Task processing method for target domain, task processing method for bioinformatics domain, task system for target domain, computing device, computer-readable storage medium, and computer program product

Abstract

Provided in the embodiments of the present disclosure are a task processing method for a target domain, a task processing method for a bioinformatics domain, a task system for a target domain, a computing device, a computer-readable storage medium, and a computer program product. The task processing method for a target domain is applied to the task system for a target domain, and comprises: in response to a target task of a target domain, retrieving, from a task library corresponding to the target domain, a domain task unit related to the target task, wherein the task library is obtained by means of performing construction by means of a target task model and on the basis of a sample task in at least one material file of the target domain and a sample task unit called for executing the sample task; on the basis of the domain task unit, generating meta task information; and on the basis of the meta task information, executing a meta task, so as to obtain a task execution result. Thus, the extendibility of task processing is improved, frequent iterative updates are avoided, costs are reduced, and the task processing efficiency is improved.

Description

[0001] TECHNICAL FIELD: Embodiments of the present disclosure relate to the technical field of deep learning, and more particularly to a task processing method in a target domain, a task processing method in the bioinformatics field, a task system in a target domain, a computing device, a computer-readable storage medium, and a computer program product. BACKGROUND: With the rapid growth of data in various fields, massive amounts of data have posed new challenges to domain task processing. Effectively utilizing this data has become crucial for domain task processing. Currently, target tasks in target domains are exhaustively enumerated based on rules, and task systems with multiple operation levels are designed. For example, task systems in the bioinformatics field include "Plants / Animals / Microorganisms / Viruses - Genetic Genes / Non-Genetic Genes - Gene Annotation / Gene Transcription / Gene Expression / Gene Analysis / Gene Compilation - Promoter Structure / Transcription Factors and Cis-Acting." Front-end users interactively select the corresponding task tool layer by layer to execute the target task. For example, CRISPR-Cas9 can be selected as the gene editing tool to perform editing operations on non-genetic plant genes. However, this traditional software engineering approach suffers from low scalability, requires frequent iterative updates, is costly, and suffers from low task processing efficiency. Therefore, there is an urgent need for a highly scalable, low-cost, and highly efficient task processing method for a target domain. In view of this, embodiments of the present disclosure provide a task system for a target domain. One or more embodiments of the present disclosure also relate to a task system for the bioinformatics field, a task processing method for a target domain, and a computing device to address the technical deficiencies of low scalability, high cost, and low task processing efficiency in the prior art. According to a first aspect of an embodiment of the present disclosure, a task processing method for a target domain is provided, which is applied to a task system for the target domain and includes: In response to a target task for the target domain, retrieving domain task units related to the target task from a task library corresponding to the target domain, wherein the task library is constructed using a target task model based on sample tasks in at least one source file for the target domain and sample task units invoked by executing the sample task; generating meta-task information based on the domain task units; and executing the meta-task based on the meta-task information to obtain a task execution result.According to a second aspect of an embodiment of the present disclosure, a task processing method in the field of bioinformatics is provided. The method is applied to a task system in the field of bioinformatics, comprising: in response to a bioinformatics task in the field of bioinformatics, retrieving domain task units related to the bioinformatics task from a task library corresponding to the field of bioinformatics, wherein the task library is constructed using a target task model based on sample tasks in at least one source file in the field of bioinformatics and sample task units invoked by executing the sample task; generating meta-task information based on the domain task units; and executing the meta-task based on the meta-task information to obtain a task execution result. According to a third aspect of an embodiment of the present disclosure, a task system in the target field is provided. The task system comprises a model component and a response component. The model component is configured to construct a task library corresponding to the target field based on sample tasks in the target field and sample task units invoked by executing the sample tasks using the target task model; the response component is configured to, in response to a target task in the target field, retrieve domain task units related to the target task from the task library and generate meta-task information based on the domain task units; and executing the meta-task based on the meta-task information to obtain a task execution result. According to a fourth aspect of an embodiment of the present disclosure, a computing device is provided, comprising: a memory and a processor; the memory is configured to store a computer program / instructions, and the processor is configured to execute the computer program / instructions, wherein the computer program / instructions, when executed by the processor, implement the steps of the aforementioned method. According to a fifth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, storing a computer program / instructions, wherein the computer program / instructions, when executed by the processor, implement the steps of the aforementioned method. According to a sixth aspect of an embodiment of the present disclosure, a computer program product is provided, comprising a computer program / instructions, wherein the computer program / instructions, when executed by the processor, implement the steps of the aforementioned method. One embodiment of the present disclosure provides a method for processing tasks in a target domain. Using a target task model, a task library corresponding to the target domain is constructed based on sample tasks in at least one source file of the target domain and sample task units invoked when executing the sample tasks. This method achieves more comprehensive coverage of tasks and meta-tasks in the target domain, improves the scalability of task processing, avoids frequent iterative updates, and reduces costs. In response to the task request of the target task, the domain task units related to the target task are retrieved from the task library corresponding to the target domain, and meta-task information is generated based on the domain task units. The meta-task is directly executed based on the meta-task information to obtain the task execution result, thereby improving the efficiency of task processing.BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a flowchart of a task processing method in a target domain provided by an embodiment of the present disclosure; Figure 2 is a flowchart of a task processing method in the bioinformatics field provided by an embodiment of the present disclosure; Figure 3 is a schematic diagram of the structure of a task system in a target domain provided by an embodiment of the present disclosure; Figure 4 is a schematic diagram of the structure of a task system in another target domain provided by an embodiment of the present disclosure; Figure 5 is a schematic diagram of the structure of a task system in another target domain provided by an embodiment of the present disclosure; Figure 6 is an architectural diagram of a task system in a target domain provided by an embodiment of the present disclosure; Figure 7 is a flowchart of the processing process of a task processing method in the bioinformatics field provided by an embodiment of the present disclosure; Figure 8 is a schematic diagram of the front-end of a task processing method in the bioinformatics field provided by an embodiment of the present disclosure; and Figure 9 is a block diagram of the structure of a computing device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION The following description sets forth numerous specific details to facilitate a thorough understanding of the present disclosure. However, the present disclosure can be implemented in many other ways than those described herein, and those skilled in the art may make similar generalizations without departing from the scope of the present disclosure. Therefore, the present disclosure is not limited to the specific implementations disclosed below. The terminology used in one or more embodiments of the present disclosure is for the purpose of describing specific embodiments only and is not intended to limit the present disclosure. As used in one or more embodiments of the present disclosure and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in one or more embodiments of the present disclosure refers to and encompasses any and all possible combinations of one or more of the associated listed items. It should be understood that while the terms "first," "second," and so on may be employed in one or more embodiments of the present disclosure to describe various information, such information should not be limited to these terms. These terms are merely used to distinguish information of the same type from one another. For example, the first could be referred to as the second, and similarly, the second could be referred to as the first, without departing from the scope of one or more embodiments of the present disclosure. Depending on the context, the word "if," as used herein, could be interpreted as "when," "at the time," or "in response to determining."Furthermore, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, storage, and display) involved in one or more embodiments of this disclosure are all authorized by the user or fully authorized by all parties. The collection, use, and processing of the relevant data must comply with the relevant laws, regulations, and standards of the relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or deny. In one or more embodiments of this disclosure, a large model refers to a deep learning model with large model parameters, typically containing hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than ten trillion model parameters. Large models, also known as foundation models, are pre-trained using large unlabeled corpora to produce pre-trained models with parameters exceeding 100 million. Such models are adaptable to a wide range of downstream tasks and have good generalization capabilities. Examples include large language models (LLMs) and multi-modal pre-training models. In practical applications, large models only require a small number of samples to fine-tune the pre-trained model and can be applied to various tasks. Large models can be widely used in fields such as natural language processing (NLP) and computer vision. Specifically, they can be applied to computer vision tasks such as visual question answering (VQA), image captioning (IC), and image generation, as well as natural language processing tasks such as text-based sentiment classification, text summarization, and machine translation. Key application scenarios for large models include digital assistants, intelligent robots, search, online education, office software, e-commerce, and intelligent design. First, let's explain the terms used in one or more embodiments of this disclosure. Artificial Intelligence Agent (AI Agent): refers to a software or hardware entity that can operate autonomously in its environment to achieve a given goal or perform a specific task. Large Language Model Agent: An AI agent that leverages the capabilities of a large language model to perform specific tasks. Bioinformatics system: An intelligent system built on a large language model, specifically used in the field of bioinformatics, capable of utilizing and managing the knowledge graph of task tools in the field of bioinformatics to complete specific bioinformatics tasks.A tool knowledge graph in the bioinformatics field: A structured knowledge system that contains the correspondence between bioinformatics tasks and corresponding tools, designed to help bioinformatics systems quickly retrieve and generate feasible tool meta-tasks. A tool meta-task: A series of bioinformatics tools organized in a specific order, used to perform complex bioinformatics analysis tasks. Retrieval-Augmented Generation (RAG): A technology that combines retrieval mechanisms and generative models to improve the model's ability to handle complex tasks by retrieving relevant information to assist in the generation process. This disclosure provides a task system in a target domain. This disclosure also relates to a task system in the bioinformatics field, a task processing method in the target domain, and a computing device, each of which is described in detail in the following embodiments. Referring to Figure 1, a flowchart of a target domain task processing method provided by one embodiment of the present disclosure is shown. The task system, applied to the target domain, includes the following specific steps: Step 102: In response to a target task in the target domain, domain task units related to the target task are retrieved from a task library corresponding to the target domain. The task library is constructed using a target task model based on sample tasks in at least one source file in the target domain and the sample task units invoked when executing the sample tasks. The target domain task system is an artificial intelligence agent system designed and constructed for a specific knowledge domain to handle complex tasks within that domain. For example, in the field of bioinformatics, the task system can be used to automatically complete a series of complex operations involving gene expression differential analysis, data acquisition, data analysis, and visualization. The target domain is the knowledge domain within which the target task to be processed resides. The target domain, to a certain extent, defines the task library and the domain task units to be retrieved and invoked. For example, the target domain is the field of bioinformatics. The target task is a pending domain task within the target domain. It must be executed through domain task units according to a specific process. This process can be divided into multiple subtasks, each of which can be understood as one or more process steps. For example, if the target domain is bioinformatics, the target task is to explore the expression differences of specific gene markers across different cell types (such as neurons and hepatocytes). This includes multiple subtasks, such as gene tagging, sampling, threshold determination, gene screening, enrichment analysis, and visualization. A metatask decomposes the target task into at least one executable, interconnected, and functionally distinct basic operation unit. Each metatask represents an independent step or process stage within the target task and is typically associated with a specific domain task unit.Metatasks are a key concept for modularizing and structuring complex tasks. They help break down complex processes into logically simpler parts and ensure that each part can be executed efficiently in a predetermined order. This metatask division enhances the scalability and automation of task system 100. For example, in the bioinformatics field, a target task is to "explore the differential expression of specific gene markers between different cell types." This can be divided into the following three metatasks: Data Acquisition Metatask: This metatask calls the NCBI GEO Data Retrieval API to retrieve research data for a specified gene in different cell types. Parameters such as the gene list, cell type, and time range must be set. The output is a gene expression data file called "expressiondata.json." oThe differential expression analysis meta-task calls the DESeq2 analysis tool to perform differential expression analysis on the data obtained in the first step, setting the input file to "data-expression.json" generated in the first step. It compares gene expression between neurons and hepatocytes and outputs the differential expression analysis results file, "DESeq2_results.csv." The visualization meta-task calls the Cytoscape visualization plug-in to create a network diagram based on the differential expression results obtained in the second step, setting the input data source to "DESeq2_results.csv," and saves the final visualization of gene expression differences as the image file "expression_network.png." The task library corresponding to the target domain is a systematic storage structure built for the target domain. It records task information for sample tasks and unit information for the sample task units called to execute the sample tasks, based on the correspondence between tasks and task units. The task library for the target domain includes, but is not limited to, structured data such as text or triples, key-value pairs, knowledge graphs, and lists. For example, a task library might be a knowledge graph in the bioinformatics domain, recording a complete gene expression data analysis task. This includes task information for a sample task, such as comparing gene expression differences between different cell types, and unit information for sample task units, including the target domain's task library data retrieval API (for acquiring research data), data analysis tools (for performing differential expression analysis), and visualization plugins (for generating a visual result network). Domain task units are component units within the target domain that can execute the target task. There must be at least one domain task unit, corresponding to each subtask within the target task. Domain task units can be software tools, algorithms, or solutions, such as gene tagging tools, sampling algorithms, gene screening tools, and enrichment analysis tools. Sample tasks are subtasks within the target domain, represented by one or more process steps. The sample task has been executed using the corresponding sample task unit. For example, the sample task is analyzing the expression changes of key genes at various stages of mouse embryonic development. A sample task unit is a component unit used to execute a sample task. It can be a software tool, an algorithm, or a solution. For example, a sample task unit utilizes RNA-seq data. The target task model is a deep learning model with the ability to build a task library. This target task model is pre-trained to understand the execution process of sample tasks in source files. It extracts sample tasks and the sample task units invoked to execute the sample tasks from at least one source file, and constructs a systematic storage structure called a task library.Target task models include, but are not limited to, Transformer models, BERT models, and large models. The target task model can be a specific model for the target domain, pre-trained using training data from the target domain (e.g., fine-tuning, transfer learning, reinforcement learning), or a general model for different domains, implemented based on prompts from a task library. This is not limited here. In response to the target task in the target domain, domain task units related to the target task are retrieved from the task library corresponding to the target domain. One optional approach is to retrieve domain task units related to the target task from the task library corresponding to the target domain based on the task information of the target task. The task information of the target task is content information related to the target task, including, but not limited to, the task name, task requirements, task data information, task execution process, and task execution constraints. For example, if the target task is to explore the differential expression of specific gene markers across different cell types (e.g., neurons and hepatocytes), the task information includes: task name (Explore the differential expression of specific gene markers across different cell types); a list of cell types (e.g., neurons, hepatocytes, cardiomyocytes, etc.); a list of gene markers (such as specific transcription factors or disease-related genes); analysis requirements (whether statistical significance testing, visualization of results, benchmarks, etc. are required); and data source restrictions (database name, time range, literature screening criteria, etc.). For example, based on the bioinformatics task name (Explore the differential expression of specific gene markers across different cell types), domain bioinformatics tools related to the task are retrieved from the corresponding relationships between multiple sample bioinformatics tasks and sample bioinformatics tools in the bioinformatics domain knowledge graph: the NCBI GEO data retrieval API, the DESeq2 analysis tool, and the Cytoscape visualization plugin. Using the target task model, a task library corresponding to the target domain is constructed based on sample tasks in at least one source file in the target domain and the sample task units invoked to execute the sample tasks. This achieves more comprehensive coverage of tasks and meta-tasks in the target domain, improves the scalability of task processing, avoids frequent iterative updates, and reduces costs. Step 104: Generate meta-task information based on the domain task units. Meta-task information records the flow of each meta-task in the target task and the process logic of the domain task units that execute each meta-task. It represents the flow and dependencies of at least one meta-task of the target task. Meta-task information has a specific format and flow. For example, meta-task information is in JSON format, specifically:

[0002] {

[0003] "task": "Explore the expression differences of specific gene markers in different cell types between different cells", "The target task defines the goal of the entire task, that is, to compare the expression differences of specified genes in different cell types

[0004] "atomic task" : [ / / Meta-task list, which describes in detail the series of operations required to complete the target task in the order of execution

[0005] {

[0006] "unit" : "NCBI GEO Data Retrieval API", / / The tool or service used by the first meta-task. In this case, it is the API interface for obtaining relevant research data from the NCBI GEO database.

[0007] "params": ( / / Parameter set, which is the specific parameters that need to be passed when calling the API

[0008] "genejist" : ["GENE1", "GENE2"], / / Specify the gene list to be studied

[0009] "cell_types": ["neurons", "hepatocytes"], "list the cell types to be compared

[0010] "time_range" : "2010-2022" / / Time range limit for research data

[0011] },

[0012] "output" : "expression_data.json" / / Expected output file name, storing the retrieved gene expression data results

[0013] "unit": "DESeq2 analysis tool", "analysis tool used in the second meta-task, in this case, the R package DESeq2 for differential expression analysis "params": (

[0014] "input file" : "expression_data.json", / / Input file name, the data generated in the first meta-task is used as analysis input

[0015] "comparison groups" : [["neurons["liver cells"]]〃 Comparison group settings, clearly indicate which cell types need to be compared in terms of gene expression

[0016] },

[0017] "output" : "DESeq2_results.csv" / / Expected output file name, containing the differential expression results after DESeq2 analysis

[0018] "unit": "Cytoscape visualization plug-in", "The visualization tool used by the third-party task, in this case, it is a network visualization plug-in based on the Cytoscape platform

[0019] "params" : (

[0020] "data source" : "DESeq2_results.csv", / / Input data source, using the differential expression analysis results generated in the second meta-task

[0021] "visualization type" : "network" / / Visualization type, specifies to display the results in the form of a network diagram

[0022] },

[0023] "output" : "expression_network.png" / / Expected output file name, the final generated gene expression difference visualization network image

[0024] }

[0025] ]

[0026] Meta-task information is generated based on domain task units. One optional method is to generate meta-task information based on domain task units using the target task model. For example, the tool information of a bioinformatics tool is input into a universal large language model to generate meta-task information in JSON format corresponding to the bioinformatics task. Exemplarily, the unit information of the domain bioinformatics tool is input into the universal large language model, and, prompted by the task information of the bioinformatics task, meta-task information in JSON format that conforms to the bioinformatics task execution logic is generated. In response to the task request of the target task, the domain task units related to the target task are retrieved from the task library corresponding to the target domain, and meta-task information is generated based on the domain task units, providing an execution basis for subsequent tasks. Step 106: Execute the meta-task based on the meta-task information to obtain a task execution result. The task execution result is the task result obtained by calling each domain task unit to execute the target task according to the process logic corresponding to the call chain information. The task execution result is a summary and integration of the results obtained during the execution of the target task and can be presented in various forms such as data sets, reports, charts, and analytical conclusions. For example, in the bioinformatics field, for example, after executing each subtask according to the call chain information, the task execution results include: Data Analysis Report: A detailed document outlining the differential expression analysis process, key findings, and statistical significance test results for specific gene markers across different cell types; Data Table: The task execution results, based on the process logic corresponding to the call chain information, record key data such as the expression level, fold change, P-value, and adjusted P-value (if using FDR correction) for each gene across different cell types; Visualization Chart: A network diagram that displays the interactions and expression differences between genes, visually illustrating the differences in gene expression patterns across different cell types; Raw Data and Intermediate Results: This may also include the research data file "expression_data.json" obtained from the NCBI GEO Data Retrieval API and any other intermediate computational result files. An optional method for executing a meta-task based on the meta-task information to obtain the task execution results is to call a domain task unit based on the meta-task information to execute the meta-task and obtain the task execution results.Exemplarily, based on meta-task information in JSON format, the meta-task information is invoked, a data retrieval API is invoked to execute the meta-task of acquiring research data, the DESeq2 analysis tool is invoked to execute the meta-task of differential expression analysis, and the differential expression analysis plug-in is invoked to execute the meta-task of visualizing the result network, thereby obtaining visualization results of the bioinformatics task. In this embodiment of the present disclosure, a target task model is used to construct a task library corresponding to the target domain based on sample tasks in at least one source file of the target domain and the sample task units invoked when executing the sample tasks. This achieves more comprehensive coverage of tasks and meta-tasks in the target domain, improves the scalability of task processing, avoids frequent iterative updates, and reduces costs. In response to a task request for the target task, domain task units related to the target task are retrieved from the task library corresponding to the target domain, meta-task information is generated based on the domain task units, and the meta-task is directly executed based on the meta-task information to obtain task execution results, thereby improving task processing efficiency. In an optional embodiment of the present disclosure, prior to step 102, the following specific step is further included: obtaining sample information for the target domain from a related database of the target domain, wherein the sample information includes sample tasks and sample task units invoked when executing the sample tasks. A linked database in a target domain is a database designed specifically for the target domain, storing sample information within that domain. Linked databases are typically tightly integrated with task systems, providing real-time or batch access to required data resources to support task processing within the target domain. For example, in the bioinformatics field, linked databases include public gene expression databases such as NCBI GEO and SRA, literature databases, and other specialized databases. Sample information in a target domain is representative sample information within the target domain, including source files, domain data, and related metadata. It includes actual case studies or research content within the target domain, as well as input data and tool usage records required to perform related tasks. For example, in the bioinformatics field, sample information might include a research paper detailing how to obtain gene expression data from the NCBI GEO database, perform differential analysis using the DESeq2 analysis tool, and visualize it using Cytoscape. Furthermore, it includes the actual gene expression dataset used in the study (raw or processed), as well as detailed information on how to use various software tools and APIs to complete the data analysis process, including the specific steps and parameter settings.For example, sample information in the field of bioinformatics is obtained from public gene expression databases, literature databases, or other specialized databases. This sample information includes a research paper detailing the use of the NCBI GEO database to obtain gene expression data, differential analysis using the DESeq2 analysis tool, and visualization using Cytoscape. Furthermore, the sample information also includes the gene expression datasets (raw or processed) actually used in the research, as well as detailed information such as the specific steps and parameter settings for invoking various software tools and APIs to complete the data analysis process. In this embodiment of the present disclosure, sample information, including sample tasks and sample task units, is obtained from a relational database to provide data support for the subsequent construction of a task library. In an optional embodiment of the present disclosure, after obtaining sample information in a target domain from a relational database for the target domain, the following specific step is further included: converting the sample information to obtain sample information in a target format. The target format is any predefined specific data format. For different application scenarios and requirements, domain data needs to be converted to the target format to achieve compatibility with other components. For example, in the field of bioinformatics, the target format is Anndata, a standard format for single-cell transcriptomics data analysis. It integrates gene expression matrices, metadata, and various observed experimental properties, making the data easier to manage and manipulate during subsequent analysis. One option for converting sample information to obtain sample information in the target format is to convert the sample information according to a preset domain data definition. Furthermore, a format conversion model can be used to convert sample information according to a preset domain data definition to obtain sample information in the target format. For example, in single-cell transcriptomics research in the bioinformatics field, domain data is typically stored in CSV format. Using a format conversion model, a data structure compliant with the Anndata standard is generated. This data structure not only contains normalized and filtered gene expression values ​​but also integrates cell clustering, marker genes, and other experimental annotation information, resulting in sample information in the target format suitable for single-cell transcriptomics analysis. For example, using the universal large model, domain data in CSV format is converted according to a preset domain data definition to obtain domain data in the Anndata standard. In the disclosed embodiment, sample information is converted to obtain sample information in the target format, ensuring the feasibility of subsequent processing.In an optional embodiment of the present disclosure, after obtaining sample information in the target domain from a related database in the target domain, the following specific steps are further included: analyzing the sample information using a data analysis model based on a preset data analysis strategy to obtain analysis results; and determining updated sample information based on the analysis results and the sample information. In related art, the representation of long-tail (newly discovered) sample information is not ideal, often resulting in computational results that are inconsistent with the facts and unusable. Therefore, it is necessary to analyze the sample information based on a preset data analysis strategy. A preset data analysis strategy is a predefined combination of data processing steps, algorithms, or methods used to perform specific analysis and interpretation of sample information. Data analysis strategies are typically defined based on field practices and research, aiming to extract meaningful information from raw data to support subsequent meta-task generation and execution. For example, in single-cell transcriptomics research, preset data analysis strategies may include, but are not limited to, data quality control (filtering low-quality cells and genes), normalization, dimensionality reduction and visualization, cluster analysis, and differentially expressed gene detection. Analysis results are specific findings or conclusions derived after processing using a predefined data analysis strategy. These can be data products in the form of numerical values, charts, or reports, reflecting key features or patterns within the sample information. For example, in the field of bioinformatics, analysis results might be a CSV file containing a list of differentially expressed genes, a PDF file showing the distribution of cell populations, or a lab report summarizing the statistical significance and biological interpretation. Updated sample information is a new, richer, or more structured dataset generated after preprocessing and data analysis. It integrates the original sample information with the target format sample information obtained through conversion, and further includes the analysis results obtained using the predefined data analysis strategy. For example, in the case of single-cell transcriptomics, after the raw data is converted into an AnnData object and a series of data analyses are performed, the updated sample information contains not only the original gene expression matrix and cell metadata, but also includes information such as cell cluster labels, marker gene information, and statistically significant biological findings. This updated sample information can directly serve as the basis for further research or decision-making.For example, based on a preset data analysis strategy, the sample information is analyzed to obtain analysis results, which are stored as the GSM2230757.pdf file. The four processed files are packaged as updated sample data, including: sample information metadata: GSM2230757.metadata; sample information: GSM2230757.h5ad; and analysis results: GSM2230757.pdf. The data analysis model performs data analysis, enabling autonomous identification and adjustment of long-tail (newly discovered) sample information. In the disclosed embodiments, the data analysis model analyzes the sample information based on a preset data analysis strategy to obtain analysis results. Based on the analysis results and the sample information, updated sample information is determined, further ensuring the feasibility of subsequent processing. In an optional embodiment of the present disclosure, after obtaining sample information in the target domain from a related database in the target domain, the following specific steps are further included: feeding the sample information back to the client; receiving the client's evaluation results of the sample information; determining an adjustment strategy corresponding to the evaluation results based on the evaluation results, and adjusting the sample information based on the adjustment strategy. The evaluation results are the conclusions drawn by the user after evaluating the sample information, including but not limited to: data quality scoring, confirmation of the effectiveness of the analysis method, authenticity judgment of newly discovered data points, and consistency evaluation of the data analysis results with domain knowledge. For example, in the field of bioinformatics, after completing the cleaning, normalization, and preliminary analysis of single-cell transcriptome data and presenting the results to the client in the form of a visual report, the evaluation results include: based on the cell clustering and differential gene expression analysis in the report, the user scores and comments on the rationality of the clustering, biological significance, and credibility of the differential gene list. The user can also make modifications to the analysis process or parameter settings based on their own professional knowledge, such as suggesting that certain statistical thresholds need to be adjusted to obtain more reliable results. Users compare experimental designs with existing literature and provide feedback, either approving or questioning, on newly discovered phenomena (such as specific cell subpopulations or marker genes). Adjustment strategies based on the user or expert response to sample information evaluation results are used to guide and improve decision-making strategies for data processing, analysis method selection, and parameter settings, thereby enhancing the accuracy and applicability of data analysis results. Adjustment strategies based on evaluation results include, but are not limited to: Data preprocessing optimization: Renormalize the data and try different normalization methods, such as Log-Normalization or SCTransform, to improve data distribution and clustering.Clustering parameter adjustment: Lower or increase the resolution parameter in clustering algorithms (such as the Leiden algorithm) to find the optimal number and structure of cell clusters. Feature selection strategy changes: Based on domain knowledge or new statistical metrics, select more discriminatory gene markers for cluster analysis. Analysis process adjustments: Consider incorporating additional hierarchical clustering or multi-perspective clustering methods to enhance the robustness and accuracy of the analysis. For example, sample information is fed back to the client. Based on the cell clustering and differential gene expression analysis reported, the user scores and comments on the rationality, biological significance, and credibility of the differential gene list. The client receives the scores and comments, determines a corresponding adjustment strategy based on the scores and comments, and adjusts the sample information based on the adjustment strategy. In the disclosed embodiments, sample information adjustment is accomplished through front-end interaction, further ensuring the feasibility of subsequent processing. In an optional embodiment of the present disclosure, the following specific steps are further included before step 102: using an information extraction model, extracting sample tasks and sample task units invoked by executing the sample tasks from the sample information in the target domain; constructing task nodes for the sample tasks and unit nodes for the sample task units, and determining the correspondence between the task nodes and unit nodes; and constructing a task library corresponding to the target domain based on the task nodes, unit nodes, and correspondence. The information extraction model is a deep learning model with information extraction capabilities. The information extraction model is pre-trained and can identify sample tasks and sample task units invoked by executing the sample tasks in the sample information, and extract the sample tasks and sample task units from them. Information extraction models include, but are not limited to, Transformer models, BERT models, and large models. The information extraction model can be a target domain-specific model, pre-trained using training data from the target domain (e.g., fine-tuning, transfer learning, reinforcement learning), or a general model across different domains, implemented based on extraction prompts. This is not a limitation here. The task node of a sample task is a structured data node for the sample task in the task library. This node contains the task information of the sample task and serves as an index point for the sample task in the task library to facilitate retrieval of domain task units. This includes, but is not limited to, metadata such as the task name, task requirements, task data, task execution process, and task execution constraints. For example, in a knowledge graph, a task node represents a sample task.The unit node of a sample task unit is a structured data node representing the sample task unit in the task library. The unit node contains the task unit information of the sample task unit and serves as the search result for domain task units in the task library. This information includes, but is not limited to, metadata such as the task unit name, task unit parameters, task unit type, and task unit call information. For example, in a knowledge graph, a task node represents a sample task unit. The correspondence between a task node and a unit node is a mapping between the sample task in the task library and the sample task unit that calls the sample task. It demonstrates how a specific sample task completes its meta-task by calling the corresponding sample task unit. For example, in the bioinformatics knowledge graph, the task node of the "differential expression analysis" sample task is connected to the unit nodes of three analysis tools via one or more edges. This indicates that when executing this sample task, the task units represented by these three unit nodes can be used to perform the data analysis meta-task. The corresponding relationship is represented by the edges connecting the nodes. Based on task nodes, unit nodes, and corresponding relationships, a task library corresponding to the target domain is constructed. One optional approach is to connect task nodes and unit nodes based on the corresponding relationships to construct the task library corresponding to the target domain. For example, multiple papers in the field of bioinformatics are obtained. Sample bioinformatics tasks and sample bioinformatics tools invoked to perform the sample bioinformatics tasks are extracted from the papers. Based on the sample bioinformatics tasks and sample bioinformatics tools, task nodes for the sample bioinformatics tasks and unit nodes for the sample bioinformatics tools are constructed. A corresponding relationship between the task nodes and unit nodes is determined. Based on the corresponding relationship, the task nodes and unit nodes are connected to construct a knowledge graph corresponding to the bioinformatics domain. In the disclosed embodiments, an information extraction model is used to extract sample tasks from sample information in the target domain and the sample task units invoked to execute the sample tasks. Task nodes for the sample tasks and unit nodes for the sample task units are constructed, and the corresponding relationships between the task nodes and unit nodes are determined. Based on the task nodes, unit nodes, and corresponding relationships, a comprehensive task library is automatically constructed. This further achieves more comprehensive coverage of tasks and meta-tasks in the target domain, improves the scalability of the task system, avoids frequent iterative updates, and further reduces costs.In an optional embodiment of the present disclosure, the following specific steps are further included before step 102: extracting task execution information from sample information in the target domain using an information extraction model; generating multiple task problems in the target domain based on the task execution information. Accordingly, step 102 includes the following specific steps: receiving a target task in the target domain sent by a client; determining the target task problem based on the task request and feeding the target task problem back to the client; receiving a target task reply to the target task problem from the client; and determining the target task in the target domain based on the target task problem and the target task reply. The information extraction model is a deep learning model with information extraction capabilities. The information extraction model is pre-trained and can identify and extract task execution information from sample information. Information extraction models include, but are not limited to, Transformer models, BERT models, and large models. The information extraction model can be a target domain-specific model, pre-trained (through fine-tuning, transfer learning, reinforcement learning, etc.) with training data from the target domain, or a general model across different domains, implemented based on extraction prompts. This is not limited here. Task execution information describes the execution content of a sample task, including but not limited to steps, parameter settings, tool call records, and task execution results. This information describes the important data and processes involved in the sample task execution process. It provides the necessary basis for constructing task nodes and unit nodes in the task library and facilitates the generation of task-specific responses to various task questions. For example, the following task execution information was extracted from a bioinformatics paper: Task execution steps: First, the NCBI GEO Data Retrieval API is used to query expression data for a specified gene in different cell types; then, the DESeq2 analysis tool is used to perform differential expression analysis on the collected data; finally, the Cytoscape visualization plug-in is used to display the network structure of the differential expression results. Parameter configuration example: When calling the NCBI GEO Data Retrieval API, the parameters used include the gene list to be studied, the cell types to be compared, and the time range. Tool call records: Command line calls or programmatic interface code snippets for tools such as the NCBI GEO Data Retrieval API, the DESeq2 analysis tool, and the Cytoscape visualization plug-in.Example Output: After the task is completed, the result files include "expression data.json" (raw dataset), "DESeq2_results.csv" (differential expression analysis results), and "expression_network.png" (network visualization). Paper Summary: A research paper on gene expression data analysis briefly describes how the authors use a series of tools and techniques to explore the differential expression of specific genes across different cells. This content can be extracted and used to construct relationships between task nodes and cell nodes. The multiple task questions in the target domain are guiding questions set for the target domain. They reflect different aspects of various complex tasks that may need to be solved within the target domain. They are designed to guide front-end users through interactive question-and-answer sessions, thereby determining the target tasks in the target domain. For example, in the field of bioinformatics, multiple task questions include: "Do you need to analyze the expression differences of specific genes between different cell types?", "Which database do you want to use to obtain gene expression data?", and "Do you need to visualize the analysis results?" The target task question is the corresponding task question determined based on the task request. The target task question is formed after the target task model analysis, is related to the user's actual needs, and can help the system accurately identify and clarify the target task in the target field that the user wants to solve. For example, a front-end user sends a task request like, "I want to analyze the differential expression of a certain gene in different cell types." Based on this task request, the target task question determined through large language model analysis is, "Which gene would you like to analyze for differential expression in which specific cell types? Do you have a preferred data source (such as NCBI GEO) or would you like us to automatically retrieve appropriate datasets for you? Furthermore, would you like us to present the final results in the form of visual charts?" The target task response is a specific and unambiguous response from the front-end user to the target task question posed by the system. It contains a detailed description of the user's task requirements and serves as a key basis for the system to further execute and achieve the user's desired target task. For example, the target task response sent by the front-end user for the target task question above is: Focus on the differential expression between neurons and hepatocytes; use the NCBI GEO database to obtain relevant data; and request that the system automatically complete the differential expression analysis and present the analysis results in the form of a network diagram generated by the Cytoscape visualization plug-in.An information extraction model is used to extract task execution information from sample information in the target domain. One optional approach is to encode the sample information in the target domain using the information extraction model to obtain a semantic encoding vector. Based on the semantic encoding vector, the information extraction model is then used to extract task execution information from the sample information. For example, in a bioinformatics paper, a sentence titled "We used the DESeq2 analysis tool to analyze gene expression differences between different cell types" can be obtained after token segmentation: "We / used / the / DESeq2 analysis tool / to / analyze / gene / expression / differences / between / different / cell / types." Through Word2Vec embedding encoding, the sentence "We / used / the / DESeq2 / analysis / tool / to / analyze / gene / expression / differences / between / different / cell / types" is encoded into a dense semantic encoding vector, which captures its semantic meaning within the field of bioinformatics. Using a large language model, the entire paper's content is understood based on the semantic encoding vectors of each sentence in the paper, generating a summary description: "This paper used the DESeq2 analysis tool to analyze gene expression differences between different cell types. Data was first retrieved from the NCBI GE0 database, differential expression was calculated, and the results were finally visualized." In this embodiment of the present disclosure, task mining is performed through an interactive question-and-answer approach, identifying target tasks in the target domain and providing clear task support for the response component to process the task. In an optional embodiment of the present disclosure, the task execution result includes an alarm message. After step 106, the following specific steps are further included: feeding the alarm message back to the client; receiving updated meta-task information sent by the client based on the alarm message; and executing the meta-task based on the updated meta-task information to obtain an updated task execution result. Alarms are system-generated warning notifications when one or more meta-tasks fail to complete successfully due to errors, abnormal conditions, or other unforeseen issues during task execution. These notifications typically include the error type, description, and possible cause analysis. Alarms serve as a crucial feedback mechanism in the task execution process, promptly alerting users or the task system of potential issues requiring further investigation or corrective action. For example, in a bioinformatics gene expression differential analysis task, data quality issues prevent effective analysis, generating the alarm message "Error in data analysis phase: DESeq2 analysis failed." In the disclosed embodiments, updating meta-task information involves regenerating the execution logic and sequence information for executing the target task based on user intervention or automatic correction strategies after receiving the alarm message.When an alarm message appears during the execution of a task guided by metatask information, the user can replace the domain task unit provided in the alarm message, generate new metatask information, and continue executing the metatask. For example, if the metatask information specifies a specific version of an RNA-seq data processing tool, but during execution, it is discovered that the tool does not fully support the current data format, resulting in an alarm, the user can change the tool version or select a more suitable tool. Based on this change, an updated metatask message is generated, replacing the original tool with the latest compatible version and rescheduling the execution order of the metatask. The updated task execution result is the updated task result obtained by executing the metatask according to the updated metatask information flow logic. For example, when executing a gene expression differential analysis task in the bioinformatics domain, the DESeq2 analysis tool, invoked based on the pre-generated metatask information, fails to perform analysis, generating an alarm message: "Error in data analysis phase: DESeq2 analysis failed due to incompliance with DESeq2 analysis tool requirements." This alarm message is quickly fed back to the client interface, allowing the user to promptly understand and address the issue. After receiving the alarm message, the client user identifies the problem and adjusts the meta-task information accordingly, upgrading or switching the existing DESeq2 analysis tool to another differential expression analysis tool that supports the new data format. For example, the version upgrade or switch updates the meta-task information, which includes the analysis tool and its parameter configuration optimized for the current data format. After receiving the updated meta-task information sent by the client based on the alarm message, the execution process is reorganized according to the new execution logic and sequence. Following the updated meta-task information, the data is converted to the format required by the newly selected tool, and the analysis tool is then called to execute the differential expression analysis meta-task. The subsequent steps in the updated meta-task are then followed to complete the meta-task execution. In this disclosed embodiment, through front-end interaction, flexible responses to alarm messages and user intervention are implemented to dynamically complete task processing, effectively resolving issues encountered during task execution and ensuring the feasibility of task processing. In an optional embodiment of the present disclosure, the task execution result includes an alarm message. Following step 106, the following specific steps are further included: generating update meta-task information based on the alarm message and the domain task unit using the information generation model; executing the meta-task based on the updated meta-task information; and obtaining the updated task execution result. In this embodiment of the present disclosure, the updated meta-task information is the execution logic and sequence information for executing the target task, regenerated by the information generation model after receiving the alarm message.When an alarm message appears during the execution of a task guided by metatask information, the information generation model generates new metatask information based on the replacement domain task unit provided in the alarm message and continues executing the metatask. For example, if the metatask information uses a specific version of an RNA-seq data processing tool, but during execution, it is discovered that the tool does not fully support the current data format, resulting in an alarm, the information generation model generates updated metatask information based on the alarm message, replaces the original tool with the latest compatible version, and replans the execution order of the metatasks. For example, when executing a gene expression differential analysis task in the bioinformatics domain, the DESeq2 analysis tool, invoked based on the pre-generated metatask information, fails to perform analysis, generating an alarm message: "Error in data analysis phase: DESeq2 analysis failed due to incompatibility with DESeq2 analysis tool requirements." Based on the alarm message, the information generation model generates updated metatask information, upgrading or switching the original DESeq2 analysis tool to another differential expression analysis tool that supports the new data format. This updated metatask information includes the analysis tool and its parameter configuration optimized for the current data format. After receiving the update meta-task information sent by the front-end user based on the alarm message, the execution process is reorganized according to the new execution logic and sequence. Based on the update meta-task information, the data is converted to the format required by the newly selected tool. The update meta-task analysis tool is then called to execute the differential expression analysis meta-task. Subsequently, the subsequent steps in the update meta-task are followed to complete the meta-task execution. In this embodiment of the present disclosure, the response component dynamically completes task processing by flexibly responding to alarm messages and responding to front-end user intervention through model generation, effectively resolving issues encountered during task execution and ensuring the feasibility of task processing. In an optional embodiment of the present disclosure, after step 106, the following specific steps are further included: generating a meta-task execution flowchart based on the meta-task information; feeding the execution flowchart back to the client; receiving update meta-task information sent by the client in response to the execution flowchart; and executing the meta-task based on the update meta-task information to obtain the task execution result. A meta-task execution flowchart is a graphical visualization of meta-task information, used to intuitively present meta-tasks and their execution process. This includes the domain task units corresponding to each meta-task, their position in the execution process, their contextual relationships, and any conditional branches or loop structures. Constructed using nodes (representing individual meta-tasks) and links (representing the calling sequence or dependencies), the execution flowchart clearly reflects the dynamic connections between the meta-tasks involved in the entire task execution process, from start to finish.For example, in the bioinformatics field of gene expression differential analysis, an execution flowchart can show a series of steps: first, calling the NCBI GEO Data Retrieval API to acquire data, then using the DESeq2 analysis tool for differential analysis, and finally using the Cytoscape plug-in to generate a visual network diagram to represent this series of steps. In this disclosed embodiment, the updated meta-task information is obtained by the front-end user updating the meta-task information based on the execution flowchart, providing new execution logic and sequence information for executing the target task. For example, the meta-task information includes: calling the NCBI GEO Data Retrieval API to acquire public research data containing target genes and corresponding cell types; calling the DESeq2 analysis tool to process the downloaded data to determine the differential expression of each gene between different cells; and calling the Cytoscape plug-in to construct and display a gene co-expression network diagram. Based on this meta-task information, an execution flowchart is automatically generated. This flowchart illustrates the entire task process through a series of nodes (representing meta-tasks) and lines (indicating execution order). After receiving this execution flowchart, the front-end user can clearly see the domain task units used in each stage and their interrelationships. The user decided to change the data source, using the ArrayExpress database API to obtain relevant research data. The subsequent analysis process was adjusted accordingly, adding a preprocessing step before using the DESeq2 analysis tool, employing the R package limma for preliminary data cleaning and standardization. The front-end user sent updated meta-task information based on the updated execution logic and sequence. In this embodiment, by generating and sending an execution flowchart to the front-end user, the execution process is proactively adjusted through front-end interaction, enhancing task processing flexibility. In an optional embodiment of this disclosure, the task execution results include task execution information. After step 106, the following specific step is included: generating a task analysis report based on the task execution information using a report generation model. Task execution information is information about the execution content of the sample task, including but not limited to steps, parameter settings, tool call records, and task execution results. Task execution information is the important data and process description involved in the sample task execution process. It provides the necessary basis for constructing task nodes and unit nodes in the task library and facilitates the generation of responses to different task questions and their corresponding tasks. The report generation model is a deep learning model with report generation capabilities. The information extraction model is pre-trained and can generate task analysis reports based on task execution information. Report generation models include, but are not limited to, the Transformer model, the BERT model, and large models.The report generation model can be a specific model for the target domain, pre-trained with training data from that domain (e.g., fine-tuning, transfer learning, reinforcement learning), or a general model across different domains, implemented based on report generation prompts. This is not a limitation here. A task analysis report is a comprehensive report document generated by the system after completing the target task. It comprehensively presents the process, methods, results, and conclusions of the task execution. A task analysis report typically includes multiple sections, such as an introduction to the task background, a description of the experimental design and methods, data analysis steps, presentation of key results, statistical tests and interpretations, and conclusions and recommendations. It aims to provide a clear and easy-to-understand overview of the results, making them easy for users to review, verify, and reference. Furthermore, the report may include recommendations for future research directions or practical applications. Taking the gene expression differential analysis task as an example, a task analysis report would include: Introduction: Briefly describe the research purpose and background; Methods: Detail the tools used (e.g., NCBI GEO Data Retrieval API, DESeq2 analysis tool), data sources, experimental design, and statistical methods; Results: Present the main contents of the data analysis report, such as tabular data on differential expression of key genes, P-values ​​and adjusted P-values, and various visualizations (e.g., expression network diagrams); Discussion: Explain the significance of the data analysis results and explore the functional relevance and biological significance of the differentially expressed genes; Conclusion: Summarize the main findings and potential application value of the analysis; References: List all data sources and tool references used. For example, in the field of bioinformatics, upon completion of a gene expression differential analysis task, the task results are summarized, including a detailed statistical report, data tables, and visualizations. For example, a raw data analysis report displays the expression changes of specific genes between different cell types and the results of statistical significance tests. In the bioinformatics field, a gene expression differential analysis task file contains the expression levels, fold change, P-value, and corrected P-value for all genes tested. Based on these rich task execution results, a pre-trained large language model is used to deeply interpret and integrate the data, generating a structured task analysis report. This report includes: Abstract: This briefly summarizes the experimental design, objectives, methods, and key findings. Materials and Methods: This report details the sample sources, experimental procedures, and the specific steps for acquiring the dataset using the NCBI GEO Data Retrieval API and performing differential expression analysis using the DESeq2 analysis tool.Results Presentation: Core results are presented with both graphic and text, such as a list of significantly differentially expressed genes, accompanied by heatmaps or volcano plots to illustrate overall expression trends and significance distributions. Gene co-expression network images generated using the Cytoscape plugin are also provided to visually demonstrate gene interactions. Statistical Analysis and Discussion: The significance of statistical test indicators is explained, and the functional annotations of significantly differentially expressed genes and the underlying biological mechanisms underlying differential expression across different cell types are discussed. Conclusion: The main findings of this analysis are summarized, and their value and significance for scientific research or clinical applications are explored. In this disclosed embodiment, the report generation model generates comprehensive and professional task analysis reports based on task execution results, significantly enhancing the richness of task processing and improving the user experience. In an optional embodiment of the present disclosure, a task system includes a model component and a response component. Step 102 includes the following specific steps: In response to a target task in a target domain, triggering the model component to retrieve domain task units related to the target task from a task library corresponding to the target domain; Step 104 includes the following specific steps: triggering the response component to generate meta-task information based on the domain task units; and Step 106 includes the following specific steps: triggering the response component to execute the meta-task based on the meta-task information and obtain task execution results. The model component is configured to build and maintain a task library corresponding to the target domain using a target task model. The target task model is deployed on the model component. Using the target task model, the model component integrates sample tasks in the target domain and the sample task units they invoke to construct a structured storage task library. This library is then used to retrieve various resources required to execute the target task and generate meta-task information, thereby achieving retrieval-enhanced generation. For example, in the bioinformatics field, the target task model integrates sample task units such as the NCBI GEO data retrieval API, the DESeq2 analysis tool, and the Cytoscape visualization plug-in to construct a task library corresponding to the bioinformatics domain. The response component is configured to respond to task requests from the target domain and trigger the task processing flow for the target task. Upon receiving the target task, the response component uses the meta-task generation interface to retrieve the task library on the model component, identify the domain task units associated with the target task, generate meta-task information (i.e., execution flow information) based on the domain task units, and schedule each domain task unit to execute the corresponding meta-tasks sequentially through the meta-task execution interface, thereby completing the execution of the entire target task.For example, in response to the bioinformatics target task of "exploring differential expression of specific gene markers between different cell types," response component 120 retrieves domain task units related to the target task, such as the data retrieval APL differential expression analysis tool and visualization plug-in, and generates corresponding meta-task information guidance. Based on the meta-task information, it sequentially schedules the data retrieval APL differential expression analysis tool and visualization plug-in to execute the corresponding meta-tasks, thereby completing the entire target task of "exploring differential expression of specific gene markers between different cell types." The specific steps in this embodiment are described above and will not be repeated here. In this embodiment, component-based processing improves automation and enhances task processing efficiency. In an optional embodiment of the present disclosure, the model component and the response component are connected via a meta-task generation interface and a meta-task execution interface. In response to a target task in a target domain, the model component is triggered to retrieve domain task units related to the target task from the task library corresponding to the target domain. This includes the following specific steps: In response to the target task in the target domain, the model component is triggered to call the meta-task generation interface to retrieve domain task units related to the target task from the task library corresponding to the target domain. The response component is triggered to execute the meta-task based on the meta-task information and obtain a task execution result. This includes the following specific steps: The response component is triggered to call the meta-task execution interface based on the meta-task information to execute the meta-task and obtain a task execution result. The meta-task generation interface is a functional interface in the target domain task system. It retrieves domain task units related to the target task from the task library on the model component and, based on the information about these task units, constructs a list of meta-tasks required to complete the target task. The meta-task generation interface decomposes a complex, high-level target task into at least one meta-task with clear execution logic and dependencies. For example, when a response component receives a bioinformatics target task, "Explore differential expression of specific gene markers between different cell types," the metatask generation interface searches the task library, identifies domain task units related to the target task (e.g., the NCBI GEO data retrieval API, the DESeq2 analysis tool, and the Cytoscape visualization plugin). Based on these functional characteristics and interdependencies, it generates a metatask sequence consisting of three metatasks. Each metatask includes the corresponding tool or service name, execution parameters, and expected output. The metatask execution interface is a functional interface within the target domain's task system. Based on metatask information, the metatask generation interface executes metatasks according to explicit execution logic and dependencies.The meta-task execution interface parses meta-task information, connects to the execution interface of domain task units, schedules domain task units, and executes the corresponding meta-tasks according to a pre-set process. This meta-task execution interface enables efficient management and automated execution of complex task processes. For example, in the aforementioned bioinformatics example, the meta-task execution interface dispatches domain task units related to the target task (such as the NCBI GEO data retrieval API, the DESeq2 analysis tool, and the Cytoscape visualization plugin) one by one according to the JSON-formatted meta-task information, sequentially executing the three meta-tasks of data acquisition, differential expression analysis, and visualization. The meta-task generation interface and meta-task execution interface are callable interfaces provided by the model component to the response component. The specific steps in this disclosed embodiment are described above and will not be further elaborated. This disclosed embodiment enhances flexibility and standardization through interface-based processing. Referring to Figure 2, a flowchart of a task processing method in the field of bioinformatics, provided by one embodiment of the present disclosure, is shown. The task system, applied to the field of bioinformatics, includes the following specific steps: Step 202: In response to a bioinformatics task in the field of bioinformatics, domain task units related to the bioinformatics task are retrieved from a task library corresponding to the field of bioinformatics. The task library is constructed using a target task model based on sample tasks in at least one source file in the field of bioinformatics and sample task units invoked when executing the sample task. Step 204: Meta-task information is generated based on the domain task units. Step 206: The meta-task is executed based on the meta-task information to obtain a task execution result. The task system in the field of bioinformatics is an artificial intelligence agent system designed and constructed for the knowledge domain of bioinformatics and used to process complex tasks within the field of bioinformatics. For example, in the field of bioinformatics, the task system can be used to automatically complete a series of complex operations involving gene expression differential analysis, data acquisition, data analysis, and visualization. This embodiment of the present disclosure shares the same inventive concept as the embodiment described in Figure 1 above, and the specific contents of steps 202 through 206 are not further described.For example, in a task system in the field of bioinformatics, a user selects the desired cell type, enters a gene list, and enters data sources for the task. The platform's front-end then generates a task request for the field based on this information and sends the request to the platform's server. In response to the target task in the target field, the platform's server invokes a meta-task generation interface. Based on the name of the task (exploring the differential expression of specific gene markers in different cell types), the server retrieves domain bioinformatics tools related to the task from the corresponding relationships between multiple sample bioinformatics tasks and sample bioinformatics tools in the knowledge graph of the field of bioinformatics: NCBI GEO data retrieval, APL DESeq2 analysis tools, and Cytoscape visualization plug-in. The unit information of the domain bioinformatics tool (such as the "data retrieval API" (for obtaining research data), the "acquire research data analysis tool" (for performing differential expression analysis), and the "differential expression analysis plug-in" (for visualizing the result network)) is input into a universal large language model to generate meta-task information corresponding to the bioinformatics task in JSON format. Based on this meta-task information in JSON format, the meta-task information is invoked to call the data retrieval API to execute the meta-task of obtaining research data, the DESeq2 analysis tool to execute the meta-task of differential expression analysis, and the differential expression analysis plug-in to execute the meta-task of visualizing the result network, thereby obtaining a visualization result of the bioinformatics task. In this disclosed embodiment, a task library corresponding to the bioinformatics domain is constructed using the bioinformatics task model based on sample tasks in at least one source file in the bioinformatics domain and the sample task units invoked to execute the sample tasks. This achieves more comprehensive coverage of tasks and meta-tasks in the bioinformatics domain, improves the scalability of task processing, avoids frequent iterative updates, and reduces costs. In response to task requests for bioinformatics tasks, domain task units related to the bioinformatics task are retrieved from the task library corresponding to the bioinformatics field, and meta-task information is generated based on the domain task units. The meta-task is directly executed based on the meta-task information to obtain the task execution results, thereby improving the efficiency of task processing in the bioinformatics field.Figure 3 shows a schematic diagram of the structure of a target domain task system provided by one embodiment of the present disclosure. Task system 300 includes a model component 310 and a response component 320. Model component 310 is configured to construct a task library corresponding to the target domain based on sample tasks in the target domain and sample task units invoked when executing the sample tasks, using a target task model. Response component 320 is configured to, in response to a target task in the target domain, retrieve domain task units related to the target task from the task library and generate meta-task information based on the domain task units. Meta-tasks are then executed based on the meta-task information to obtain task execution results. In this embodiment of the present disclosure, the model component constructs a task library corresponding to the target domain based on sample tasks in at least one source file in the target domain and sample task units invoked when executing the sample tasks, using the target task model. This achieves more comprehensive coverage of tasks and meta-tasks in the target domain, improves the scalability of task processing, avoids frequent iterative updates, and reduces costs. In response to the target task's task request, the response component retrieves domain task units related to the target task from the target domain's corresponding task library. Based on the domain task units, it generates meta-task information and directly executes the meta-task based on the meta-task information to obtain task execution results, improving task processing efficiency. Figure 4 shows a schematic diagram of the architecture of another target domain task system provided by one embodiment of the present disclosure. This task system 300 also includes an acquisition component 410, which connects to a relational database for the target domain. Acquisition component 410 is configured to retrieve sample information for the target domain from the relational database. The sample information includes sample tasks and sample task units invoked to execute the sample tasks. In this embodiment of the present disclosure, the acquisition component retrieves sample information, including sample tasks and sample task units, from the relational database, providing data support for the subsequent construction of the task library. Figure 5 shows a schematic diagram of the structure of a task system for another target domain provided by one embodiment of the present disclosure. Task system 300 also includes a pre-processing component 510. Pre-processing component 510 is configured to analyze sample information in a target format using a data analysis model based on a preset data analysis strategy, obtain analysis results, and determine updated sample information based on the analysis results and the sample information in the target format. In this embodiment of the present disclosure, the pre-processing component analyzes sample information using a data analysis model based on a preset data analysis strategy, obtains analysis results, and determines updated sample information based on the analysis results and the sample information, further ensuring the feasibility of subsequent processing.Figure 6 illustrates the architecture of a task system for a target domain, provided by one embodiment of the present disclosure. As shown in Figure 6, the task system includes an acquisition component 610, a model component 620, a preprocessing component 630, and a response component 640. The acquisition component 610 is configured to acquire source files, a domain task unit library, and sample information of private and open-source data using a data acquisition tool. The model component 620 is configured to perform word segmentation, embedding, and summarization on the source files to extract task execution information. Based on the task execution information, questions and answers are generated to implement task mining. Based on the task execution information, a task library corresponding to the target domain is constructed. The preprocessing component 630 is configured to extract and convert domain data, including private and open-source data, perform data evaluation on the domain data, and interact with front-end users to provide data feedback, thereby obtaining domain data in the target format. The response component 640 is configured to receive task requests from front-end users and, through task mining, determine target tasks for the target domain. Using the target task model, a metatask generation interface is called to generate metatask information. Based on the metatask information, a metatask execution interface is called to execute the metatask on the domain data in the target format, obtaining a task execution result. The task execution result is evaluated to obtain updated metatask information. Based on the updated metatask information, the metatask execution interface is called to execute the metatask on the domain data in the target format, obtaining an updated task execution result. Using the target task model, a task analysis report is generated based on the task execution result or the updated task execution result. The following, combined with FIG7 , further illustrates the target domain task processing method provided by the present disclosure, using its application in the bioinformatics field as an example. FIG7 illustrates a flowchart of a process for a bioinformatics task processing method provided by one embodiment of the present disclosure. The method includes a response component in a task system for the target domain. The system includes a model component and a response component. The model component and the response component are connected via a metatask generation interface and a metatask execution interface. The method includes the following specific steps: Step 702: Receive a task request in the bioinformatics field from a front-end user. For example, the task request may be "Functionally annotate the genome sequence of a newly cultivated plant." Step 704: Call the meta-task generation interface and use the large language model to retrieve domain tools related to the bioinformatics task from the corresponding knowledge graph in the bioinformatics field. Domain tools include: Sequencing data processing tools such as FastQC and Trimmomatic, which are used for quality assessment and preprocessing of raw sequencing data.Genome assembly software, such as SPAdes, Canu, or HiSAT2+Flye, is used to assemble short-read or long-read sequencing data into chromosome-level draft genomes. Repeat sequence identification tools, such as RepeatMasker, are used to identify and filter out repetitive sequence regions in the genome. Gene prediction software, such as Augustus, GlimmerHMM, and GeneMark.hmm, predict the gene structure of the target plant based on known species homology and gene structure characteristics. Protein functional domain prediction tools, such as InterProScan, are used to predict the functional domains of proteins encoding genes. Database alignment tools, such as BLAST or Diamond, are used to perform homology comparison analysis between predicted gene sequences and public databases such as NR and SwissProt. Transcription factor binding site (TFBS) prediction software, such as JASPAR or other TFBS prediction algorithms, are used to predict transcription factor binding sites within gene promoter regions. Step 706: Based on the domain tools, meta-task information is generated using a large language model. The meta-task information is in JSON format:

[0027] "task": "Genome quality assessment seven

[0028] "unit" : "FastQC",

[0029] "params" : (

[0030] "input_files" : ["raw_sequence_reads.fastq.gz"],

[0031] " output_dir" : " quality _check"

[0032] },

[0033] "dependencies" : [],

[0034] "output" : ["fastqc_report.html"]

[0035] },

[0036] {

[0037] "task": "Genome assembly seven

[0038] "unit" : "Canu",

[0039] "params" : (

[0040] "input_file" : "trimmed_sequences.fastq.gz",

[0041] "output_prefix" : "assembly",

[0042] "genome_size" : "500Mb",

[0043] "threads": "16"

[0044] },

[0045] "dependencies" : ["Genome quality assessment

[0046] "output" : [" assembly. contigs. fasta" , "assembly.scaflblds.fasta"]

[0047] },

[0048] {

[0049] "task": "repeated sequence removal",

[0050] "unit" : "RepeatMasker",

[0051] "params" : (

[0052] "input_fasta" : " assembly.contigs.fasta" ,

[0053] "output_format" : "gfF',

[0054] "species" : "plants"

[0055] "task": "Gene prediction seven

[0056] "unit" : "Augustus",

[0057] "params" : (

[0058] "reference_genome" : "assembly_masked.fasta",

[0059] "model_species" : "related_plant_species",

[0060] "output_gfF' : true

[0061] },

[0062] "dependencies" : ["Repeated sequence removal

[0063] "output" : ["predicted_genes.gfF']

[0064] }

[0065] {

[0066] "task": "Protein functional domain prediction seven

[0067] "unit" : "InterProScan",

[0068] "params" : (

[0069] "input_fasta" : "predicted_proteins.fa",

[0070] " output_format" : "xml" ,

[0071] "programs" : ["all"]

[0072] },

[0073] "output" : ["protein domains.xml"]

[0074] },

[0075] Step 708: Based on the meta-task information, call the meta-task execution interface to call the domain tool to execute the meta-task and obtain the task execution result. Step 710: If the execution fails, obtain an alarm message, feed the alarm message back to the front-end user, receive the updated meta-task information sent by the front-end user based on the alarm message, and call the meta-task execution interface based on the updated meta-task information to execute the meta-task and obtain the updated task execution result. Step 712: Based on the updated task execution result, generate a task analysis report using the large language model. For example, the task analysis report is as follows: I. Project Background and Objectives This study performs functional annotation on the whole genome sequence of newly cultivated plant varieties. The sequence was obtained and assembled using high-throughput sequencing technology. The main goal is to reveal its genetic information and potential biological functions to guide breeding improvement, molecular marker development, and subsequent physiological and biochemical function verification research. II. Raw Data Processing and Quality Assessment Sequence Acquisition and Assembly: High-quality genome sequences were successfully obtained by performing second-generation or third-generation sequencing on the samples, and chromosome-level genome drafts were generated using advanced assembly algorithms. Repeat Sequence Identification and Removal: Repeat sequence analysis and filtering were performed on the assembly results using tools such as RepeatMasker to reduce interference during the annotation process. Gene Prediction and Structure Analysis: Based on homology information from known plant species, gene structural characteristics, and expression data, gene structure prediction was performed using software such as Augustus and GlimmerHMM. III. Functional Annotation Key Results Protein-Coding Gene Annotation Gene Location and Structure Description: A total of X protein-coding genes were predicted, distributed across Y chromosomes. The exon and intron distribution and complete CDS regions of each gene are detailed. Protein Functional Domain Prediction: Using tools such as InterProScan, functional domains were searched for all predicted protein amino acid sequences to identify their potential involvement in biochemical reactions or cellular processes. Homology analysis compared to public databases such as NR and SwissProt revealed that a large number of genes shared high homology with known proteins. Based on this, we hypothesized the potential functions of these new genes in metabolic pathways, signal transduction, and other areas. Transcription factor binding site prediction using JASPAR or other TFBS prediction tools identified multiple potential transcription factor binding sites in gene promoter regions. These sites are important for regulating gene expression patterns. IV. Highlights and Discussion: New Gene Discovery: The report highlighted several new genes not previously reported in existing databases and their potential functional associations. Important Regulatory Networks: Based on the analysis of transcription factor binding sites, a transcriptional regulatory network model was constructed that may influence the development of key traits.Evolutionary Relationship Exploration: Phylogenetic tree construction and gene family analysis reveal evolutionary relationships and unique gene clusters between closely related species. In this disclosed embodiment, a large language model is used to integrate domain tools in the bioinformatics field. This model also links user tasks with tool orchestration and autonomous invocation, thereby enabling the understanding and automated execution of bioinformatics tasks. Knowledge graphs are used to quickly and accurately obtain domain tools related to bioinformatics tasks, generating meta-task information. Front-end user feedback is incorporated into the automated execution and usage phases, combining the autonomous judgment of the large language model to deliver results more appropriate for the user's task and generate richer task analysis reports. Figure 8 shows a front-end schematic diagram of a bioinformatics task processing method provided by one embodiment of this disclosure. As shown in Figure 8, the front-end interface includes a content display area, an input box, a send control, and an export control. The front-end user enters the target task information, "Functionally annotate a newly cultivated plant genome sequence," in the input box and clicks the Send control. Steps 702-712 are executed until the task analysis report is obtained and displayed in the display area. The user can then click the Export control to export the task analysis report in a file format such as .doc or .pdf. Figure 9 shows a block diagram of a computing device according to one embodiment of the present disclosure. Components of computing device 900 include, but are not limited to, a memory 910 and a processor 920. Processor 920 is connected to memory 910 via a bus 930, and database 950 is used to store data. Computing device 900 also includes an access device 940, which enables computing device 900 to communicate via one or more networks 960. Examples of these networks include the Public Switched Telephone Network (PSTN), a Local Area Network (LAN), a Wide Area Network (WAN), and a Personal Area Network (PAN). or a combination of communication networks such as the Internet.The access device 940 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a Worldwide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a near field communication (NFC) interface. In one embodiment of the present disclosure, the aforementioned components of the computing device 900 and other components not shown in FIG. 9 may also be connected to each other, for example, via a bus. It should be understood that the computing device structure block diagram shown in FIG. 9 is for illustrative purposes only and does not limit the scope of the present disclosure. Those skilled in the art may add or replace other components as needed. The computing device 900 can be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), a mobile phone (e.g., smartphone), a wearable computing device (e.g., smartwatches, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or personal computer (PC). The computing device 900 can also be a mobile or stationary server. The computing device 900 is equipped with a task system in the target domain or a task system in the bioinformatics field. The above is a schematic diagram of a computing device in this embodiment. It should be noted that the technical solution of this computing device and the technical solutions of the task processing method in the target domain and the task processing method in the bioinformatics field are based on the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solutions of the task processing method in the target domain or the task processing method in the bioinformatics field. The above describes specific embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve desirable results.Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multi-tasking and parallel processing are also possible or may be advantageous. The computer instructions include computer program code, which may be in source code form, object code form, executable files, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signals, telecommunication signals, and software distribution media. It should be noted that the content of the computer-readable medium may be appropriately increased or decreased based on the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media does not include electric carrier signals and telecommunication signals. It should be noted that, for ease of description, the aforementioned method embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments of the present disclosure are not limited by the order of the actions described, as certain steps may be performed in a different order or simultaneously, depending on the embodiments of the present disclosure. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are preferred embodiments, and the actions and modules described are not necessarily required for the embodiments of the present disclosure. In the above embodiments, the description of each embodiment has its own emphasis. For portions not described in detail in a particular embodiment, reference should be made to the relevant descriptions of other embodiments. The preferred embodiments disclosed above are merely intended to help illustrate the present disclosure. The optional embodiments do not describe all details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and variations are possible based on the content of the embodiments of the present disclosure. The present disclosure selects and describes these embodiments in detail to better explain the principles and practical applications of the embodiments of the present disclosure, thereby enabling those skilled in the art to better understand and utilize the present disclosure. The present disclosure is limited only by the claims and their full scope and equivalents.

Claims

27 Claims 1. A task processing method in a target domain, applied to a task system in the target domain, comprising: In response to the target task of the target domain, domain task units related to the target task are retrieved from a task library corresponding to the target domain, wherein the task library is constructed through a target task model based on sample tasks in at least one material file of the target domain and sample task units called for executing the sample tasks; meta-task information is generated based on the domain task units; and the meta-task is executed based on the meta-task information to obtain a task execution result.

2. The method according to claim 1, further comprising: before the step of retrieving domain task units related to the target task from a task library corresponding to the target domain in response to the target task of the target domain, Acquire sample information of the target domain from a related database of the target domain, wherein the sample information includes a sample task and a sample task unit called to execute the sample task.

3. The method according to claim 2, further comprising, after obtaining the sample information of the target domain from the associated database of the target domain: Analyze the sample information using a data analysis model based on a preset data analysis strategy to obtain analysis results; Based on the analysis result and the sample information, updated sample information is determined.

4. The method according to claim 2, further comprising, after obtaining the sample information of the target domain from the associated database of the target domain: Feedback the sample information to the client; receiving an evaluation result for the sample information sent by the client; Based on the evaluation result, an adjustment strategy corresponding to the evaluation result is determined, and based on the adjustment strategy, the sample information is adjusted.

5. The method according to any one of claims 1 to 4, further comprising: before the step of retrieving domain task units related to the target task from a task library corresponding to the target domain in response to the target task of the target domain; Extracting sample tasks from the sample information of the target domain and sample task units called for executing the sample tasks through an information extraction model; Constructing the task nodes of the sample tasks and the unit nodes of the sample task units, and determining the corresponding relationship between the task nodes and the unit nodes; and constructing a task library corresponding to the target domain based on the task nodes, the unit nodes, and the corresponding relationship.

6. The method according to any one of claims 1 to 4, further comprising: before the step of retrieving domain task units related to the target task from a task library corresponding to the target domain in response to the target task of the target domain; extracting task execution information from sample information in the target domain using an information extraction model; generating a plurality of task problems in the target domain based on the task execution information; The method of responding to the target task of the target domain and retrieving domain task units related to the target task from a task library corresponding to the target domain includes: receiving a target task of the target domain sent by a client; determining a target task problem based on the task request, and feeding back the target task problem to the client; receiving a target task reply fed back by the client in response to the target task problem; and determining the target task of the target domain based on the target task problem and the target task reply.

7. The method according to any one of claims 1 to 4, wherein the task execution result includes alarm information; after executing the meta-task based on the meta-task information and obtaining the task execution result, further comprising: Feedback the alarm information to the client; receiving update meta-task information sent by the client based on the alarm information; The meta-task is executed based on the updated meta-task information to obtain an updated task execution result.

8. The method according to any one of claims 1 to 4, wherein the task execution result includes alarm information; after executing the meta-task based on the meta-task information and obtaining the task execution result, further comprising: An information generation model is used to generate update meta-task information based on the alarm information and the domain task unit, and a meta-task is executed based on the update meta-task information to obtain an update task execution result.

9. The method according to any one of claims 1 to 4, further comprising, after executing the meta-task based on the meta-task information and obtaining a task execution result: Based on the meta-task information, generate a meta-task execution flowchart; Feedback the execution flow chart to the client; receiving update meta-task information sent by the client for the execution flowchart; The meta-task is executed based on the updated meta-task information to obtain a task execution result.

10. The method according to any one of claims 1 to 4, wherein the task execution result comprises task execution information; and after the step of calling the meta-task execution interface to execute the meta-task based on the meta-task information and obtaining the task execution result, further comprising: Generate a task analysis report based on the task execution information through a report generation model.

11. The method according to claim 1, wherein the task system comprises a model component and a response component; wherein the step of responding to the target task in the target domain, retrieving domain task units related to the target task from a task library corresponding to the target domain, comprises: In response to the target task of the target domain, triggering the model component to retrieve domain task units related to the target task from a task library corresponding to the target domain; The generating meta-task information based on the domain task unit, comprising: triggering the response component, based on the domain task unit generating meta-task information; The executing the meta-task based on the meta-task information and obtaining the task execution result includes: triggering the response component, executing the meta-task based on the meta-task information, and obtaining the task execution result.

12. The method according to claim 11, wherein the model component is connected to the response component via a meta-task generation interface and a meta-task execution interface; and wherein, in response to the target task of the target domain, triggering the model component to retrieve domain task units related to the target task from a task library corresponding to the target domain comprises: In response to the target task of the target domain, trigger the model component to call the meta-task generation interface to retrieve domain task units related to the target task from the task library corresponding to the target domain; The triggering of the response component, executing the meta-task based on the meta-task information, and obtaining the task execution result includes: triggering the response component, calling the meta-task execution interface based on the meta-task information to execute the meta-task, and obtaining the task execution result.

13. A task processing method in the field of bioinformatics, applied to a task system in the field of bioinformatics, comprising: In response to the bioinformatics task in the bioinformatics field, domain task units related to the bioinformatics task are retrieved from a task library corresponding to the bioinformatics field, wherein the task library is constructed through a target task model based on sample tasks in at least one source file in the bioinformatics field and sample task units called to execute the sample tasks; meta-task information is generated based on the domain task units; and the meta-task is executed based on the meta-task information to obtain a task execution result.

14. A task system for a target domain, the task system comprising a model component and a response component; the model component being configured to construct a task library corresponding to the target domain based on sample tasks in the target domain and sample task units invoked to execute the sample tasks using a target task model; The response component is configured to respond to the target task of the target domain, retrieve domain task units related to the target task from the task library, and generate meta-task information based on the domain task units; The meta-task is executed based on the meta-task information to obtain a task execution result.

15. The task system according to claim 14, wherein the model component and the response component are connected via a meta-task generation interface and a meta-task execution interface; the response component is configured to respond to the target task of the target domain, call the meta-task generation interface, retrieve the domain task units related to the target task from the task library, and generate meta-task information based on the domain task units; and call the meta-task execution interface to execute the meta-task based on the meta-task information to obtain the task execution result.

16. The task system according to claim 14, further comprising an acquisition component, wherein the acquisition component is connected to a relational database of the target domain; the acquisition component is configured to acquire sample information of the target domain from the relational database, wherein: The sample information includes a sample task and a sample task unit called to execute the sample task.

17. The task system according to claim 14, further comprising a preprocessing component; the preprocessing component is configured to analyze the sample information in the target format through a data analysis model based on a preset data analysis strategy to obtain an analysis result, and determine updated sample information based on the analysis result and the sample information in the target format.

18. A computing device, comprising: memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer programs / instructions are executed by the processor, the steps of the method according to any one of claims 1 to 13 are implemented.

19. A computer-readable storage medium storing a computer program / instruction, wherein the computer program / instruction, when executed by a processor, implements the steps of the method according to any one of claims 1 to 13.

20. A computer program product, comprising a computer program / instructions, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 13.

Images