, we enable the CRISPR agent to retrieve and synthesize information from published protocols, peer-reviewed research papers and expert-written guidelines, and to utilize external tools and conduct web searches via Tool provider agents (Fig. 2a<\/a>).<\/p>\n For an end-to-end gene-editing workflow, CRISPR-GPT typically constructs a chain of tasks that includes selecting the appropriate CRISPR system, recommending delivery methods, designing gRNAs, predicting off-target effects, selecting experimental protocols, planning validation assays and performing data analysis (Fig. 2b<\/a>). The system\u2019s modular architecture facilitates easy integration of additional functionalities and tools. CRISPR-GPT serves as a prototype LLM-powered AI co-pilot for scientific research, with potential applications extending beyond gene editing.<\/p>\n CRISPR-GPT agents automate gene-editing research tasks<\/p>\n CRISPR-GPT is able to automate gene-editing research via several key functionalities. For each functionality we discuss the agentic implementation and evaluation results.<\/p>\n Experiment planning<\/p>\n The Task planner agent is charged with directing the entire workflow and breaking down the user\u2019s meta-request into a task chain (Fig. 2b<\/a>). While the Planner selects and follows a predefined workflow in the Meta mode, it is able to take in freestyle user requests and auto-build a customized workflow in the Auto mode. For example, a user may only need part of the pre-designed workflow including CRISPR-Cas system selection, delivery method selection, gRNA design and experimental protocol selection before the experiment. Then the Task planner agent extracts the right information from the user request and assembles a customized workflow to suit user needs (Fig. 3a<\/a>). To evaluate CRISPR-GPT\u2019s ability to correctly layout gene-editing tasks and manage intertask dependence, we compiled a planning test set, as part of the Gene-editing bench, with user requests and golden answers curated by human experts. Using this test set, we evaluated CRISPR-GPT in comparison with prompted general LLMs, showing that CRISPR-GPT outperforms general LLMs in planning gene-editing tasks (Fig. 3b<\/a>). The CRISPR-GPT agent driven by GPT-4o scored over 0.99 in accuracy, precision, recall and F1 score, and had <0.05 normalized Levenshtein distance between agent-generated plans and golden answers (Fig. 3b<\/a>). For extensive description of the test set and evaluation, please see Supplementary Note C1<\/a>.<\/p>\n Fig. 3: Task decomposition and experiment planning in CRISPR-GPT Auto mode with performance evaluation. a, The LLM Planner agent automatically breaks down the user\u2019s meta-request to a sequence of tasks. Then it assembles a customized workflow of the chained tasks to meet the user\u2019s needs. Part of the panel was created with BioRender.com\/qy4v<\/a>. b, Evaluation of the LLM Planner using a gene-editing planning test set. For each test case, we generate three independent answers from each model and report the average scores (Supplementary Note C<\/a>). Data shown are the mean\u2009\u00b1\u2009s.d. (n\u2009=\u20093 per group).<\/p>\n Delivery method selection<\/p>\n We present and evaluate the delivery agent of CRISPR-GPT (Fig. 4a,b<\/a>). Delivery is a critical step for all gene-editing experiments. CRISPR-GPT equips LLM with expert-tailored instructions and external tools to choose delivery methods. Specifically, the agent first tries to understand the biological system that the user is planning to edit. It extracts keywords for the target cell\/tissues\/organisms, performs Google search and summarizes the results. Then, given its own knowledge and search results, CRISPR-GPT matches the user case with a major biological category (cell lines, primary cells, in vivo and so on) which reduces the possible options to a focused set of candidate methods. Next, CRISPR-GPT performs literature search with user and method-specific keywords, and ranks the candidate methods on the basis of citation metrics to suggest a primary and a secondary delivery method (Fig. 4a<\/a>). To evaluate the performance of this module, we compiled test cases including 50 biological systems as part of the Gene-editing bench. For each case, we invited three human experts to score potential delivery options and utilized those as ground truth. We then evaluated the output of CRISPR-GPT and baseline models by comparing to the pre-compiled ground-truth score sheet. We found that CRISPR-GPT outperforms the baseline GPT-4 and GPT-3-turbo models (Fig. 4b<\/a>). The agent has a substantial edge on difficult tasks such as those involving hard-to-transfect cell lines and primary cell types. We also noticed that including an additional literature search step improves the agent\u2019s performance only moderately. More details about the delivery selection evaluation can be found in Supplementary Note C2<\/a>.<\/p>\n Fig. 4: CRISPR-GPT automates gene-editing research and experiment tasks. a, Design of delivery method selection agent in CRISPR-GPT, showing the workflow, example request and a series of agent thoughts\u2013actions to identify most suitable delivery methods for the user\u2019s needs. b, Evaluation results of delivery method selection using CRISPR-GPT and baseline models. Data shown are the mean\u2009\u00b1\u2009s.d. (n\u2009=\u20093 per group). c, Design of gRNA design agent in CRISPR-GPT, showing the workflow, example request and a series of agent thoughts\u2013actions to select top-ranked gRNA customized to user\u2019s request. d, Evaluation results of gRNA design using CRISPR-GPT and baseline models. Models were prompted to generate functions and associated parameters to design gRNAs requested by the user. Data shown are the mean\u2009\u00b1\u2009s.d. (n\u2009=\u20093 per group). e, Design of Q&A mode in CRISPR-GPT, showing the workflow, example request and a series of agent thoughts\u2013actions to answer gene-editing questions. f, Evaluation of CRISPR-GPT and baseline models for answering gene-editing research questions. Models were prompted to generate answers, which were anonymized, evaluated by three human experts in a fully blind setup. Scores range from 1 (lowest) to 5 (highest). All scores above were from three independent trials (details on evaluation in Supplementary Note C<\/a>). Data shown are the mean\u2009\u00b1\u2009s.d. (n\u2009=\u20093 per group).<\/p>\n gRNA design<\/p>\n Good gRNA design is crucial for the success of CRISPR experiments. Various gRNA design tools and software, such as CRISPick39<\/a>,40<\/a>,41<\/a>,42<\/a> and ChopChop43<\/a>, are available. However, we believe there are two key challenges in general usage: (1) choosing a trustworthy source and (2) difficulty in quickly identifying gRNAs that suit specific user requirements or experiment contexts, often requiring lengthy sorting, ranking or literature review. To address these issues, we utilized pre-designed gRNA tables from CRISPick, a reputable and widely used tool. We leverage the reasoning capabilities of LLMs to accurately identify regions of interest and quickly extract relevant gRNAs. This approach is similar to the recently proposed \u2018chain-of-tables\u2019 methodology44<\/a> (Fig. 4c<\/a>, Extended Data Fig. 1a<\/a>, and Supplementary Demo Videos 1<\/a> and 2<\/a>). To evaluate the ability of CRISPR-GPT to correctly retrieve gRNAs, we compiled a gRNA design test set with ground truth from human experts (detailed in Supplementary Note C3<\/a>). CRISPR-GPT agent outperforms the baseline LLMs in accurately selecting gRNA design actions and configuring the arguments (Fig. 4d<\/a>).<\/p>\n Further, we picked a real-world test case from a cancer biology study, in which many highly ranked gRNA designs did not generate biological phenotypes, even when their editing efficiencies were high45<\/a>. Instead, the authors of the study had to design gRNAs manually against exons encoding important functional domains within a gene, and exon-selected gRNAs induced expected cancer-killing effects. We tested CRISPR-GPT for designing gRNAs targeting BRD4 gene from this study, and compared results with those generated by CRISPick and CHOPCHOP (Extended Data Fig. 1<\/a>). CRISPR-GPT was uniquely able to select the key exons, Exon3\u20134, within BRD4. In contrast, gRNAs designed by CRISPick or CHOPCHOP would be likely ineffective, as 7 out of 8 gRNAs mapped to non-essential exons (Extended Data Fig. 1<\/a>). Taken together, our results support the benefit and validity of this module.<\/p>\n Other functions and tools<\/p>\n CRISPR-GPT provides specific suggestions on the choice of the CRISPR system, experimental and validation protocol by leveraging LLM\u2019s reasoning ability and retrieving information from an expert-reviewed knowledge base46<\/a>,47<\/a>,48<\/a>,49<\/a>,50<\/a>,51<\/a>. It also offers automated gRNA off-target prediction, primer design for validation experiments, and data analysis. In particular, the agent provides fully automated solutions to run external software, such as Primer3 (ref. 52<\/a>), CRISPRitz53<\/a>,54<\/a>,55<\/a>,56<\/a> and CRISPResso2 (ref. 57<\/a>) (Supplementary Table 1<\/a>). We focused on implementing these tools as they are considered gold standard in respective tasks and have been extensively validated in previous work.<\/p>\n Problem solving via fine-tuning LLM on scientific discussion<\/p>\n General-purpose LLMs may possess broad knowledge but often lack the deep understanding of science needed to solve research problems. To enhance the CRISPR-GPT agent\u2019s capacity in answering advanced research questions, we build a Q&A mode that synthesizes information from multiple resources, including published literature, established protocols and discussions between human scientists, utilizing a combination of RAG technique, a fine-tuned specialized model and a general LLM (for which we picked GPT-4o; Methods<\/a>).<\/p>\n To enhance the Q&A mode\u2019s capacity to \u2018think\u2019 like a scientist for problem solving, we sought to train a specialized language model using real scientific discussions among domain experts. The fine-tuned model is used as one of the multiple sources of knowledge for the Q&A mode (Fig. 4e<\/a>). To this end, we collected 11\u2009years of open-forum discussions from a public Google Discussion Group on CRISPR gene-editing, starting from 2013 (Supplementary Note B<\/a>). The discussion group involved a diverse cohort of scientists worldwide. This dataset, comprising ~4,000 discussion threads, was curated into an instructional dataset with over 3,000 question-and-answer pairs (Supplementary Note B<\/a>). Using this dataset, we fine tuned an 8-billion-parameter LLM on the basis of the Llama3-instruct model58<\/a>,59<\/a>,60<\/a>,61<\/a>,62<\/a>,63<\/a>,64<\/a>,65<\/a>. The fine-tuned model, which we call CRISPR-Llama3, has improved abilities in gene-editing questions, outperforms the baseline model on basic questions by a moderate 8% and on real-world research questions by ~20% (Supplementary Figs. 2<\/a> and 3<\/a>). We integrate this fine-tuned LLM into the Q&A mode as a \u2018brainstorming source\u2019, enabling the agent to generate ideas like a human scientist and provide a second opinion for difficult queries (Fig. 4e<\/a>).<\/p>\n To assess the performance of the Q&A mode, we used the Gene-editing bench Q&A test set (Supplementary Note C<\/a>). The test questions encompass basic gene-editing knowledge, experimental troubleshooting, CRISPR application in various biological systems, ethics and safety66<\/a>,67<\/a>. We prompted CRISPR-GPT, GPT-3.5-turbo and GPT-4o to generate responses to test questions. Three human experts scored the answers in a fully blinded setting. The test demonstrated that the Q&A mode outperformed baseline LLMs in accuracy, reasoning and conciseness, with improvement of 12%, 15% and 32%, respectively, versus GPT-4o (Fig. 4f<\/a>). Human evaluators observed that general-purpose LLMs sometimes make factual errors and tend to provide extensive answers not all of which are relevant to the questions. For example, one question was about solving cell growth issues in an experiment where a scientist performed Cas9 editing followed by single-cell sorting using MCF-7 cells. For this question, the Q&A mode provided a concise, accurate summary of potential reasons and actionable solutions. In contrast, GPT-4o responded with a long list of 9 factors\/options, but at least 2 of them were not applicable to MCF-7 cells (Extended Data Fig. 2<\/a>). This and other examples (Extended Data Figs. 3<\/a> and 4<\/a>) showcase the advantage of the CRISPR-GPT Q&A mode. Overall, evaluation results confirmed that the multisource Q&A mode is better at answering advanced research questions about gene editing.<\/p>\n CRISPR-GPT excels in human\u2013AI collaboration<\/p>\n To further evaluate the human user experience of CRISPR-GPT, we assembled a panel of eight gene-editing experts to assess the agents\u2019 performance for end-to-end experiment covering all 22 individual tasks (Supplementary Table 2<\/a> and 3<\/a> demos). The experts were asked to rate their experiences in four dimensions: accuracy, reasoning and action, completeness and conciseness (see Supplementary Note C<\/a> for detailed rubrics). CRISPR-GPT demonstrated improved accuracy and strong capabilities in reasoning and action, whereas general LLMs, such as GPT-4o, often included errors and were prone to hallucination (Fig. 5a,b<\/a>).<\/p>\n Fig. 5: CRISPR-GPT outperforms general-purpose LLM for gene-editing research in human user experiences. a, Human user experience: evaluation of CRISPR-GPT for end-to-end gene-editing copiloting. Human experts scored performances from 1 (lowest) to 5 (highest). See detailed procedure and rubrics in Supplementary Note C<\/a> (full chat history and video demo provided in Supplementary Table 2<\/a>). b, Human user experience: evaluation results breakdown by major gene-editing tasks. Data shown are the mean\u2009\u00b1\u2009s.d. (n\u2009=\u20098 per group). c, User observations on the strengths and limitations of CRISPR-GPT compared to baseline LLMs.<\/p>\n Highlighted by human evaluators\u2019 observations (Fig. 5c<\/a>), the CRISPR-GPT agent provides users with more accurate, concise and well-rounded instructions to execute the planned experiments. The ability of CRISPR-GPT to perform specialty gene-editing tasks, such as exon-selected gRNA design, customized off-target prediction and automated sequencing data analysis, reinforced its advantage versus general-purpose LLMs. This is confirmed by the task-specific evaluation results (Fig. 5b<\/a>). Despite its strengths, CRISPR-GPT struggled with complex requests and rare biological cases, highlighting areas for improvement (limitations in Supplementary Note D<\/a>).<\/p>\n Real-world demonstration for fully AI-guided gene editing<\/p>\n To showcase and validate CRISPR-GPT\u2019s ability as an AI co-pilot to general biologists, we enlisted two junior researchers unfamiliar with gene editing. They used CRISPR-GPT in two real-world experiments: (1) designing and conducting a multigene knockout and (2) epigenetic editing, from scratch.<\/p>\n In the first experiment, the junior researcher conducted gene knockouts in the human A549 lung adenocarcinoma cell line, targeting four genes involved in tumour survival and metastasis: TGF\u03b2R1, SNAI1, BAX and BCL2L1 (Fig. 6<\/a>). The experiment was designed from scratch with CRISPR-GPT (Fig. 6a<\/a>). On the basis of the user\u2013AI interaction, enAsCas12a was selected for its multitarget editing capability and low off-target effects. For delivery, CRISPR-GPT recommended lentiviral transduction for stable Cas and gRNA expression. The gRNAs for the four target genes were designed through CRISPR-GPT. Furthermore, CRISPR-GPT provided step-by-step protocols for gRNA cloning, lentivirus production and viral delivery into A549 cells. To validate the editing, the researcher followed CRISPR-GPT\u2019s next-generation sequencing (NGS) protocol, using assay primers designed via the integrated Primer3 tool. After generating the NGS data, the raw sequencing files were uploaded into CRISPR-GPT for automated analysis through the CRISPResso2 pipeline. The analysis reports, sent directly via email, summarized the editing outcomes and showed consistently ~80% high efficiency across all target genes (Fig. 6b<\/a>, Supplementary Demo Video 3<\/a>, user interactions summarized in Supplementary Table 4<\/a>, full chat history listed in Supplementary Table 2<\/a>). To further assess the biological phenotypes of TGF\u03b2R1 and SNAI1 knockout in A549 cells, the researcher conducted an epithelial\u2013mesenchymal transition (EMT) induction experiment by treating A549 cells with TGF\u03b2 (Fig. 6c<\/a> and Methods<\/a>). The qPCR results revealed that the knockout A549 cell lines (A549 TGF\u03b2R1 KO and A549 SNAI1 KO) showed up to 9-fold reduction in CDH1 expression change and up to 34-fold reduction in VIM expression change, which are both key marker genes in the EMT process. This confirms the biological role of TGF\u03b2R1 and SNAI1 signalling in driving EMT progression (a crucial driver of metastasis) in lung cancer cells (Fig. 6d<\/a>).<\/p>\n Fig. 6: Wet-lab demonstrations of CRISPR-GPT in knockout and activation experiments. a, The full workflow of CRISPR-GPT-guided knockout experiment of TGF\u03b2R1, SNAI1, BAX1 and BCL2L1 through multiple rounds of human\u2013AI interaction (TGF\u03b2R1 knockout is shown as an example, see Supplementary Demo Video 3<\/a> and full chat history listed in Supplementary Table 2<\/a>). Panel was originally created with BioRender.com\/iim1m2t<\/a> and processed. b, Editing efficiencies for TGF\u03b2R1, SNAI1, BAX1 and BCL2L1 measured via NGS, analysed using CRISPResso2 and CRISPR-GPT. c, Schematic of the EMT induction process via TGF\u03b2 treatment (Methods<\/a>). Panel was originally created with BioRender.com\/iim1m2t<\/a> and processed. d, Functional outcomes of TGF\u03b2R1 and SNAI1 knockout in A549 cells after EMT induction by TGF\u03b2. qPCR analysis shows reduced expression changes in EMT marker genes (CDH1, VIM) in A549 TGF\u03b2R1 and SNAI1 knockout cells compared with A549 wildtype (WT) cells, confirming successful knockouts of TGF\u03b2R1\/SNAI1. Data shown are the mean\u2009\u00b1\u2009s.d. (n\u2009=\u20093 per group). One-way analysis of variance (ANOVA) was used to calculate statistics. ****p\u2009<\u20090.0001. e, Simplified workflow of a beginner researcher activating NCR3LG1 and CEACAM1 expression through multiround interactions with CRISPR-GPT (full chat history listed in Supplementary Table 2<\/a>). f, Editing outcomes of NCR3LG1 and CEACAM1 activation using CRISPR-GPT-designed sgRNAs, measured via flow cytometry (Methods<\/a>).<\/p>\n In the second experiment, the junior researcher performed epigenetic editing to activate two genes involved in cancer immunotherapy resistance in a human melanoma model cell line (Fig. 6e<\/a>, user interactions summarized in Supplementary Table 4<\/a>, full chat history listed in Supplementary Table 2<\/a>). CRISPR-GPT guides the researcher through the full workflow: identifying the most suitable CRISPR activation system, selecting an appropriate delivery method for A375 cells, designing dCas9 gRNAs (three gRNAs per gene) and generating protocols for validating editing outcomes. After editing was completed, measurements of target protein expression level confirmed successful activation of both genes, with up to 56.5% efficiency for NCR3LG1, and 90.2% efficiency for CEACAM1, when comparing gRNA-edited groups vs negative control gRNAs (Fig. 6f<\/a>).<\/p>\n Overall, CRISPR-GPT enabled successful completion of the first set of AI-guided gene-editing experiments. Interactions between the researchers and LLM-powered agents led to efficient, accurate and ethically mindful gene-editing on the first attempt, even by users new to the technique.<\/p>\n","protected":false},"excerpt":{"rendered":"Building AI co-pilot harnessing LLM\u2019s reasoning ability CRISPR-GPT supports four major gene-editing modalities and 22 gene-editing experiment tasks…\n","protected":false},"author":2,"featured_media":63395,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[25],"tags":[18924,3251,23203,11958,19314,3250,916,90,56,54,55],"class_list":{"0":"post-63394","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-genetics","8":"tag-biomedical-engineering-biotechnology","9":"tag-biomedicine","10":"tag-computational-biology-and-bioinformatics","11":"tag-computational-science","12":"tag-crispr-cas-systems","13":"tag-general","14":"tag-genetics","15":"tag-science","16":"tag-uk","17":"tag-united-kingdom","18":"tag-unitedkingdom"},"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts\/63394","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/comments?post=63394"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts\/63394\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/media\/63395"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/media?parent=63394"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/categories?post=63394"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/tags?post=63394"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}![\"figure](\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2025\/08\/41551_2025_1463_Fig3_HTML.png\")
<\/a><\/p>\n![\"figure](\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2025\/08\/41551_2025_1463_Fig4_HTML.png\")
<\/a><\/p>\n![\"figure](\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2025\/08\/41551_2025_1463_Fig5_HTML.png\")
<\/a><\/p>\n![\"figure](\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2025\/08\/41551_2025_1463_Fig6_HTML.png\")
<\/a><\/p>\n