A Bioinformatics Automated Analysis System Based on AI Large Model

Abstract

This invention relates to an automated bioinformatics analysis system based on an AI large-scale model. The system includes a data input interface for receiving sequencing data; a multi-agent analysis engine driven by a large language model and comprising multiple dedicated intelligent agents, which collaboratively execute various stages of the de novo genome assembly analysis process based on the sequencing data; and a heterogeneous resource scheduling module for adapting and submitting the analysis task instructions generated by the dedicated intelligent agents in the multi-agent analysis engine to the corresponding computing cluster scheduling system. By employing a multi-agent analysis engine driven by a large language model, and including multiple dedicated intelligent agents, the various stages of the de novo genome assembly process can be collaboratively executed through these dedicated intelligent agents, thereby reducing the fragmentation of the various stages and improving the efficiency of de novo genome assembly.

Description

Technical Field

[0001] This invention relates to the field of biological detection technology, and in particular to an automated bioinformatics analysis system based on an AI large model. Background Technology

[0002] De novo genome assembly is a core technology for deciphering the genetic blueprint of species, playing a crucial role in fields such as species evolution research, functional gene mining, and biobreeding. Its analysis workflow typically includes multiple stages such as data quality control, genome characterization surveys, sequence assembly, chromosome mounting, and gene annotation. With the widespread adoption of high-throughput sequencing technology, the amount of data from de novo genome assembly is growing exponentially, thus placing higher demands on the automation, intelligence, and reliability of each stage of its analysis workflow.

[0003] Traditionally, the various stages of the de novo genome assembly analysis process typically rely on a series of decentralized bioinformatics tools (such as Nextflow, Snakemake, etc.) to manage each stage independently. However, this management approach can easily lead to the fragmentation of the various stages of the de novo genome assembly analysis process, requiring parameter re-adaptation when switching between different bioinformatics tools, which affects the efficiency of de novo genome assembly. Summary of the Invention

[0004] In view of this, it is necessary to provide a bioinformatics automated analysis system based on AI large models to solve the problem that existing technologies rely on operating a series of scattered bioinformatics tools for independent management, resulting in the fragmentation of each step in the genome de novo assembly analysis process.

[0005] To address the aforementioned problems, this invention provides a bioinformatics automated analysis system based on a large AI model, comprising: a data input interface, a multi-agent analysis engine connected to the data input interface, and a heterogeneous resource scheduling module connected to the multi-agent analysis engine, wherein: The data input interface is used to receive sequencing data; The multi-agent analysis engine is driven by a large language model and includes multiple dedicated intelligent agents, which are used to collaboratively execute each step of the genome de novo assembly analysis process based on the sequencing data. The heterogeneous resource scheduling module is used to adapt and submit the analysis task instructions generated by each dedicated intelligent agent in the multi-agent analysis engine during the analysis process to the corresponding computing cluster scheduling system.

[0006] In one possible implementation, the multi-agent analysis engine further includes a shared memory pool connected to each dedicated intelligent agent, wherein the shared memory pool is used to store the analysis results generated by any of the dedicated intelligent agents and to allow access by other dedicated intelligent agents, so as to realize the flow and collaboration of analysis results among multiple dedicated intelligent agents.

[0007] In one possible implementation, the multiple dedicated intelligent agents specifically include: a genome survey agent, a sequence assembly agent, a chromosome mounting agent, and a gene annotation agent, wherein: The genome survey agent is used to perform quality assessment, genome feature prediction, and comparative analysis with closely related species on the sequencing data; The sequence assembly agent is used to automatically select assembly strategies and parameters based on the analysis results of the genome survey agent, and to perform sequence assembly and assembly quality assessment. The chromosome mounting agent is used to perform chromosome-level mounting, error correction, and integrity assessment on the assembled sequences output by the sequence assembly agent using three-dimensional genomics data. The gene annotation agent is used to perform repetitive sequence annotation, gene structure prediction, and multi-database-based functional annotation on the chromosome-level sequences output by the chromosome mounting agent.

[0008] In one possible implementation, the genome survey agent specifically includes: An automated quality control module is used to assess the quality of sequencing data by comparing the quality control indicators of the sequencing data with multiple preset quality control thresholds. A genome feature prediction module is used to perform K-mer analysis on sequencing data that has passed quality assessment in order to predict the feature parameters of the target genome, wherein the feature parameters include at least one of genome size, heterozygosity and repetitive sequence ratio; The closely related species comparison analysis module is used to access external databases to obtain genomic feature data of at least one closely related species, compare and analyze the feature parameters estimated by the genomic feature prediction module with the genomic feature data of the closely related species, and output assembly parameter suggestions based on the comparison analysis results.

[0009] In one possible implementation, at least one of the preset quality control thresholds is dynamically calibrated based on the genomic characteristic data of closely related species obtained from the closely related species comparison analysis module.

[0010] In one possible implementation, the sequence assembly agent specifically includes: The data type identification and strategy selection module is used to identify the corresponding sequencing data type based on the sequence length distribution of the sequencing data, and, in combination with the sequencing data type and the genome feature analysis results provided by the genome survey agent, select the corresponding assembly tool and generate suitable assembly parameters for gene sequence assembly. The assembly quality assessment module is used to perform contamination detection, sequence consistency quality assessment, and assembly integrity assessment on the assembled gene sequences. The exception handling module is used to automatically generate an evaluation report and trigger an external confirmation process when the assembly quality assessment module outputs an indication that the assembly quality does not meet the preset standard.

[0011] In one possible implementation, the chromosome mounting agent specifically includes: The data comparison and evaluation module is used to compare the input three-dimensional genomics data with the assembled gene sequences and to screen effective pairs based on a preset comparison quality threshold. The intelligent clustering mounting module is used to perform chromosome-level clustering and sorting orientation of the gene sequences assembled based on the effective interaction pairs, so as to realize chromosome mounting. The visual error correction and filling module is used to generate a visual interaction map based on the effective interaction pairs to help identify and correct assembly errors, and to fill gaps in the gene sequence after chromosome mounting using long read sequence data; The quality approval module is used to calculate the chromosome anchoring rate and interaction strength distribution of the gene sequence after chromosome mounting to evaluate the mounting quality, and trigger an external approval process containing mounting details when the quality assessment is passed.

[0012] In one possible implementation, the intelligent clustering mounting module is further configured to dynamically adjust the minimum number of connections required by the clustering algorithm based on the proportion of effective interaction pairs and the genome size of the assembled gene sequence.

[0013] In one possible implementation, the gene annotation agent specifically includes: The repetitive sequence annotation module is used to annotate tandem repeats and scattered repeats in gene sequences at the chromosome level. The gene structure prediction module is used to predict and integrate the structure of protein-coding genes in chromosome-level gene sequences based on homology alignment, de novo prediction, and transcriptome evidence. The functional annotation module is used to compare and analyze the predicted gene structures using multiple functional databases to achieve gene function annotation. The annotation quality assessment module is used to evaluate the integrity of gene annotations based on a preset integrity threshold and to count the proportion of genes supported by different evidence, so as to trigger an external review process that includes annotation details.

[0014] In one possible implementation, the bioinformatics automated analysis system further includes an enterprise communication and approval integration module connected to each dedicated intelligent agent, which is used to generate and send alarm messages or approval request messages at preset key decision nodes or anomaly detection nodes through an integrated external communication application interface during the collaborative execution of each dedicated intelligent agent. as well as, The system receives approval result instructions through the external communication application interface and controls the continuation, retry, or termination of subsequent analysis processes based on these instructions.

[0015] The beneficial effects of this invention are: Compared with existing technologies, the bioinformatics automated analysis system based on an AI large model provided in this application includes a data input interface, a multi-agent analysis engine connected to the data input interface, and a heterogeneous resource scheduling module connected to the multi-agent analysis engine. The data input interface receives sequencing data; the multi-agent analysis engine is driven by a large language model and includes multiple dedicated intelligent agents, which collaboratively execute various stages of the de novo genome assembly analysis process based on the sequencing data; the heterogeneous resource scheduling module adapts and submits the analysis task instructions generated by the dedicated intelligent agents in the multi-agent analysis engine to the corresponding computing cluster scheduling system. Because this bioinformatics automated analysis system uses a multi-agent analysis engine driven by a large language model, and this multi-agent analysis engine includes multiple dedicated intelligent agents, the various stages of the de novo genome assembly process can be collaboratively executed through these dedicated intelligent agents, thereby reducing the fragmentation of the various stages in the de novo genome assembly process, reducing the need for parameter configuration, and improving the efficiency of de novo genome assembly. Attached Figure Description

[0016] Figure 1 A schematic diagram of the structure of the AI-based large-scale bioinformatics automated analysis system provided by this invention; Figure 2 A schematic diagram of the structure of a genome survey agent in the bioinformatics automated analysis system provided by this invention; Figure 3 A schematic diagram of the sequence assembly agent in the bioinformatics automated analysis system provided by this invention; Figure 4 A schematic diagram of the chromosome mounting agent structure in the bioinformatics automated analysis system provided by this invention; Figure 5 This is a schematic diagram of the structure of the gene annotation agent in the bioinformatics automated analysis system provided by the present invention. Detailed Implementation

[0017] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0018] As mentioned earlier, the current de novo genome assembly analysis process typically relies on a series of decentralized bioinformatics tools to manage each step independently. However, this management approach can easily lead to the fragmentation of the de novo genome assembly analysis process, requiring parameter re-adaptation during the switching between different bioinformatics tools, which affects the efficiency of de novo genome assembly.

[0019] In view of this, embodiments of this application provide a bioinformatics automated analysis system based on an AI large model, which can be used to solve the problems in the prior art. For example... Figure 1 The diagram shown is a schematic of the specific structure of the bioinformatics automated analysis system. The bioinformatics automated analysis system includes a data input interface 10, a multi-agent analysis engine 20, and a heterogeneous resource scheduling module 30. The multi-agent analysis engine 20 is connected to the data input interface 10, and the heterogeneous resource scheduling module 30 is connected to the multi-agent analysis engine 20.

[0020] The data input interface 10 refers to a logical port or physical channel in a bioinformatics automated analysis system used to access raw biological sequencing data. The data input interface 10 functions as the data source for the entire analysis process, responsible for accepting and receiving various types of sequencing data inputs. In practical applications, the sequencing data received by this data input interface can be configured according to specific needs. For example, it can be second-generation sequencing data (such as paired-end reads generated by DNBseq or Illumina platforms), third-generation sequencing data (such as long-read data generated by PacBio HiFi or Oxford Nanopore), Hi-C or Pore-C data for chromosome conformation capture, and RNA-seq transcriptome data for annotation. This data input interface 10 connects to the multi-agent analysis engine 20, enabling the standardized reception of heterogeneous data to be transmitted to the downstream multi-agent analysis engine 20 to initiate subsequent automated analysis processes.

[0021] For example, in practical applications, users can input second-generation sequencing data via the second-generation data input path / data / demo_R1.fq.gz / data / demo_R2.fq.gz, or third-generation sequencing data via the third-generation data input path / data / hifi.fastq.gz. They can also input sequencing data using a given genome file assembly.fasta and Pore-C data porec.fastq.gz, or a given genome file chromosome.fasta and RNA-seq data rna.fq.gz. This data input interface 10 can receive sequencing data from these different sources and transmit it to the downstream multi-agent analysis engine 20.

[0022] It is important to note that the multi-agent analysis engine 20 provided in this embodiment is driven by a Large Language Model (LLM). Therefore, the multi-agent analysis engine 20 is based on the core decision-making and execution unit of the LLM. In this embodiment, the multi-agent analysis engine 20 includes multiple dedicated intelligent agents. These dedicated intelligent agents do not operate in isolation, but are used to collaboratively execute various steps in the genome de novo assembly analysis process based on the sequencing data received by the data input interface 10. The LLM can be the Qwen3-Thinking Large Language Model, or other Large Language Models; the specific type of Large Language Model is not limited here.

[0023] In the overall technical solution of this application, since the multi-agent analysis engine 20 is based on LLM-driven architecture, its function is to simulate the analytical thinking of a senior bioinformatics expert, autonomously planning and executing the entire process of de novo genome assembly. Within this framework, multiple dedicated intelligent agents cooperate with the data input interface 10 to collaboratively execute each step of the de novo genome assembly analysis process based on the sequencing data received by the data input interface 10.

[0024] For example, when the data input interface 10 receives third-generation long-read data, a specific agent in the multi-agent analysis engine can be activated to perform the assembly task; when Hi-C data is received, another specific agent is activated to perform the mounting task. Furthermore, the dedicated intelligent agents can achieve seamless integration between stages through information flow; the output of one dedicated intelligent agent can directly serve as the input for the next, collaboratively completing the construction process from raw data to high-quality genome sequences without manual intervention.

[0025] The heterogeneous resource scheduling module 30 is used to adapt and submit the analysis task instructions generated by each dedicated intelligent agent in the multi-agent analysis engine 20 in the analysis process to the corresponding computing cluster scheduling system. Therefore, the heterogeneous resource scheduling module 30 is actually an adaptation layer component between the multi-agent analysis engine 20 and the underlying computing infrastructure (used to provide different types of computing resources), thereby solving the problem of differences in job submission commands and resource configuration syntax between different high-performance computing (HPC) clusters.

[0026] In this embodiment, the heterogeneous resource scheduling module 30 receives analysis task instructions generated by various dedicated intelligent agents in the multi-agent analysis engine 20 during the analysis process. These instructions contain abstract descriptions such as the required computing resource type, memory size, runtime, and number of parallel cores. The heterogeneous resource scheduling module 30 converts these abstract descriptions into specific job scripts recognizable by a particular computing cluster scheduling system. For example, if the target cluster is an SGE system, the heterogeneous resource scheduling module 30 can adapt the task instructions to a script containing parameters such as qsub and -pe mpi; if the target cluster is a Slurm system, it can adapt them to a script containing parameters such as sbatch and --ntasks. Through this adaptation mechanism, the heterogeneous resource scheduling module 30 can decouple the analysis logic from the underlying hardware environment, enabling the same analysis process to be seamlessly deployed in different computing environments.

[0027] Therefore, the bioinformatics automated analysis system provided in this application embodiment constructs a three-layer automated architecture consisting of a data input interface 10, a multi-agent analysis engine 20, and a heterogeneous resource scheduling module 30. This architecture utilizes the reasoning capabilities of a large language model to drive multiple dedicated intelligent agents in the multi-agent analysis engine 20, integrating the originally fragmented bioinformatics analysis process into a continuous automated pipeline, and shielding the differences in the underlying computing environment through the heterogeneous resource scheduling module 30.

[0028] In the actual operation of this bioinformatics automated analysis system, the system first receives sequencing data submitted by the user through the data input interface 10, and further identifies the data type of the sequencing data. Subsequently, the multi-agent analysis engine 20 is activated. The dedicated intelligent agents within this engine coordinately execute various stages of the de novo genome assembly analysis process based on the sequencing data. For example, some dedicated intelligent agents are responsible for data quality assessment and genome feature prediction; some automatically select assembly strategies and execute sequence assembly based on the assessment results; and others utilize auxiliary data for chromosome mounting and functional annotation. When computational resources are needed, the dedicated intelligent agents in the multi-agent analysis engine 20 generate analysis task instructions and send them to the heterogeneous resource scheduling module 30. The heterogeneous resource scheduling module 30 then parses and translates these instructions into job submission commands adapted to the target computing cluster (such as SGE or Slurm), and submits them to the corresponding computing cluster scheduling system for the scheduling of appropriate computing resources.

[0029] Therefore, the bioinformatics automated analysis system based on an AI large model provided in this application includes a data input interface, a multi-agent analysis engine connected to the data input interface, and a heterogeneous resource scheduling module connected to the multi-agent analysis engine. The data input interface receives sequencing data; the multi-agent analysis engine is driven by a large language model and includes multiple dedicated intelligent agents, which collaboratively execute various stages of the de novo genome assembly analysis process based on the sequencing data; the heterogeneous resource scheduling module adapts and submits the analysis task instructions generated by the dedicated intelligent agents in the multi-agent analysis engine to the corresponding computing cluster scheduling system. Because this bioinformatics automated analysis system uses a multi-agent analysis engine driven by a large language model, and this multi-agent analysis engine includes multiple dedicated intelligent agents, various stages of the de novo genome assembly process can be collaboratively executed through these dedicated intelligent agents, thereby reducing the fragmentation of the various stages in the de novo genome assembly process, reducing the need for parameter configuration, and improving the efficiency of de novo genome assembly.

[0030] It is important to note that, in addition to including multiple dedicated intelligent agents, the multi-agent analysis engine 20 in this application may further include a shared memory pool 201 connected to each dedicated intelligent agent. The shared memory pool 201 is used to store the analysis results generated by any dedicated intelligent agent and to allow other dedicated intelligent agents to access them, so as to realize the flow and collaboration of the analysis results among multiple dedicated intelligent agents.

[0031] The shared memory pool 201 can be a data storage and exchange medium deployed inside or communicating with the multi-agent analysis engine, and its function is to serve as a central information hub between various dedicated intelligent agents. The structure of the shared memory pool 201 can be set according to the actual situation. For example, it can be a high-speed key-value database based on memory (such as Redis), or it can be a specific data table in a distributed file system or relational database. This application embodiment does not impose any special limitations on this.

[0032] In this application, the shared memory pool 201 has a bidirectional data interaction relationship with multiple dedicated intelligent agents: on the one hand, after completing its specific analysis task, any dedicated intelligent agent can write the generated intermediate analysis results, status flags, or decision parameters into the shared memory pool 201; on the other hand, other dedicated intelligent agents that need to rely on the aforementioned results can read the corresponding data from the shared memory pool 201. Through this write, store, and read mechanism, the shared memory pool 201 can achieve seamless transfer of analysis results between different dedicated intelligent agents, enabling subsequent dedicated intelligent agents to directly obtain the processing results of previous stages without recalculation or inefficient transfer through external files, thereby further improving the coherence of the analysis process and the consistency of data.

[0033] It should be further explained that the aforementioned multi-agent analysis engine 20 includes multiple dedicated intelligent agents. In this application, in conjunction with the various stages of the de novo assembly process, these multiple dedicated intelligent agents are further configured as a genome survey agent 202 (Survey Agent), a sequence assembly agent 203 (Assembly Agent), a chromosome mounting agent 204 (Scaffolding Agent), and a gene annotation agent 205 (Annotation Agent). In this application, the genome survey agent 202, sequence assembly agent 203, chromosome mounting agent 204, and gene annotation agent 205 are all connected to the shared memory pool 201, so that they can respectively transmit their analysis results to the shared memory pool 201 for access by other dedicated intelligent agents.

[0034] Among them, the genome survey agent 202 is used to perform quality assessment, genome feature prediction, and comparison analysis with closely related species on sequencing data; the sequence assembly agent 203 is used to automatically select assembly strategies and parameters based on the analysis results of the genome survey agent, and perform sequence assembly and assembly quality assessment; the chromosome mounting agent 204 is used to perform chromosome-level mounting, error correction, and integrity assessment on the assembled sequences output by the sequence assembly agent using three-dimensional genomics data; and the gene annotation agent 205 is used to perform repetitive sequence annotation, gene structure prediction, and multi-database-based functional annotation on the chromosome-level sequences output by the chromosome mounting agent.

[0035] For example, suppose a user initiates a de novo assembly process for a plant genome. After the data input interface 10 receives the sequencing data and transmits it to the multi-agent analysis engine 20, the genome survey agent 202 is first triggered. This agent performs quality assessment, genome feature prediction, and comparison analysis with closely related species on the sequencing data. The genome survey agent 202 can then write some or all of the analysis results to the shared memory pool 201 (e.g., to a specified namespace within the shared memory pool 201). Subsequently, the sequence assembly agent 203 is scheduled to execute. It can first query the shared memory pool 201, read some or all of the analysis results written by the genome survey agent 202, and automatically select assembly strategies and parameters based on these results. It then performs sequence assembly and assembly quality assessment. 203 can also write the analysis results to the shared memory pool 201; then the chromosome mounting agent 204 is scheduled to be executed, which can also obtain the analysis results of various upstream agents (including the genome survey agent 202 and the sequence assembly agent 203) from the shared memory pool 201, and then use the three-dimensional genomics data to perform chromosome-level mounting, error correction and integrity assessment on the assembled sequence output by the sequence assembly agent, and write the analysis results to the shared memory pool 201; then the gene annotation agent 205 is scheduled to be executed, which can also obtain the analysis results of various upstream agents from the shared memory pool 201, and perform repetitive sequence annotation, gene structure prediction and multi-database-based functional annotation on the chromosome-level sequence output by the chromosome mounting agent.

[0036] Through the above technical solution, this application sets up a shared memory pool 201 connected to each dedicated intelligent agent, so that the analysis results generated by any agent can be stored in a standardized manner and shared with other agents in real time. This solves the problem of information silos caused by inconsistent data formats or reliance on manual intervention in traditional workflows, thereby achieving the technical effects of improving the collaborative efficiency among multiple agents, reducing the consumption of redundant computing resources, and ensuring the consistency of data throughout the analysis process.

[0037] The genome survey agent 202 can be further described here. This agent is used for quality assessment of sequencing data, genomic feature prediction, and comparative analysis with closely related species. In practical applications, the agent 202 may further include an intelligent processing unit deployed at the front end of the multi-agent analysis engine 20, connected to the data input interface 10. This unit receives the sequencing data transmitted from the data input interface 10 and performs quality assessment, genomic feature prediction, and comparative analysis with closely related species. Specifically, the agent 202 can perform quality indicator statistics on the sequencing data (such as Q20 / Q30 distribution, GC content, etc.), estimate genome size, heterozygosity, and repetitive sequence ratio based on K-mer frequency distribution fitting, and obtain genomic feature data of closely related species from external public databases for comparative analysis, thereby outputting an analysis report including assembly difficulty assessment and preliminary parameter suggestions. Of course, the specific algorithm model of the genome survey agent 202 can be set according to the actual situation. For example, it can be an inference module based on LLM, or an automated script container that integrates traditional bioinformatics tools such as FastK and GenomeScope. This application embodiment does not make any special limitations on this.

[0038] like Figure 2 As shown, in practical applications, the genome survey agent 202 may further include an automated quality control module 2021, a genome feature prediction module 2022, and a closely related species comparison analysis module 2023. The automated quality control module 2021 is used to assess the quality of sequencing data by comparing the quality control indicators of the sequencing data with multiple corresponding preset quality control thresholds. Specifically, for example, when at least one quality control indicator exceeds its corresponding preset quality control threshold, a corresponding alarm message is triggered and output. Only when none of the quality control indicators exceed their corresponding preset quality control thresholds is the data determined to have passed quality control.

[0039] In practical applications, the automated quality control module 2021 can extract quality control indicators from sequencing data, such as base quality distribution (Q20 / Q30), GC content distribution, N base ratio, sequence length distribution, and adapter contamination. These quality control indicators are compared with multiple preset quality control thresholds. If at least one quality control indicator exceeds the corresponding preset quality control threshold, such as Q30 being below 90% or GC distribution showing abnormal multi-peaks, it indicates that the quality control has not passed (there is a quality risk), triggering an alarm mechanism and outputting corresponding alarm information. This alarm mechanism can generate structured alarm information, and the quality control is only considered passed when all quality control indicators do not exceed their corresponding preset quality control thresholds.

[0040] The genome feature prediction module 2022 is used to perform K-mer analysis on sequencing data that has passed quality assessment to predict the characteristic parameters of the target genome. These characteristic parameters include at least one of genome size, heterozygosity, and the proportion of repetitive sequences. This module 2022 is connected to the automated quality control module 2021, enabling it to receive sequencing data that has passed quality control (i.e., quality assessment) and perform K-mer analysis on it. Specifically, the genome feature prediction module 2022 can fit key characteristic parameters such as the size, heterozygosity, and proportion of repetitive sequences of the target genome by constructing a K-mer frequency distribution histogram.

[0041] In this application, the genome feature prediction module 2022 and the automated quality control module 2021 work in tandem. The genome feature prediction module 2022 only starts its calculation task after the automated quality control module 2021 confirms that the sequencing data has passed quality control, thus avoiding invalid feature estimation of low-quality data. Simultaneously, the target genome feature parameters output by the genome feature prediction module 2022 will serve as the basic input data for subsequent analysis steps.

[0042] The closely related species comparison analysis module 2023 is used to access external databases to obtain genomic characteristic data of at least one closely related species. It compares and analyzes the characteristic parameters estimated by the genomic characteristic estimation module 2022 with the genomic characteristic data of closely related species, and outputs assembly parameter suggestions based on the comparison analysis results. Specifically, the closely related species comparison analysis module 2023 can access public bioinformatics databases (such as the NCBI NT database) via a network interface to obtain genomic characteristic data of closely related species with similar taxonomic positions to the species currently being analyzed.

[0043] In this application, the closely related species comparison analysis module 2023 is connected to the genome feature prediction module 2022. The closely related species comparison analysis module 2023 receives the feature parameters output by the genome feature prediction module 2022 and performs multi-dimensional comparison analysis with the acquired closely related species feature data. Based on the difference rate of the comparison results (e.g., the heterozygosity of this sample is significantly lower than that of closely related species), the closely related species comparison analysis module can identify potential risk levels and output targeted assembly parameter suggestions accordingly. For example, when the comparison results show that the proportion of repetitive sequences in this sample is much lower than that in closely related species, the closely related species comparison analysis module 2023 can suggest a specific redundancy removal strategy.

[0044] The assembly parameters in this assembly parameter suggestion can be generated based on an automatic assembly parameter generation algorithm. This algorithm is based on the heterozygosity rate h and the repetition rate r, and its calculation formula is n_hap = round(1 + 2×h / (1-h) × (1-r)). This allows for the dynamic calculation of the n-hap parameter in HiFiasm, avoiding reliance on human experience. In this formula, n-hap is the n-hap parameter in HiFiasm; h is the heterozygosity rate; r is the repetition rate; and round is the rounding operator.

[0045] It should be further clarified that at least one preset quality control threshold in the aforementioned automated quality control module 2021 can be dynamically calibrated based on the genomic characteristic data of closely related species obtained from the closely related species comparison analysis module 2023. In other words, at least one of these preset quality control thresholds is not a fixed static constant, but a dynamic parameter that can be flexibly adjusted according to the specific biological characteristics of the sample to be analyzed.

[0046] The system dynamically calibrates the preset quality control threshold. Specifically, if the close-related species comparison analysis shows that the current sample belongs to a high-GC-content species (e.g., GC content significantly higher than 45% in conventional model organisms), the system can automatically lower the default Q30 threshold from 95% to 93% or other suitable values ​​according to calibration rules. This tolerates the inherent sequencing quality fluctuations in high-GC regions and prevents a large number of valid reads from being incorrectly removed. Conversely, if close-related species data show that the species typically has extremely high sequence consistency requirements, the system may raise the relevant threshold. Generally, the dynamic adjustment formula for the Q30 threshold for high-GC-content species is: threshold = 95 - 0.5 × (gc_content - 45), where threshold is the dynamically adjusted Q30 threshold, and gc_content is the GC content of the high-GC-content species.

[0047] The aforementioned sequence assembly agent 203 can automatically select assembly strategies and parameters based on the analysis results of the genome survey agent 202, and perform sequence assembly and assembly quality assessment. For example, the sequence assembly agent 203 can communicate via the shared memory pool 201 or the internal bus to read the analysis results of the genome survey agent 202, including key features such as heterozygosity and genome size. Based on this, it dynamically adjusts the operating parameters of the assembly tool (such as haploid / diploid mode switching, k-mer value selection, etc.) and calls the corresponding assembly engine to execute the sequence splicing task. After completing the initial assembly, the sequence assembly agent 203 can also perform assembly quality assessments on the generated gene sequences, such as consistency checks, contamination screening, and integrity assessments (e.g., BUSCO scoring), and pass the qualified assembled sequences to the subsequent chromosome mounting agent 204. If the assessment fails, the sequence assembly agent 203 can trigger an anomaly handling process or request manual intervention. In practical applications, the sequence assembly agent 203 can integrate any one or more of the assembly tools (HiFiasm, wtdbg2, Canu, etc.), and the specific selection depends on the type and feature parameters of the input data.

[0048] like Figure 3 As shown, the sequence assembly agent 203 may further include a data type identification and strategy selection module 2031, an assembly quality assessment module 2032, and an anomaly handling module 2033. The data type identification and strategy selection module 2031 is a functional unit that can automatically determine the data type based on the sequence length distribution of the sequencing data and intelligently select assembly tools and generate adaptation parameters in conjunction with genomic feature analysis results. For example, when the sequencing data is identified as having a long read length and a heterozygosity rate below a preset threshold, it can automatically select the haploid mode assembly tool and configure the corresponding redundancy removal parameters; or, when a high heterozygosity characteristic is identified, it switches to the genotyping assembly mode and adjusts the relevant algorithm parameters.

[0049] In actual implementation, the data type identification and strategy selection module 2031 can set the criteria for determining the sequence length distribution according to the actual situation. For example, it can be determined as HiFi data based on N50 greater than 15kb, or it can be determined as ONT Ultra-long data based on average read length greater than 30kb. This application embodiment does not make any special limitation on this. The assembly tool selected can also be HiFiasm, wtdbg2 or other genome assembly software suitable for specific data types.

[0050] The assembly quality assessment module 2032 is used to perform contamination detection, sequence consistency quality assessment, and assembly integrity assessment on the assembled gene sequences. In this embodiment, the assembly quality assessment module 2032 can utilize external databases or alignment tools (such as performing purge_haplotigs v1.1.3 for redundancy removal and identifying heterozygous contigs based on depth distribution) to calculate the proportion of non-target species sequences to complete contamination detection, and use short read data backfitting or K-mer consistency analysis to calculate QV values ​​to assess sequence consistency, and use indicators such as conserved single-copy homologous genes (BUSCO) to assess assembly integrity. Of course, the assembly quality assessment module 2032 can also summarize the above multiple assessment results into a quality score or status indicator, serving as the basis for determining whether subsequent processes should continue. Specifically, if the assessment result of the assembly quality assessment module 2032 shows that the quality meets the standards, it outputs a corresponding pass signal for downstream modules to use; if it does not meet the standards, it generates a quality report containing specific defect indicators. The specific values ​​for the quality compliance standards can be set according to the actual project requirements or species characteristics. For example, the integrity assessment threshold can be BUSCO integrity ≥90% or ≥95%. This application does not impose any special limitations on this.

[0051] The anomaly handling module 2033 is used to automatically generate an assessment report and trigger an external confirmation process when the assembly quality assessment module 2032 outputs an indication that the assembly quality has not met a preset standard. Specifically, the anomaly handling module 2033 is connected to the assembly quality assessment module 2032 and specifically responds to the latter's output of non-compliance status signals. For example, when the assembly quality assessment module 2032 detects anomalies such as N50 significantly lower than expected, insufficient BUSCO integrity, or excessive contamination rate, the anomaly handling module 2033 is immediately activated. It can then generate an assessment report containing key quality indicator comparisons, error type analysis, and recommended measures, and send a confirmation request to a designated user or management group through an integrated external communication interface (such as WeChat or DingTalk API). This allows for timely human decision-making, preventing erroneous data from flowing into subsequent chromosome mounting or gene annotation stages, and ensuring the accuracy of the final analysis results. Of course, the specific form of the external confirmation process triggered by the anomaly handling module 2033 can be a pop-up approval, email reply, or card confirmation within instant messaging software; this embodiment does not impose any special limitations on this.

[0052] The aforementioned chromosome mounting agent 204 is used to perform chromosome-level mounting, error correction, and integrity assessment on the assembled sequences output by the sequence assembly agent 203 using three-dimensional genomics data. For example, the chromosome mounting agent 204 can receive contig or scaffold sequences output by the sequence assembly agent 203 and, combined with input three-dimensional genomics data (such as Hi-C, Pore-C, etc.), perform clustering, sorting, and orientation operations by calculating the interaction frequency matrix between sequence fragments, thereby achieving chromosome-level mounting. Furthermore, the chromosome mounting agent 204 also has error correction capabilities, capable of identifying and correcting chimerism errors or orientation errors during the assembly process, filling sequence gaps using long-read data, and finally outputting a mounting quality assessment report.

[0053] like Figure 4 As shown, the chromosome mounting agent 204 may include a data alignment evaluation module 2041, an intelligent clustering mounting module 2042, a visual error correction and imputation module 2043, and a quality approval module 2044. The data alignment evaluation module 2041 is used to align the input three-dimensional genomics data to the assembled gene sequences and filter effective interaction pairs based on a preset alignment quality threshold, thereby extracting high-confidence spatial interaction information from the original three-dimensional data and providing basic data support for subsequent chromosome construction. In specific implementation, the data alignment evaluation module 2041 can map the input three-dimensional genomics data to the assembled gene sequences (Contig or Scaffold) using an alignment algorithm (such as minimap2), and filter the alignment results according to a preset alignment quality threshold (such as alignment score, mapping quality value MAPQ, alignment length ratio, etc.) to select effective interaction pairs.

[0054] In this application, the data comparison and evaluation module 2041 and the subsequent intelligent clustering mounting module 2042 have a close data flow relationship. The effective interaction pairs output by the module directly serve as the input data for the intelligent clustering mounting module 2042, guiding the clustering and sorting of sequences. Therefore, if the comparison quality does not meet the preset standard, the data comparison and evaluation module 2041 can determine that the current data is insufficient to support high-quality mounting, thereby terminating the subsequent process or triggering an exception handling mechanism. Depending on the actual application scenario, the preset comparison quality threshold can be set according to the actual situation. For example, it can be a mapping quality value greater than 10 and a comparison length greater than 150bp, or a sequence consistency greater than 90%. This application embodiment does not impose any special limitations on this.

[0055] The intelligent clustering mounting module 2042 is used to perform chromosome-level clustering and sorting orientation of gene sequences assembled based on effective interaction pairs, in order to achieve chromosome mounting. Specifically, the intelligent clustering mounting module 2042 can utilize the spatial proximity effect of chromosomes within the cell nucleus to cluster sequence fragments that are linearly far apart but spatially close into the same chromosome group, and determine their relative positions and orientations on the chromosome.

[0056] In this embodiment, the intelligent clustering mounting module 2042 receives valid interaction pairs from the data comparison and evaluation module, constructs an interaction matrix, and uses a clustering algorithm (e.g., hierarchical clustering, spectral clustering, or a deep learning-based clustering model) to perform chromosome-level partitioning of the assembled gene sequences. Furthermore, the intelligent clustering mounting module 2042 can sort and orient the sequences according to the decay law of interaction strength, outputting chromosome-level mounted sequences. For different genomic characteristics, the clustering algorithm and parameters used by the intelligent clustering mounting module 2042 can be set according to the actual situation; for example, it can be heatmap clustering based on contact frequency, or a community detection algorithm based on graph theory. This embodiment does not impose any special limitations on this.

[0057] It should be further explained that the intelligent clustering mounting module 2042 is also used to dynamically adjust the minimum number of connections required by the clustering algorithm based on the proportion of effective interaction pairs and the genome size of the assembled gene sequence. For example, it can be dynamically adjusted based on the formula min_links = max(5, round(10 × (1-v) ×log10(g / 1e6))), where min_links is the minimum number of connections required by the clustering algorithm; v is the proportion of effective interaction pairs; g is the genome size of the assembled gene sequence; max is the maximum value operator; and round is the rounding operator.

[0058] The visual error correction and gap-filling module 2043 is used to generate a visual interaction map based on effective interaction pairs to assist in identifying and correcting assembly errors, and to fill gaps in the gene sequences after chromosome mounting using long-read sequence data. This visual error correction and gap-filling module 2043 can provide a human-computer interaction interface within a visualization tool to display the interaction map and perform gap-filling operations. This visualization tool can be a mainstream tool such as Juicebox or HiGlass, or a custom-developed web-based visualization interface.

[0059] In practical applications, the visual error correction and imputation module 2043 can overcome the limitations of fully automated algorithms when handling complex repetitive regions or anomalous structures. It incorporates expert experience through visualization for correction and utilizes long-read data to improve sequence integrity. Specifically, based on the effective interaction pairs selected by the data comparison and evaluation module 2041, the visual error correction and imputation module 2043 can generate a two-dimensional interaction heatmap or contact map, visually displaying the distribution of interaction strength within and between chromosomes, assisting users in identifying assembly errors such as chimeras, inversions, and translocations. Simultaneously, the visual error correction and imputation module 2043 can call long-read sequence data (such as Nanopore or PacBio Ultra-long data) to perform local assembly and imputation of gaps in the sequence after chromosome mounting.

[0060] The quality approval module 2044 is used to calculate the chromosome anchoring rate and interaction strength distribution of the gene sequence after chromosome mounting to assess the mounting quality. Upon successful quality assessment, it triggers an external approval process containing mounting details. This quality approval module 2044 ensures that the mounting results meet preset quality standards and achieves enterprise-level compliance management by triggering external approval processes.

[0061] In practical applications, the quality approval module 2044 calculates key metrics of the gene sequence after chromosome mounting, including but not limited to chromosome anchoring rate (the proportion of sequences successfully mounted onto chromosomes) and interaction strength distribution (the ratio of intrachromosomal interaction strength to interchromosomal interaction strength). When these metrics meet preset standards, the quality approval module 2044 generates an approval request containing mounting details (such as chromosome length, anchoring rate, and interaction map snapshots) and sends it to designated personnel via integrated external communication interfaces (such as DingTalk and WeChat Work). Only after receiving an approval instruction will the system allow subsequent analysis processes (such as gene annotation) to continue.

[0062] The aforementioned gene annotation agent 205 is used to perform repetitive sequence annotation, gene structure prediction, and multi-database-based functional annotation on chromosome-level sequences output by chromosome-mounted agent 204. Specifically, gene annotation agent 205 can refer to a terminal analysis unit that interprets the biological meaning of chromosome-level sequences. It first identifies and masks repetitive sequences, then integrates homologous protein evidence, transcriptomic evidence, and de novo prediction results to accurately predict the structure of protein-coding genes (exon-intron boundaries, UTR regions, etc.), and performs functional annotation on the predicted genes based on multiple functional databases (such as GO, KEGG, SwissProt, etc.).

[0063] In practical applications, the gene annotation agent 205 functions as a transformation from sequence data to biological knowledge within the overall technical solution. Its output annotation results can be directly used for downstream comparative genomics analysis or breeding applications. Typically, the database source upon which the gene annotation agent 205 relies for functional annotation can be set according to actual research needs. For example, it can be a general database such as NR, TremBL, or InterPro, or a species-specific database. This application does not impose any special limitations on this.

[0064] In the embodiments of this application, such as Figure 5 As shown, the gene annotation agent 205 may specifically include a repetitive sequence annotation module 2051, a gene structure prediction module 2052, a functional annotation module 2053, and an annotation quality assessment module 2054. The repetitive sequence annotation module 2051 is used to annotate tandem repetitive and scattered repetitive sequences in chromosome-level gene sequences, thereby preprocessing the chromosome-level sequences output by the chromosome-mounted agent to shield the repetitive regions from interference with subsequent gene prediction. In this embodiment, there is a data flow relationship between the repetitive sequence annotation module 2051 and the gene structure prediction module 2052; that is, the repetitive sequence mask file output by the repetitive sequence annotation module 2051 serves as the input basis for the gene structure prediction module 2052.

[0065] In practical applications, the repetitive sequence annotation module 2051 can first use tandem repeat lookup tools (such as TRF) to identify tandem repeat sequences such as microsatellites, then use homology alignment tools (such as RepeatMasker) to identify scattered repeats based on a known repetitive sequence database, and then use de novo prediction tools (such as RepeatModeler) to construct a species-specific repetitive sequence library and perform supplementary annotation. Of course, other mature repetitive sequence identification algorithms or combinations can also be selected according to the characteristics of the actual sequencing data.

[0066] This gene structure prediction module 2052 is used to predict and integrate the structure of protein-coding genes from chromosome-level gene sequences based on homology alignment, de novo prediction, and transcriptome evidence. In practical applications, this gene structure prediction module 2052 can employ a multi-evidence fusion strategy, specifically including: integrating evidence from homologous species protein sequence alignment, evidence from de novo prediction based on hidden Markov models, and evidence from RNA-seq transcriptome assembly. For example, this gene structure prediction module 2052 can first run the homology alignment program, the de novo prediction program, and the transcriptome assembly program separately to obtain their respective gene structure prediction results. Then, it can use evidence integration software (such as MAKER2 or a self-developed integrator) to weight and merge the three types of evidence, prioritizing the retention of gene models that simultaneously possess protein homology, transcriptome support, and conform to the de novo prediction rules (i.e., genes supported by PHC evidence), thereby generating a high-confidence gene structure set. Specific integration weights or screening criteria can be set according to actual conditions; this application embodiment does not impose special limitations on this.

[0067] The functional annotation module 2053 can be used to perform comparative analysis of predicted gene structures using multiple functional databases to achieve gene functional annotation. In this application, the functional annotation module 2053 is a functional unit that assigns a biological function description to the predicted gene, thus placing the link after the gene structure prediction and aiming to resolve the potential function of the gene. In a specific implementation, the functional annotation module 2053 can utilize multiple functional databases to perform parallel comparative analysis of predicted protein sequences. For example, it can call tools such as Diamond or Blastp to compare with general protein databases such as NR and SwissProt to obtain homologous functional descriptions; call Kofamscan to compare with the KEGG database to obtain metabolic pathway information; call InterProScan to scan protein domains to obtain GO terms and pathway information; and it can also compare with transcription factor databases (such as AnimalTFDB), carbohydrate active enzyme databases (such as dbCAN), or transmembrane protein prediction databases (such as TMHMM) for specific needs. The selection and combination of the above databases can be flexibly adjusted according to research objectives, and this application does not impose any special limitations on this.

[0068] The annotation quality assessment module 2054 can be used to assess the integrity of gene annotations based on a preset integrity threshold and to statistically analyze the proportion of genes supported by different sources of evidence, thereby triggering an external review process that includes annotation details. Specifically, the annotation quality assessment module 2054 can be a functional unit used to quantitatively evaluate the reliability of gene annotation results and trigger a manual review process, ensuring that the output meets scientific or industry standards. Specifically, the annotation quality assessment module 2054 can assess the integrity of the annotation set based on a preset integrity threshold (such as the BUSCO integrity index); it can also statistically analyze the proportion of genes supported by different sources of evidence, for example, calculating the ratio of genes supported by only a single piece of evidence to genes supported by multiple pieces of evidence (such as PHC). When the proportion of genes supported by a single piece of evidence is detected to be too high or the overall integrity is below a preset threshold, the module can generate a quality report containing detailed statistical charts (such as an Upset plot or an evidence distribution pie chart) and trigger an external review process through an integrated external communication interface, requesting professional intervention for verification.

[0069] It should be further explained that the bioinformatics automated analysis system provided in this application embodiment may further include an enterprise communication and approval integration module 40 connected to each dedicated intelligent agent. This module is used to generate and send alarm messages or approval request messages at preset key decision nodes or anomaly detection nodes during the collaborative execution of each dedicated intelligent agent, through an integrated external communication application interface. It also receives approval result instructions through the external communication application interface and controls the continued execution, retry, or termination of subsequent analysis processes based on the approval result instructions.

[0070] In practical applications, the enterprise communication and approval integration module 40 can be a middleware service deployed at the edge or in the cloud of a bioinformatics automated analysis system. Its function is to act as a two-way communication bridge between the multi-agent analysis engine 20 and enterprise-level office collaboration platforms (such as DingTalk and WeChat Work). This module 40 can establish connections with the aforementioned genome survey agent 202, sequence assembly agent 203, chromosome mounting agent 204, and gene annotation agent 205, respectively, and monitor the status signals generated by each agent during operation in real time. When any dedicated intelligent agent detects a preset key decision node (e.g., assembly strategy selection point, mounting quality assessment point) or an anomaly detection node (e.g., exceeding quality control standards, excessive contamination rate) while performing a task, this module can be triggered. Its working principle involves intercepting structured event data generated by the agents, encapsulating it into a message body conforming to an external communication protocol, and pushing it to the designated enterprise communication terminal by calling a pre-configured application programming interface (API).

[0071] External communication application programming interfaces (APIs) can refer to open data interfaces provided by third-party enterprise instant messaging software or collaborative office platforms, such as DingTalk API, WeCom API, or other interfaces that support message push and workflow approval. The function of this interface is to convert the machine language state within the bioinformatics analysis system into human-readable alarm text or standardized approval forms. In specific implementations, this interface can support various message formats, including plain text notifications, rich text cards (Markdown), interactive buttons, or embedded approval links. Through this interface, the system can transform complex bioinformatics analysis anomalies (such as BUSCO integrity below the threshold, insufficient N50 length, etc.) into concise and clear work orders, which are then sent to the project manager's mobile terminal or desktop client, thereby achieving immediate access to technical issues.

[0072] Key decision-making nodes or anomaly detection nodes refer to predefined logical judgment points in the de novo genome assembly analysis workflow. Key decision-making nodes typically occur at stages where downstream parameters need to be dynamically adjusted based on upstream analysis results. For example, after genome feature prediction, a decision must be made on whether to use haploid or diploid assembly mode based on heterozygosity; or after chromosome mounting, it must be confirmed whether gap filling should be performed. Anomaly detection nodes, on the other hand, refer to stages where data quality or operational status deviates from the normal range. Examples include sequencing data quality control indicators (Q30, GC content) exceeding preset safety ranges, memory overflow errors occurring during assembly, or alignment rates falling below expected standards. At these nodes, the enterprise communication and approval integration module is activated, pausing the current automated workflow and initiating a manual intervention request.

[0073] In this way, when multiple agents collaboratively execute genome analysis tasks, each dedicated intelligent agent reports its operational status to the shared memory pool 201 in real time, and the enterprise communication and approval integration module 40 monitors these statuses synchronously. Once a specific agent is detected to have triggered a preset abnormal condition (such as the sequence assembly agent finding a contamination rate exceeding 1%), the enterprise communication and approval integration module 40 immediately sends an approval request message containing details of the abnormality to the designated project team through the integrated external communication application interface. After sending the message, the enterprise communication and approval integration module 40 enters a waiting state, suspending the automatic scheduling of subsequent tasks. When it receives the approval result instruction returned by the user through the interface, the module parses the instruction content and converts it into an internal control signal, which is then distributed to the corresponding intelligent agent. This drives the entire analysis process to branch and flow according to the user's decision, realizing closed-loop management from automatic detection to manual confirmation and then to automatic handling.

[0074] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0075] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A bioinformatics automated analysis system based on an AI large-scale model, characterized in that, include: The system includes a data input interface, a multi-agent analysis engine connected to the data input interface, and a heterogeneous resource scheduling module connected to the multi-agent analysis engine, wherein: The data input interface is used to receive sequencing data; The multi-agent analysis engine is driven by a large language model and includes multiple dedicated intelligent agents, which are used to collaboratively execute each step of the genome de novo assembly analysis process based on the sequencing data. The heterogeneous resource scheduling module is used to adapt and submit the analysis task instructions generated by each dedicated intelligent agent in the multi-agent analysis engine during the analysis process to the corresponding computing cluster scheduling system.

2. The bioinformatics automated analysis system according to claim 1, characterized in that, The multi-agent analysis engine also includes a shared memory pool connected to each dedicated intelligent agent. The shared memory pool is used to store the analysis results generated by any of the dedicated intelligent agents and to allow other dedicated intelligent agents to access them, so as to realize the flow and collaboration of analysis results among multiple dedicated intelligent agents.

3. The bioinformatics automated analysis system according to claim 1 or 2, characterized in that, The aforementioned specialized intelligent agents specifically include: genome survey agent, sequence assembly agent, chromosome mounting agent, and gene annotation agent, wherein: The genome survey agent is used to perform quality assessment, genome feature prediction, and comparative analysis with closely related species on the sequencing data; The sequence assembly agent is used to automatically select assembly strategies and parameters based on the analysis results of the genome survey agent, and to perform sequence assembly and assembly quality assessment. The chromosome mounting agent is used to perform chromosome-level mounting, error correction, and integrity assessment on the assembled sequences output by the sequence assembly agent using three-dimensional genomics data. The gene annotation agent is used to perform repetitive sequence annotation, gene structure prediction, and multi-database-based functional annotation on the chromosome-level sequences output by the chromosome mounting agent.

4. The bioinformatics automated analysis system according to claim 3, characterized in that, The genome survey agent specifically includes: An automated quality control module is used to assess the quality of sequencing data by comparing the quality control indicators of the sequencing data with multiple preset quality control thresholds. A genome feature prediction module is used to perform K-mer analysis on sequencing data that has passed quality assessment in order to predict the feature parameters of the target genome, wherein the feature parameters include at least one of genome size, heterozygosity and repetitive sequence ratio; The closely related species comparison analysis module is used to access external databases to obtain genomic feature data of at least one closely related species, compare and analyze the feature parameters estimated by the genomic feature prediction module with the genomic feature data of the closely related species, and output assembly parameter suggestions based on the comparison analysis results.

5. The bioinformatics automated analysis system according to claim 4, characterized in that, At least one of the preset quality control thresholds is dynamically calibrated based on the genomic characteristic data of closely related species obtained from the closely related species comparison analysis module.

6. The bioinformatics automated analysis system according to claim 3, characterized in that, The sequence assembly agent specifically includes: The data type identification and strategy selection module is used to identify the corresponding sequencing data type based on the sequence length distribution of the sequencing data, and, in combination with the sequencing data type and the genome feature analysis results provided by the genome survey agent, select the corresponding assembly tool and generate suitable assembly parameters for gene sequence assembly. The assembly quality assessment module is used to perform contamination detection, sequence consistency quality assessment, and assembly integrity assessment on the assembled gene sequences. The exception handling module is used to automatically generate an evaluation report and trigger an external confirmation process when the assembly quality assessment module outputs an indication that the assembly quality does not meet the preset standard.

7. The bioinformatics automated analysis system according to claim 3, characterized in that, The chromosome mounting agent specifically includes: The data comparison and evaluation module is used to compare the input three-dimensional genomics data with the assembled gene sequences and to screen effective pairs based on a preset comparison quality threshold. The intelligent clustering mounting module is used to perform chromosome-level clustering and sorting orientation of the gene sequences assembled based on the effective interaction pairs, so as to realize chromosome mounting. The visual error correction and filling module is used to generate a visual interaction map based on the effective interaction pairs to help identify and correct assembly errors, and to fill gaps in the gene sequence after chromosome mounting using long read sequence data; The quality approval module is used to calculate the chromosome anchoring rate and interaction strength distribution of the gene sequence after chromosome mounting to evaluate the mounting quality, and trigger an external approval process containing mounting details when the quality assessment is passed.

8. The bioinformatics automated analysis system according to claim 7, characterized in that, The intelligent clustering mounting module is also used to dynamically adjust the minimum number of connections required by the clustering algorithm based on the proportion of effective interaction pairs and the genome size of the assembled gene sequence.

9. The bioinformatics automated analysis system according to claim 3, characterized in that, The gene annotation proxy specifically includes: The repetitive sequence annotation module is used to annotate tandem repeats and scattered repeats in gene sequences at the chromosome level. The gene structure prediction module is used to predict and integrate the structure of protein-coding genes in chromosome-level gene sequences based on homology alignment, de novo prediction, and transcriptome evidence. The functional annotation module is used to compare and analyze the predicted gene structures using multiple functional databases to achieve gene function annotation. The annotation quality assessment module is used to evaluate the integrity of gene annotations based on a preset integrity threshold and to count the proportion of genes supported by different evidence, so as to trigger an external review process that includes annotation details.

10. The bioinformatics automated analysis system according to claim 1, characterized in that, The bioinformatics automated analysis system also includes an enterprise communication and approval integration module that is connected to each dedicated intelligent agent. This module is used to generate and send alarm messages or approval request messages at preset key decision nodes or anomaly detection nodes during the collaborative execution of each dedicated intelligent agent through an integrated external communication application interface. as well as, The system receives approval result instructions through the external communication application interface and controls the continuation, retry, or termination of subsequent analysis processes based on these instructions.

Images