Plant genome sequencing data analysis method and analysis system based on artificial intelligence

Abstract

The present application relates to the field of artificial intelligence technology, and in particular provides a plant genome sequencing data analysis method and analysis system Plantsyn based on artificial intelligence, which accurately constructs a genome expression description vector relationship network of a target plant object through a long short-term memory network, and is adapted to a variety of feature sets. The introduction of AI sliding sampling kernel technology realizes the accurate extraction of local description vector relationship networks from complex data, and constructs vector relationship network tuples based on data distribution characteristics, thereby improving the precision of data analysis. By integrating vector relationship network tuples, a genome sequencing integrated vector set is formed, thereby improving the sensitivity and accuracy of noise detection. In this way, the noise disturbance of the sequencing data set can be comprehensively evaluated, important data quality information can be provided to scientific researchers, and subsequent data cleaning and experimental design can be assisted, significantly improving the accuracy and reliability of genome sequencing data analysis. The method and system can also be applied to plant transcriptome data analysis.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based plant genome sequencing data analysis method and analysis system. Background Art

[0002] The application of artificial intelligence and neural network models in genome sequencing data analysis has become increasingly common. These advanced technologies, with their powerful data processing capabilities and efficient algorithmic accuracy, are gradually changing the way we understand and mine genomic data.

[0003] Artificial intelligence technologies, particularly deep learning, have been widely applied to the classification, prediction, and analysis of genomic data. They can extract useful biological information from massive amounts of sequencing data, helping researchers uncover the complex relationships between genes and phenotypes. Neural network models, with their exceptional pattern recognition and feature extraction capabilities, can uncover hidden patterns and associations in genomic data that are difficult to capture using traditional methods.

[0004] However, despite the tremendous potential of artificial intelligence and neural network models for genomic sequencing data analysis, several challenges and issues remain. Traditional sequencing data analysis methods, even when combined with these advanced technologies, remain limited by data processing capabilities and algorithmic precision, making it difficult to fully exploit the deep insights within genomic data. These limitations become particularly pronounced when faced with complex feature sets such as genomic expression levels, transcript structures, or epigenetic markers, often failing to meet the data depth and breadth required by modern scientific research.

[0005] Even more serious is the widespread problem of noise disturbance in sequencing data. This noise not only arises from technical limitations in the sequencing process but can also be introduced by multiple steps, including sample processing and data reading. The presence of noise seriously affects the accuracy and reliability of data, posing a major challenge that researchers must confront. Unfortunately, while artificial intelligence and neural network models have performed exceptionally well in data processing and analysis, current noise detection methods still need to be improved in accurately identifying and quantifying noise disturbances in sequencing data. Summary of the Invention

[0006] In order to improve the above problems, the present application provides a plant genome sequencing data analysis method and analysis system based on artificial intelligence.

[0007] The present invention provides an artificial intelligence-based plant genome sequencing data analysis method, which is applied to an artificial intelligence analysis system. The method includes:

[0008] Determining a first genome expression description vector relationship network of a first initial genome sequencing dataset and a second genome expression description vector relationship network of a second initial genome sequencing dataset using a long short-term memory network, wherein the first initial genome sequencing dataset and the second initial genome sequencing dataset correspond to a target plant object, and the first genome expression description vector relationship network and the second genome expression description vector relationship network are respectively any one of a genome expression feature set, a transcript structure feature set, or an epigenetic marker feature set;

[0009] Determine at least one local description vector relationship network from the first genome expression description vector relationship network and the second genome expression description vector relationship network based on the AI ​​sliding sampling kernel, and determine at least one vector relationship network binary based on the data distribution characteristics of each local description vector relationship network in the corresponding genome expression description vector relationship network, wherein the first local description vector relationship network and the second local description vector relationship network in each vector relationship network binary belong to different genome expression description vector relationship networks and have the same data distribution characteristics in the corresponding genome expression description vector relationship network;

[0010] Integrating the first local description vector relationship network and the second local description vector relationship network in each vector relationship network binary to obtain a genome sequencing integrated vector set, and determining sequencing noise perturbation results of the first local description vector relationship network and the second local description vector relationship network in the corresponding vector relationship network binary based on each genome sequencing integrated vector set;

[0011] Based on the sequencing noise perturbation results of each of the vector relationship network binary groups, the sequencing noise perturbation results of the first initial genome sequencing data set and the second initial genome sequencing data set are determined, and the sequencing noise perturbation results are the presence of sequencing noise perturbation or the absence of sequencing noise perturbation.

[0012] Furthermore, for the first local description vector relationship network and the second local description vector relationship network in each of the vector relationship network tuples, the first local description vector relationship network and the second local description vector relationship network are integrated to obtain a genome sequencing integrated vector set, including:

[0013] Generate a genome sequencing integrated vector set based on the sequencing attention variables of each characteristic element of the first local description vector relationship network and the second local description vector relationship network, wherein each characteristic element of the genome sequencing integrated vector set includes at least one attention dimension, and the sequencing attention variable of each attention dimension is determined by at least one of the sequencing attention variable corresponding to the characteristic element of the first local description vector relationship network or the sequencing attention variable corresponding to the characteristic element of the second local description vector relationship network;

[0014] Among them, the sequencing attention variable of each characteristic element of the genomic expression feature set is the genomic expression mapping parameter, the sequencing attention variable of each characteristic element of the transcript structure feature set is the transcript structure mapping parameter, and each characteristic element of the epigenetic mark feature set corresponds to multiple sequencing attention variables and each sequencing attention variable is the epigenetic mapping parameter of an epigenetic attention dimension.

[0015] Furthermore, the AI ​​sliding sampling kernel is used to determine at least one local description vector relationship network from the first genome expression description vector relationship network and the second genome expression description vector relationship network, respectively, including:

[0016] Determine at least one first target data distribution feature in the first genome expression description vector relationship network, and determine a data distribution feature in the second genome expression description vector relationship network that is identical to each of the first target data distribution features as a second target data distribution feature, where each of the first target data distribution features is used to reflect a core sequencing embedding vector of the first genome expression description vector relationship network;

[0017] At least one local description vector relationship network is determined from each of the first target data distribution features and each of the second target data distribution features based on the AI ​​sliding sampling kernel and the selected sampling strategy.

[0018] Furthermore, determining the sequencing noise perturbation results of the first initial genome sequencing dataset and the second initial genome sequencing dataset based on the sequencing noise perturbation results of each of the vector relationship network binary groups includes:

[0019] Determine at least one target data distribution feature cluster, each of the target data distribution feature clusters including one of the first target data distribution features and a second target data distribution feature corresponding to the first target data distribution feature;

[0020] Determine the first number of vector relationship network bigrams corresponding to each target data distribution feature cluster, and the second number of vector relationship network bigrams in which the sequencing noise perturbation result in the corresponding vector relationship network bigrams is no sequencing noise perturbation;

[0021] For each of the target data distribution feature clusters, determining sequencing noise perturbation results of the first target data distribution feature and the second target data distribution feature in the target data distribution feature cluster based on the first number and the second number corresponding to the target data distribution feature cluster;

[0022] Based on the sequencing noise perturbation results of each of the target data distribution feature clusters, the sequencing noise perturbation results of the first initial genome sequencing dataset and the second initial genome sequencing dataset are determined.

[0023] Furthermore, for each of the target data distribution feature clusters, determining the sequencing noise perturbation results of the first target data distribution feature and the second target data distribution feature in the target data distribution feature cluster based on the first number and the second number corresponding to the target data distribution feature cluster includes:

[0024] Determine a first ratio of the second number corresponding to the target data distribution feature cluster to the first number corresponding to the target data distribution feature cluster;

[0025] If the first proportion corresponding to the target data distribution feature cluster is less than the first threshold value, determining the noise disturbance analysis weight of the target data distribution feature cluster based on the first proportion corresponding to the target data distribution feature cluster and the second number, and determining the sequencing noise disturbance results of the first target data distribution feature and the second target data distribution feature in the target data distribution feature cluster based on the noise disturbance analysis weight;

[0026] If the first proportion corresponding to the target data distribution feature cluster is not less than the first threshold value, it is determined that the sequencing noise disturbance results of the first target data distribution feature and the second target data distribution feature in the target data distribution feature cluster are that there is no sequencing noise disturbance.

[0027] Furthermore, determining the sequencing noise perturbation results of the first initial genome sequencing dataset and the second initial genome sequencing dataset based on the sequencing noise perturbation results of each of the target data distribution feature clusters includes:

[0028] If the sequencing noise perturbation result of at least one of the target data distribution feature clusters is that there is no sequencing noise perturbation, determining that the sequencing noise perturbation results of the first initial genome sequencing dataset and the second initial genome sequencing dataset are that there is no sequencing noise perturbation;

[0029] If the sequencing noise perturbation results of each of the target data distribution feature clusters indicate the presence of sequencing noise perturbation, it is determined that the sequencing noise perturbation results of the first initial genome sequencing dataset and the second initial genome sequencing dataset indicate the presence of sequencing noise perturbation.

[0030] Furthermore, determining the sequencing noise perturbation results of the first local description vector relationship network and the second local description vector relationship network in the corresponding vector relationship network tuple based on each genome sequencing integrated vector set is achieved through a long short-term memory network, and the long short-term memory network is debugged by the following steps:

[0031] Determine a plurality of genome expression vector relationship network sample tuples, each of the genome expression vector relationship network sample tuples comprising a first genome expression description vector relationship network sample of a first initial genome sequencing dataset sample and a second genome expression description vector relationship network sample of a second initial genome sequencing dataset sample, the first initial genome sequencing dataset sample and the second initial genome sequencing dataset sample corresponding to a target plant object, and the first genome expression description vector relationship network sample and the second genome expression description vector relationship network sample are respectively any one of a genome expression feature set, a transcript structure feature set, or an epigenetic marker feature set;

[0032] Integrating the first genome expression description vector relationship network sample and the second genome expression description vector relationship network sample in each genome expression vector relationship network sample binary group to obtain a genome sequencing integrated vector set sample, inputting each genome sequencing integrated vector set sample into an initial recurrent neural network to obtain a discriminant viewpoint corresponding to each genome sequencing integrated vector set sample, and determining a sequencing noise perturbation prediction result of the first genome expression description vector relationship network sample and the second genome expression description vector relationship network sample in the genome expression vector relationship network sample binary group corresponding to each genome sequencing integrated vector set sample based on the discriminant viewpoint;

[0033] Based on the actual sequencing noise perturbation results and sequencing noise perturbation prediction results of the first genome expression description vector relationship network sample and the second genome expression description vector relationship network sample in each genome expression vector relationship network sample binary, a debugging error function is determined, and the initial recurrent neural network is debugged based on the debugging error function and each genome expression vector relationship network sample binary until the debugging is completed when the debugging error function meets the debugging termination requirements, and the recurrent neural network obtained after the debugging is determined as the long short-term memory network.

[0034] Furthermore, when the sequencing noise disturbance result indicates that sequencing noise disturbance exists, the method further includes:

[0035] Based on the sequencing noise disturbance result, denoising is performed on the first initial genome sequencing dataset and the second initial genome sequencing dataset.

[0036] Furthermore, the method further comprises:

[0037] The genome sequencing system of the target plant object is updated.

[0038] An embodiment of the present application provides an artificial intelligence analysis system, comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the above-mentioned method.

[0039] An embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the above method when executed.

[0040] In this embodiment, by utilizing a long short-term memory (LSTM) network, a first genome expression vector relationship network corresponding to a first initial genome sequencing dataset and a second genome expression vector relationship network corresponding to a second initial genome sequencing dataset of a target plant object can be accurately constructed. This network model is not only applicable to genomic expression feature sets, but can also effectively process transcript structure feature sets or epigenetic marker feature sets, demonstrating its broad applicability.

[0041] Secondly, the present embodiment innovatively introduces AI sliding sampling kernel technology. This technology accurately extracts local description vector relationship networks from complex genomic expression description vector relationship networks and further constructs vector relationship network tuples based on the similarity of data distribution characteristics. This process not only improves the precision of data analysis but also provides strong support for subsequent noise perturbation analysis.

[0042] Furthermore, the present embodiment integrates the local description vector relationship networks in the vector relationship network binary to form a genome sequencing integrated vector set. This step enables more accurate determination of sequencing noise perturbation results, thereby effectively improving the sensitivity and accuracy of noise detection.

[0043] Finally, by integrating the sequencing noise perturbation results for each vector relationship network binary, this embodiment of the present application can comprehensively assess the sequencing noise perturbation of the first and second initial genome sequencing datasets. This assessment result not only provides researchers with important information on data quality but also helps guide subsequent data cleaning and experimental design, thereby significantly improving the accuracy and reliability of genome sequencing data analysis.

[0044] In summary, the embodiments of the present application cleverly combine the long short-term memory network and the AI ​​sliding sampling kernel technology, bringing innovation to genome sequencing data analysis. Through the long short-term memory network, the genome expression description vector relationship network of the target plant object is accurately constructed, which is adaptable to a variety of feature sets. The introduction of AI sliding sampling kernel technology realizes the accurate extraction of local description vector relationship networks from complex data, and constructs vector relationship network tuples based on data distribution characteristics, thereby improving the precision of data analysis. Furthermore, by integrating the vector relationship network tuples, a genome sequencing integrated vector set is formed, thereby improving the sensitivity and accuracy of noise detection. In this way, the noise disturbance of the sequencing data set can be comprehensively evaluated, important data quality information can be provided to scientific researchers, and subsequent data cleaning and experimental design can be assisted, significantly improving the accuracy and reliability of genome sequencing data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A flowchart of an artificial intelligence-based plant genome sequencing data analysis method provided in an embodiment of the present application.

[0046] Figure 2 A schematic diagram of the structure of an artificial intelligence analysis system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to better understand the above technical solution, the technical solution of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations on the technical solution of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0048] Figure 1 A plant genome sequencing data analysis method based on artificial intelligence is shown, which is applied to an artificial intelligence analysis system. The method includes the following steps 110 to 140.

[0049] Step 110: Use a long short-term memory network to determine a first genome expression description vector relationship network of the first initial genome sequencing data set and a second genome expression description vector relationship network of the second initial genome sequencing data set.

[0050] The first initial genome sequencing dataset and the second initial genome sequencing dataset correspond to target plant objects, and the first genome expression description vector relationship network and the second genome expression description vector relationship network are respectively any one of a genome expression feature set, a transcript structure feature set or an epigenetic marker feature set.

[0051] In the embodiments of the present application, a long short-term memory network (LSTM) is a special recurrent neural network (RNN) that can learn long-term dependencies. In genomic sequencing data analysis, LSTM can be trained to identify complex patterns in sequence data, such as specific structural and functional regions in gene sequences. Through its unique gating mechanism, LSTM can effectively handle long-term dependencies in sequences, making it perform well when processing large-scale sequences such as genomic data.

[0052] A first-pass genome sequencing dataset is the data set generated from the first round of genome sequencing of a target plant. This data contains detailed information about the plant genome and serves as the foundation for subsequent analysis and comparison. The quality and completeness of this first-pass genome sequencing dataset are crucial for accurately elucidating the structure and function of plant genomes.

[0053] The first genome expression description vector relationship network is a data structure generated by processing the first initial genome sequencing dataset using specific algorithms (such as LSTM). It converts complex genomic data into vector form, facilitating mathematical operations and comparative analysis. This relationship network captures the expression level characteristics, transcript structure characteristics, and epigenetic marker characteristics of the genome, providing researchers with an intuitive and quantitative way to understand and compare genomic differences and similarities.

[0054] Second initial genome sequencing dataset: Similar to the first initial genome sequencing dataset, but can be sequenced at different time points, under different conditions, or from different parts of the plant. By comparing the first and second initial genome sequencing datasets, researchers can explore changes in gene expression in plants under different environments or developmental stages.

[0055] The second genome expression description vector relationship network is a vector relationship network built based on the second initial genome sequencing dataset. Similar to the first genome expression description vector relationship network, it also captures specific features of the genome, but reflects the information in the second initial genome sequencing dataset.

[0056] A target plant is a specific plant species or individual selected by researchers and is the focus of genome sequencing and analysis. By deeply studying the genome of a target plant, scientists can reveal its genetic characteristics, physiological functions, and adaptability to the environment.

[0057] Furthermore, the genomic expression feature set is a set of features that describe the expression level of genes under specific conditions. Expression refers to the amount of gene transcribed into mRNA, which reflects the activity level of the gene in the cell. The genomic expression feature set helps to understand which genes are activated or inhibited under specific conditions. The transcript structure feature set focuses on the structural information of gene transcripts, such as the boundaries and lengths of exons and introns. Transcript structure is crucial for understanding gene function and regulatory mechanisms because it determines the composition and splicing of protein coding sequences. The epigenetic mark feature set is a set of features that describe epigenetic modifications on the genome (such as DNA methylation, histone modifications, etc.). These modifications can affect gene expression without changing the DNA sequence. The epigenetic mark feature set is of great significance for analyzing the complex regulatory network of gene expression.

[0058] In some examples, the first genomic expression description vector relationship network is a high-dimensional data structure in which each gene or gene expression product is represented as a numerical feature vector. For example, if there are 1,000 genes, the expression level of each gene can be quantified as a specific numerical value, resulting in a 1,000-dimensional feature vector. Each element in this vector represents the expression level of the corresponding gene, and the magnitude of the numerical value reflects the intensity of gene expression. The relationship network formed by these vectors in space demonstrates the expression correlations and interactions between genes. Similar to the first genomic expression description vector relationship network, the second genomic expression description vector relationship network is also composed of numerical feature vectors. However, these vectors are derived from a second initial genomic sequencing dataset. Therefore, while the dimensionality and structure can be similar to the first relationship network, the numerical values ​​reflect the gene expression under different conditions or time points. By comparing these two relationship networks, it is possible to analyze the changes in gene expression under different environments or states. The genomic expression feature set is a collection of multiple numerical features, each representing the expression level of a gene. For example, a feature vector might look like this: [5.6, 3.2, 8.9, ..., 1.2], where each value represents the expression level of a specific gene. This feature set can help understand which genes are upregulated or downregulated under different conditions. The transcript structure feature set focuses on structural properties of transcripts, such as exon length and number of introns. These features can also be quantified into numerical values ​​and formed into a feature vector. For example, a transcript structure feature vector might include: [300, 4, 200, ...], where the values ​​represent the length of the first exon, the number of introns, the length of the second exon, and so on. These features help understand the complexity of transcripts and their potential functional impact. The epigenetic mark feature set contains a series of numerical features describing epigenetic modifications, such as DNA methylation levels and histone modification status. These features can also be represented as numerical feature vectors. For example, a DNA methylation feature vector might look like: [0.8, 0.3, 0.9, ...], where the values ​​represent the methylation levels at different gene loci. These features are crucial for understanding the epigenetic regulation of gene expression. These examples provide a powerful tool for quantifying and comparing genomic data, contributing to a deeper understanding of the complexity and function of the genome.

[0059] Based on the above, in step 110, the AI ​​analysis system begins its complex genomic data analysis process. First, the system utilizes the Long Short-Term Memory (LSTM) network, a powerful deep learning tool, to process the first initial genomic sequencing dataset. The LSTM network, with its unique memory units and gating mechanism, can capture long-term dependencies and complex patterns in genomic sequences, thereby generating a first genomic expression description vector network. This network accurately represents the expression characteristics, transcript structure characteristics, or epigenetic mark characteristics of the genome in the form of vectors, depending on the analysis objectives and requirements. Similarly, the AI ​​analysis system performs the same processing on the second initial genomic sequencing dataset. This dataset can be derived from samples from different tissues, developmental stages, or environmental conditions of the same plant. Through processing with the LSTM network, the system generates a second genomic expression description vector network, which also captures specific genomic features in the form of vectors. These two vector networks provide the basis for subsequent comparison and analysis. By comparing these two networks, researchers can explore changes in gene expression, differences in transcript structure, or the distribution of epigenetic marks in plants under different conditions. The accuracy and effectiveness of this step are crucial to the entire genome sequencing data analysis process because it provides high-quality data input and analysis foundation for subsequent steps.

[0060] Step 120: Determine at least one local description vector relationship network from the first genome expression description vector relationship network and the second genome expression description vector relationship network based on the AI ​​sliding sampling kernel, and determine at least one vector relationship network tuple based on the data distribution characteristics of each local description vector relationship network in the corresponding genome expression description vector relationship network.

[0061] The first local description vector relationship network and the second local description vector relationship network in each vector relationship network tuple belong to different genome expression description vector relationship networks and have the same data distribution characteristics in the corresponding genome expression description vector relationship networks.

[0062] In the embodiments of this application, the AI ​​sliding sampling kernel is an advanced data processing technology that can perform sliding sampling in large and complex datasets. In genomic analysis, the AI ​​sliding sampling kernel is used to effectively extract local data fragments, namely local description vector relationship networks, from the vast genomic expression description vector relationship network. This technology can capture subtle changes and important features in the dataset through intelligent sliding and sampling, providing strong support for subsequent data analysis and comparison.

[0063] The local description vector relationship network (LDVN) is a subset of data extracted from the entire genomic expression description vector relationship network using an AI sliding sampling kernel. It represents the characteristics and relationships of a specific region or segment in the original dataset. These LVNs can more precisely reflect the local characteristics and variation patterns of the genome, helping researchers to delve deeper into the complex structure and function of the genome.

[0064] Data distribution characteristics refer to the distribution and characteristics of data within a specific space or range. In genomics analysis, data distribution characteristics can reveal key information such as gene expression patterns, transcript abundance, and the distribution of epigenetic markers. By comparing the data distribution characteristics of different local description vector networks, researchers can identify similarities and differences across genomes, thereby gaining a deeper understanding of gene function and regulatory mechanisms.

[0065] A vector relationship network dyad is a pair of two local description vector relationship networks (LDVs) from different genomic expression description vector relationship networks. Although these two LVs are in different original datasets, their data distribution characteristics in their respective datasets are identical or highly similar. This construction of dyads facilitates cross-dataset comparisons and analyses, identifying commonalities and differences between different genomes.

[0066] The first local description vector relationship network is a local data segment extracted from the first genome expression description vector relationship network using an AI sliding sampling kernel. It represents the characteristics and relationships of a specific region in the first dataset, providing an important data foundation for subsequent comparison and analysis. Similar to the first local description vector relationship network, the second local description vector relationship network is a local data segment extracted from the second genome expression description vector relationship network. It reflects the characteristics and relationships of the corresponding region in the second dataset and, together with the first local description vector relationship network, forms a vector relationship network tuple for cross-dataset comparison and analysis.

[0067] In step 120, the artificial intelligence analysis system begins using its powerful AI sliding sampling kernel technology to conduct an in-depth analysis of the previously constructed first and second genome expression description vector networks. The goal of this step is to identify at least one local description vector network and discover similarities and differences between the data. First, the system uses the AI ​​sliding sampling kernel to perform sliding sampling within the first genome expression description vector network, intelligently identifying and extracting representative local data segments. These segments constitute the first local description vector network. This process accurately captures local features and subtle changes within the genome, providing a rich data foundation for subsequent comparative analysis. Similarly, the system performs a similar sliding sampling operation on the second genome expression description vector network, generating a second local description vector network. These two local description vector networks represent specific regions or features in two different datasets. Next, the system compares the data distribution characteristics of these local description vector networks in their respective original datasets. Through precise data analysis and comparison, the system identifies local description vector networks with highly similar data distributions and pairs them together to form vector network tuples. Each vector relationship network binary contains a first local description vector relationship network from the first genome expression description vector relationship network and a second local description vector relationship network from the second genome expression description vector relationship network. The key to this step lies in leveraging the intelligence and precision of AI sliding sampling kernel technology to extract valuable local information from complex genomic data and to identify commonalities and differences between different datasets by comparing data distribution characteristics. This lays a solid foundation for subsequent genome sequencing data integration and noise perturbation analysis.

[0068] Step 130: Integrate the first local description vector relationship network and the second local description vector relationship network in each vector relationship network binary to obtain a genome sequencing integrated vector set, and determine the sequencing noise perturbation results of the first local description vector relationship network and the second local description vector relationship network in the corresponding vector relationship network binary based on each genome sequencing integrated vector set.

[0069] In the present application embodiment, the genome sequencing integrated vector set is a comprehensive data set formed by integrating genome sequencing data from different sources or under different conditions. In this data set, each gene or gene expression product is represented as a numerical feature vector, and these vectors are combined to form a data set that comprehensively reflects the characteristics of the genome. This integration not only helps researchers obtain more comprehensive gene expression information, but also reveals the dynamic changes and regulatory mechanisms of gene expression by comparing data under different conditions.

[0070] Sequencing noise refers to the deviation between sequencing data and the true state caused by various factors during the genome sequencing process, such as sequencing instrument errors and sample processing variations. This perturbation can manifest as abnormal fluctuations in gene expression, misidentification of transcript structures, or inaccurate measurement of epigenetic marks. By analyzing sequencing noise, researchers can assess the accuracy and reliability of sequencing data, providing important insights for subsequent data analysis and interpretation.

[0071] Specifically, step 130 is a key step in the entire genome sequencing data analysis process. It involves integrating the vector relationship network dyads identified and paired in the previous steps and determining sequencing noise perturbations based on the integration results. First, the system traverses each vector relationship network dyad. These dyads are formed by pairing the first local description vector relationship network with the second local description vector relationship network, whose data distribution characteristics in their respective original datasets are identical or highly similar. For each dyad, the system integrates them using specific algorithms and techniques. This integration process is not a simple data merging process; rather, it takes into account the differences between different datasets and uses methods such as weighted averaging and data normalization to ensure that the integrated data truly reflects the characteristics of the original data. After integration is complete, the system obtains a genome sequencing integrated vector set. This vector set integrates information from different datasets and provides a more comprehensive and accurate description of genomic characteristics. Next, based on this integrated vector set, the system analyzes the first local description vector relationship network and the second local description vector relationship network in each dyad for sequencing noise perturbations. To analyze sequencing noise perturbations, the system uses statistical methods and machine learning algorithms to compare the differences between the integrated vector set and the original data. These differences can come from various noise factors in the sequencing process, such as instrument errors and sample preparation differences. The system quantitatively evaluates these differences and generates a sequencing noise perturbation result. This result not only helps researchers understand the accuracy and reliability of sequencing data, but also provides an important reference for subsequent data interpretation and experimental result verification. It can be seen that step 130 is a comprehensive data processing and analysis process. By integrating and comparing information from different data sets, it reveals the noise perturbation in genomic sequencing data and provides researchers with more accurate and comprehensive data support.

[0072] Step 140: Based on the sequencing noise perturbation results of each of the vector relationship network binary groups, determine the sequencing noise perturbation results of the first initial genome sequencing data set and the second initial genome sequencing data set, where the sequencing noise perturbation result indicates the presence of sequencing noise perturbation or the absence of sequencing noise perturbation.

[0073] In an embodiment of the present application, the goal of the system is to comprehensively determine whether the first initial genome sequencing data set and the second initial genome sequencing data set have sequencing noise disturbances based on the sequencing noise disturbance results of each vector relationship network binary obtained in the previous step.

[0074] First, the system aggregates and analyzes the sequencing noise perturbation results for all vector relationship network pairs. These results contain detailed information about how the local description of the vector relationship network differs from the original data during the integration process. The system carefully reviews these perturbation results, focusing specifically on pairs that show significant noise perturbation.

[0075] Next, the system analyzes these noise perturbation results. It calculates the proportion of pairs with noise perturbations, as well as the magnitude and distribution of these perturbations. This information helps the system comprehensively assess the sequencing quality of the entire dataset.

[0076] Based on this detailed noise perturbation analysis, the system then makes a comprehensive judgment. If the majority of vector relationship network bigrams exhibit significant noise perturbation, or if bigrams in certain key regions exhibit severe noise issues, the system may determine that the entire first or second initial genome sequencing dataset exhibits sequencing noise perturbation. Conversely, if only a few bigrams exhibit minor noise perturbation, while the majority of the data remains stable and consistent, the system may determine that the dataset does not exhibit significant sequencing noise perturbation.

[0077] Finally, the system generates a detailed report that includes the analysis results of sequencing noise perturbations and the system's comprehensive judgment. This report will provide researchers with valuable information to help them understand the quality and reliability of sequencing data, thereby guiding subsequent experimental design and data analysis.

[0078] As can be seen, step 140 is a key step in comprehensively assessing the noise perturbation of the sequencing data. By analyzing the noise perturbation results of each vector network binary, it provides researchers with an accurate assessment of the quality of the entire sequencing dataset, helping to ensure the accuracy and reliability of subsequent research.

[0079] In summary, in a complete plant genome study, an AI analysis system was used to deeply analyze and process two sets of genome sequencing data to detect potential sequencing noise. First, the AI ​​analysis system acquired a first and second initial genome sequencing dataset of the target plant. These datasets can be derived from different tissue samples from the same plant, or from sequencing the same tissue at different time points or under different conditions. The system utilized the Long Short-Term Memory (LSTM) network, a powerful deep learning model, to construct a first and second genome expression description vector relationship network, respectively. These networks can capture genomic expression characteristics, transcript structure characteristics, or epigenetic mark characteristics, depending on the analysis needs and objectives. Next, the AI ​​analysis system applied AI sliding sampling kernel technology to perform local sampling within the two genome expression description vector relationship networks. This step aims to identify and extract local description vector relationship networks with similar data distribution characteristics, forming vector relationship network tuples. The two local description vector relationship networks in each tuple come from different initial genome sequencing datasets, but exhibit the same data distribution characteristics within their respective datasets. Subsequently, the system integrates the local description vector relationship networks in each vector relationship network binary to generate a genome sequencing integrated vector set. This process helps to comprehensively compare the corresponding parts in the two initial data sets, thereby more accurately identifying sequencing noise. By comparing and analyzing these integrated vector sets, the artificial intelligence analysis system can determine the sequencing noise perturbation results in each vector relationship network binary. Finally, based on the sequencing noise perturbation analysis results of all vector relationship network binary groups, the artificial intelligence analysis system draws an overall sequencing noise perturbation conclusion about the first initial genome sequencing data set and the second initial genome sequencing data set. This conclusion clearly indicates whether there is sequencing noise perturbation in the data set, providing researchers with valuable information to help them evaluate data quality and guide subsequent experimental design and data analysis. In this way, the artificial intelligence analysis system not only improves the efficiency and accuracy of genome sequencing data processing, but also assists in in-depth understanding and exploration of the complexity and dynamics of plant genomes.

[0080] The embodiment of the present application provides an accurate and efficient analysis method for the genome sequencing data of the target plant object by innovatively combining the long short-term memory network with the AI ​​sliding sampling kernel technology. Specifically, through the long short-term memory network, the first genome expression description vector relationship network of the first initial genome sequencing data set and the second genome expression description vector relationship network of the second initial genome sequencing data set can be accurately constructed, and whether it is a genome expression feature set, a transcript structure feature set or an epigenetic marker feature set, they can all be effectively expressed. Furthermore, with the help of AI sliding sampling kernel technology, it is possible to accurately extract a local description vector relationship network from a complex vector relationship network, and construct a vector relationship network binary based on the similarity of the data distribution characteristics. This step not only improves the meticulousness of data analysis, but also lays a solid foundation for subsequent noise perturbation analysis.

[0081] In addition, another significant advantage of the embodiment of the present application is that, by integrating the local description vector relationship network in the vector relationship network binary, a genome sequencing integrated vector set is formed, so that the sequencing noise disturbance result can be determined more accurately. This method not only improves the sensitivity of noise detection, but also provides a comprehensive assessment of the quality of sequencing data. Ultimately, by integrating the sequencing noise disturbance results of each vector relationship network binary, it is possible to accurately determine whether the first initial genome sequencing data set and the second initial genome sequencing data set have sequencing noise disturbances, providing strong data support for scientific researchers and helping to improve the accuracy and reliability of genome sequencing data.

[0082] In some optional embodiments, for the first local description vector relationship network and the second local description vector relationship network in each of the vector relationship network tuples, the first local description vector relationship network and the second local description vector relationship network are integrated to obtain a genome sequencing integrated vector set, including: generating a genome sequencing integrated vector set based on the sequencing attention variables of each characteristic element of the first local description vector relationship network and the second local description vector relationship network, each characteristic element of the genome sequencing integrated vector set includes at least one attention dimension, and the sequencing attention variable of each attention dimension is determined by at least one of the sequencing attention variables corresponding to the characteristic element of the first local description vector relationship network or the sequencing attention variables corresponding to the characteristic element of the second local description vector relationship network; wherein, the sequencing attention variable of each characteristic element of the genome expression feature set is a genome expression mapping parameter, the sequencing attention variable of each characteristic element of the transcript structure feature set is a transcript structure mapping parameter, and each characteristic element of the epigenetic mark feature set corresponds to multiple sequencing attention variables and each sequencing attention variable is an epigenetic mapping parameter of an epigenetic attention dimension.

[0083] Based on this embodiment, the AI ​​analysis system uses a unique method to integrate the first local description vector relationship network and the second local description vector relationship network in each vector relationship network tuple to obtain the genome sequencing integrated vector set. The key to this process is the use of the sequencing attention variable, which is a metric that measures the importance of each feature element in the genome sequencing process.

[0084] First, the system examines each characteristic element in the first and second local description vector networks. These characteristic elements can represent genomic expression levels, transcript structure, or epigenetic markers, which are all important components of biological genetic information.

[0085] Next, the system calculates one or more sequencing attention variables for each feature element. These variables are essentially mapping parameters that reflect which feature elements are of greater concern and which are relatively less important during the sequencing process. Specifically, if it is a genomic expression feature set, the sequencing attention variable for each feature element is the genomic expression mapping parameter; if it is a transcript structure feature set, then the sequencing attention variable for each feature element is the transcript structure mapping parameter; and for the epigenetic mark feature set, due to its complexity, each feature element will correspond to multiple sequencing attention variables, and each variable represents a mapping parameter for an epigenetic attention dimension.

[0086] After determining these sequencing attention variables, the AI ​​analysis system uses them to generate an integrated genome sequencing vector set. Each feature element in this integrated vector set contains at least one attention dimension, and the sequencing attention variable for each attention dimension is determined by the sequencing attention variable of the corresponding feature element in the first or second local description vector relationship network. This processing approach not only preserves the richness of the original data but also, by introducing an attention mechanism, enables the system to focus more closely on important feature elements when processing genome sequencing data.

[0087] In this way, the AI ​​analysis system can more effectively mine deep information from genome sequencing data and improve its ability to process complex genomic feature sets. At the same time, due to the introduction of the attention mechanism, the system can also demonstrate greater robustness in the face of noise perturbations, because it can more accurately identify and focus on those characteristic elements that have a greater impact on the analysis results. This not only improves the accuracy and efficiency of the AI ​​analysis system in processing genome sequencing data, but also enhances the system's stability and reliability in the face of complex data and noise perturbations. This is a major technological breakthrough in the application of AI in the field of genome sequencing data analysis, and is expected to provide researchers with more accurate and efficient data analysis tools, promoting the in-depth development of genomic research.

[0088] Under some other preferred design ideas, the AI ​​sliding sampling kernel is used to determine at least one local description vector relationship network from the first genome expression description vector relationship network and the second genome expression description vector relationship network, including: determining at least one first target data distribution feature in the first genome expression description vector relationship network, and determining the data distribution feature in the second genome expression description vector relationship network that is identical to each first target data distribution feature as a second target data distribution feature, and each first target data distribution feature is used to reflect the core sequencing embedding vector of the first genome expression description vector relationship network; based on the AI ​​sliding sampling kernel and the selected sampling strategy, at least one local description vector relationship network is determined from each first target data distribution feature and each second target data distribution feature.

[0089] Under this design concept, the AI ​​analysis system uses a refined approach to identify at least one local description vector relationship network from the first and second genome expression description vector relationship networks. This process involves identifying and matching data distribution features and extracting local descriptions based on an AI sliding sampling kernel.

[0090] First, the system conducts an in-depth analysis of the first genome expression description vector network to identify at least one first target data distribution feature. These data distribution features reflect the core information and key patterns in the genomic sequencing data, acting as a kind of "fingerprint" or "marker" in the data. Using complex algorithms, the system identifies these features, which represent the core sequencing embedding vectors in the vector network and serve as an important foundation for subsequent analysis.

[0091] Next, the system searches for data distribution features in the second genome expression description vector relationship network that match the first target data distribution features and identifies them as the second target data distribution features. This process is similar to finding a common "marker" or "fingerprint" in two sets of data, ensuring that the local description vector relationship networks extracted from the two vector relationship networks have similar or identical data characteristics, allowing for effective comparison and analysis.

[0092] After determining the target data distribution characteristics, the AI ​​analysis system uses an AI sliding sampling kernel and a selected sampling strategy to extract local descriptive vector networks from these characteristics. The AI ​​sliding sampling kernel is an advanced data processing technique that slides across the entire dataset, capturing local features based on the selected sampling strategy. This allows the system to accurately extract local descriptive vector networks that match the target data distribution characteristics, providing an accurate data foundation for subsequent noise detection and data analysis.

[0093] This not only improves the accuracy of data analysis but also enables the system to more efficiently process complex genomic sequencing data. It ensures that the local description vector relationship networks extracted from two different genomic expression description vector relationship networks are highly similar and comparable, greatly improving the accuracy and reliability of subsequent noise detection and data analysis. Through this approach, the AI ​​analysis system can more deeply explore the hidden information and association patterns in genomic sequencing data, providing researchers with more comprehensive and accurate data analysis results.

[0094] In the next step, the sequencing noise perturbation results of the first initial genome sequencing data set and the second initial genome sequencing data set are determined based on the sequencing noise perturbation results of each of the vector relationship network binary groups, including: determining at least one target data distribution feature cluster, each of the target data distribution feature clusters including one of the first target data distribution features and a second target data distribution feature corresponding to the first target data distribution feature; determining the first number of the vector relationship network binary groups corresponding to each of the target data distribution feature clusters, and the second number of the vector relationship network binary groups whose sequencing noise perturbation results in the corresponding vector relationship network binary groups are the ones without sequencing noise perturbation; for each of the target data distribution feature clusters, based on the first number and the second number corresponding to the target data distribution feature cluster, determining the sequencing noise perturbation results of the first target data distribution feature and the second target data distribution feature in the target data distribution feature cluster; based on the sequencing noise perturbation results of each of the target data distribution feature clusters, determining the sequencing noise perturbation results of the first initial genome sequencing data set and the second initial genome sequencing data set.

[0095] Based on this embodiment, the AI ​​analysis system performs a series of sophisticated operations to determine the sequencing noise perturbation results for the first and second initial genome sequencing datasets. This process involves identifying target data distribution feature clusters, counting vector relationship network tuples, and determining the sequencing noise perturbation results based on these data.

[0096] First, the system identifies at least one target data distribution feature cluster. These feature clusters are composed of first target data distribution features and their corresponding second target data distribution features. Simply put, each feature cluster contains similar or matching data distribution features from two initial genome sequencing datasets, providing a basis for subsequent comparison and analysis.

[0097] Next, the system counts the number of vector network bigrams corresponding to each target data distribution feature cluster (the first number). Simultaneously, the system also counts the number of bigrams within these bigrams where sequencing noise perturbation results indicate no sequencing noise perturbation (the second number). These two statistics are crucial for subsequent assessment of sequencing noise perturbation.

[0098] Then, for each target data distribution feature cluster, the system determines the sequencing noise perturbation results for the first and second target data distribution features within that feature cluster based on the corresponding first and second numbers. Specifically, if the ratio of the second number to the first number is high, it indicates that most vector relationship network dyads in that feature cluster are not perturbed by sequencing noise, and therefore the sequencing noise perturbation result for that feature cluster can be judged to be good. Conversely, a low ratio indicates that the sequencing noise perturbation is more severe.

[0099] Finally, based on the sequencing noise perturbation results for each target data distribution feature cluster, the system comprehensively assesses the sequencing noise perturbation results for the first and second initial genome sequencing datasets. This process is similar to a "summary scoring" process, where the system comprehensively considers the performance of each feature cluster and provides an overall sequencing noise perturbation assessment result.

[0100] Through this method, the artificial intelligence analysis system can comprehensively and accurately evaluate the sequencing noise perturbation of two initial genome sequencing data sets. This not only helps researchers understand the quality and reliability of the data, but also provides an important reference for subsequent data cleaning, experimental design, and data analysis. At the same time, this evaluation method based on target data distribution feature clusters and vector relationship network tuples is also highly versatile and flexible, and can be widely used in noise detection and quality control scenarios for various genome sequencing data. In other words, through this series of refined operations and analyses, not only the accuracy and sensitivity of sequencing noise perturbation detection are improved, but also a new solution is provided for the quality control of genome sequencing data.

[0101] In some exemplary embodiments, for each of the target data distribution feature clusters, determining the sequencing noise perturbation results of the first target data distribution feature and the second target data distribution feature in the target data distribution feature cluster based on the first number and the second number corresponding to the target data distribution feature cluster includes: determining a first ratio of the second number corresponding to the target data distribution feature cluster to the corresponding first number; if the first ratio corresponding to the target data distribution feature cluster is less than a first threshold value, determining the noise perturbation analysis weight of the target data distribution feature cluster based on the first ratio and the second number corresponding to the target data distribution feature cluster, and determining the sequencing noise perturbation results of the first target data distribution feature and the second target data distribution feature in the target data distribution feature cluster based on the noise perturbation analysis weight; if the first ratio corresponding to the target data distribution feature cluster is not less than the first threshold value, determining the sequencing noise perturbation results of the first target data distribution feature and the second target data distribution feature in the target data distribution feature cluster as the absence of sequencing noise perturbation.

[0102] Based on this embodiment, the AI ​​analysis system uses a specific method to determine the sequencing noise perturbation results for the first target data distribution feature and the second target data distribution feature in each target data distribution feature cluster. This process combines statistical data with preset thresholds to ensure the accuracy and reliability of the results.

[0103] First, the system determines the ratio of the second number to the first number corresponding to each target data distribution feature cluster. This ratio is called the first fraction. This ratio reflects the proportion of vector relationship network tuples in that feature cluster that are not perturbed by sequencing noise and is an important indicator for assessing sequencing noise perturbations.

[0104] Next, the system determines whether the first ratio is less than a preset first threshold value. This threshold value is set based on experience or actual needs and is used to distinguish whether sequencing noise disturbance exists.

[0105] If the first proportion is less than the first threshold, it indicates that there are many vector relationship network dyads in the target data distribution feature cluster that are perturbed by sequencing noise, and it cannot be simply assumed that there is no sequencing noise perturbation. In this case, the system determines a noise perturbation analysis weight based on the first proportion and the second number. This weight reflects the degree to which the feature cluster is perturbed by sequencing noise; the larger the weight, the more severe the perturbation. The system then uses this weight to determine the sequencing noise perturbation results for the first target data distribution feature and the second target data distribution feature in the feature cluster. This result takes into account the weight and other relevant factors to provide a quantitative evaluation indicator.

[0106] If the first proportion is not less than the first threshold value, it means that most of the vector relationship network tuples in the target data distribution feature cluster are not disturbed by sequencing noise. Therefore, the system can determine that the sequencing noise disturbance results of the first target data distribution feature and the second target data distribution feature in the feature cluster are that there is no sequencing noise disturbance.

[0107] Through this method, the AI ​​analysis system can provide different sequencing noise perturbation assessment results for different situations, which are more in line with actual conditions. At the same time, this method also has a certain degree of flexibility and scalability, and can be adjusted and optimized according to different needs. It can be seen that this sequencing noise perturbation assessment method based on the first proportion and noise perturbation analysis weights improves the accuracy and reliability of the assessment, helps researchers better understand the quality and reliability of genome sequencing data, and provides strong support for subsequent data analysis and research.

[0108] In some possible embodiments, the sequencing noise perturbation results of the first initial genome sequencing dataset and the second initial genome sequencing dataset are determined based on the sequencing noise perturbation results of each of the target data distribution feature clusters, including: if the sequencing noise perturbation result of at least one of the target data distribution feature clusters is that there is no sequencing noise perturbation, determining that the sequencing noise perturbation results of the first initial genome sequencing dataset and the second initial genome sequencing dataset are that there is no sequencing noise perturbation; if the sequencing noise perturbation results of each of the target data distribution feature clusters are that there is sequencing noise perturbation, determining that the sequencing noise perturbation results of the first initial genome sequencing dataset and the second initial genome sequencing dataset are that there is sequencing noise perturbation.

[0109] In this example, the AI ​​analysis system comprehensively assesses the sequencing noise perturbations of the first and second initial genome sequencing datasets based on the sequencing noise perturbation results for each target data distribution feature cluster. This process utilizes a comprehensive and meticulous evaluation approach to ensure the accuracy and reliability of the final conclusions.

[0110] First, the system examines the sequencing noise perturbation results for all target data distribution feature clusters. These feature clusters were previously identified through a series of complex analyses, and each feature cluster represents a specific data distribution pattern in the dataset. Therefore, the noise perturbation analysis of these feature clusters directly reflects the noise situation of the entire dataset.

[0111] If the system finds that the sequencing noise perturbation result for at least one target data distribution feature cluster is not affected by sequencing noise, it will determine that the sequencing noise perturbation results for the first and second initial genome sequencing datasets are not affected by sequencing noise. This is because as long as there is one feature cluster that is not affected by noise, it indicates that at least a portion of the data in the entire dataset is reliable and not affected by noise.

[0112] On the other hand, if the sequencing noise perturbation results for all target data distribution feature clusters indicate the presence of sequencing noise perturbation, the system determines that the sequencing noise perturbation results for both the first and second initial genome sequencing datasets are present. This means that the entire dataset is extensively affected by noise, and no single component is completely reliable.

[0113] This comprehensive assessment approach enables the AI ​​analysis system to provide a clear conclusion about the sequencing noise perturbation of the entire dataset. This conclusion not only considers individual components of the dataset but is also based on a detailed feature cluster analysis, resulting in a high degree of accuracy and confidence.

[0114] In some alternative embodiments, the sequencing noise perturbation results of the first local description vector relationship network and the second local description vector relationship network in the corresponding vector relationship network tuple determined based on each of the genome sequencing integrated vector sets are achieved through a long short-term memory network, and the long short-term memory network is debugged by the following steps: determining multiple genome expression vector relationship network sample tuples, each of the genome expression vector relationship network sample tuples includes a first genome expression description vector relationship network sample of a first initial genome sequencing data set sample and a second genome expression description vector relationship network sample of a second initial genome sequencing data set sample, the first initial genome sequencing data set sample and the second initial genome sequencing data set sample correspond to target plant objects, the first genome expression description vector relationship network sample and the second genome expression description vector relationship network sample are respectively any one of a genome expression feature set, a transcript structure feature set or an epigenetic mark feature set; the first genome expression description vector relationship network sample in each of the genome expression vector relationship network sample tuples is converted into a sequence of the first genome expression description vector relationship network sample and the second genome expression description vector relationship network sample. The genome expression vector relationship network sample and the second genome expression description vector relationship network sample are integrated to obtain a genome sequencing integrated vector set sample, and each of the genome sequencing integrated vector set samples is input into the initial recurrent neural network to obtain a discriminant viewpoint corresponding to each of the genome sequencing integrated vector set samples, and the sequencing noise perturbation prediction results of the first genome expression description vector relationship network sample and the second genome expression description vector relationship network sample in the genome expression vector relationship network sample binary group corresponding to each of the genome sequencing integrated vector set samples are determined based on the discriminant viewpoint; based on the actual sequencing noise perturbation results and the sequencing noise perturbation prediction results of the first genome expression description vector relationship network sample and the second genome expression description vector relationship network sample in each of the genome expression vector relationship network sample binary groups, a debugging error function is determined, and the initial recurrent neural network is debugged based on the debugging error function and each of the genome expression vector relationship network sample binary groups until the debugging is completed when the debugging error function meets the debugging termination requirements, and the recurrent neural network obtained after the debugging is determined as the long short-term memory network.

[0115] Based on this embodiment, the artificial intelligence analysis system uses a long short-term memory (LSTM) network to determine the sequencing noise perturbation results for corresponding vector relationship network pairs (i.e., the first local description vector relationship network and the second local description vector relationship network) in the genome sequencing integrated vector set. LSTM networks are a special type of recurrent neural network (RNN) that can efficiently process sequence data and remember long-term dependencies, thus offering significant advantages in processing genome sequencing data.

[0116] To train this LSTM network, the system first identifies multiple example pairs of genomic expression vector relationship networks. These example pairs are extracted from the genome sequencing data of the target plant. Each example pair includes a first genomic expression description vector relationship network example from the first initial genome sequencing dataset and a second genomic expression description vector relationship network example from the second initial genome sequencing dataset. These examples can be any of a genomic expression feature set, a transcript structure feature set, or an epigenetic mark feature set, representing different levels and characteristics of the genome.

[0117] Next, the system integrates the first and second genome expression description vector relationship network samples in these sample pairs to obtain genome sequencing integrated vector set samples. These integrated vector set samples are then input into an initial recurrent neural network to obtain a discriminant view corresponding to each sample. These discriminant views are the network's initial judgment on whether each sample is perturbed by sequencing noise.

[0118] Based on these discriminant viewpoints, the system further determines the sequencing noise perturbation prediction results for the first and second genome expression description vector relationship network examples in the genome expression vector relationship network example pair corresponding to each genome sequencing ensemble vector set example. These prediction results are the network's specific predictions of sequencing noise perturbation and are compared with the actual results.

[0119] To train and optimize the network, the system determines a tuning error function based on the actual sequencing noise perturbation results and the predicted results of the first and second genomic expression description vector relationship network samples in each sample pair. This function measures the difference between the network's predictions and the actual results and is a key metric for optimizing network performance.

[0120] By continuously using these sample pairs and the debugging error function to debug and optimize the initial recurrent neural network until the debugging error function meets the preset debugging termination requirements (such as the error rate is lower than a certain threshold or the maximum number of iterations is reached), the system will eventually determine the recurrent neural network that has completed debugging as the required long short-term memory network.

[0121] This not only improves the accuracy and efficiency of analysis, but also enables the system to process complex genomic sequencing data and effectively identify and predict sequencing noise perturbations. This is of great significance for subsequent data cleaning, experimental design, and data analysis.

[0122] In some exemplary embodiments, when the sequencing noise disturbance result indicates that sequencing noise disturbance exists, the method further includes: performing denoising processing on the first initial genome sequencing dataset and the second initial genome sequencing dataset based on the sequencing noise disturbance result.

[0123] In this embodiment, the artificial intelligence analysis system demonstrated its powerful intelligent analysis capabilities when processing genome sequencing data. When the system detected the presence of sequencing noise perturbations, it took a series of precise measures to remove noise from the first and second initial genome sequencing datasets.

[0124] First, the AI ​​analysis system carefully analyzes the source and nature of sequencing noise. This noise can arise from errors in the sequencing instrument itself, contamination during sample processing, or interference during data reading. The system uses advanced algorithms, such as deep learning models, to identify and distinguish this noise from true genomic sequencing signals.

[0125] Next, the system uses sophisticated statistical methods and machine learning techniques to clean the data. This may include using filters to remove high-frequency noise or using modeling to predict and correct sequencing errors. During the noise removal process, the AI ​​analysis system pays special attention to protecting the integrity and accuracy of the genomic data, ensuring that the noise-free data truly reflects the genomic information of the sample.

[0126] For example, if the system detects an abnormal gene sequence read, it will compare it with the known genome database to confirm whether the read is due to sequencing noise. If it is confirmed to be noise, the system will remove it from the data set; if not, the system will retain the read to ensure data integrity.

[0127] After denoising the first and second initial genome sequencing datasets, the AI ​​analysis system generates a more accurate and reliable genome sequencing dataset. This optimized dataset can be used for subsequent genomic research, such as gene variation analysis and disease risk assessment.

[0128] Through this noise removal process, the AI ​​analysis system not only improves the accuracy of genome sequencing data but also provides researchers with a higher-quality data foundation, further promoting the research and application of genomics. This precise data processing method is crucial for improving the accuracy of medical diagnoses, the efficiency of drug development, and the realization of personalized medicine.

[0129] In some independent embodiments, the denoising process of the first initial genome sequencing dataset and the second initial genome sequencing dataset includes: combining the sequencing noise disturbance result, the first initial genome sequencing dataset, and the second initial genome sequencing dataset to obtain a sequencing technology noise detection vector and a sequencing biological noise detection vector for the target plant object, wherein the sequencing technology noise detection vector is used to characterize the sequencing technology noise factor in the target plant object, and the sequencing biological noise detection vector is used to characterize the sequencing biological noise factor in the target plant object; obtaining a first linkage noise positioning vector for the sequencing technology noise detection vector based on a noise detection vector in the sequencing biological noise detection vector that is related to the sequencing technology noise detection vector. The first linkage noise localization vector is used to represent a sequencing technology noise detection vector that integrates a sequencing biological noise detection vector; for the sequencing biological noise detection vector, based on a noise detection vector in the sequencing technology noise detection vector that is related to the sequencing biological noise detection vector, a second linkage noise localization vector is obtained, and the second linkage noise localization vector is used to represent the sequencing biological noise detection vector that integrates the sequencing technology noise detection vector; the first linkage noise localization vector and the second linkage noise localization vector are fused to obtain a global noise localization vector; based on the global noise localization vector, a noise removal viewpoint for the target plant object is determined; and the first initial genome sequencing data set and the second initial genome sequencing data set are denoised according to the noise removal viewpoint.

[0130] In this embodiment, the AI ​​analysis system performs noise removal on the first and second initial genome sequencing datasets. This process is highly sophisticated and technically demanding. This process involves complex vector operations and noise identification techniques to ensure the accuracy and effectiveness of noise removal.

[0131] First, the system combines the sequencing noise perturbation results with the first and second initial genome sequencing datasets to obtain a sequencing technical noise detection vector and a sequencing biological noise detection vector for the target plant. These two vectors form the basis for noise removal, characterizing the sequencing technical noise and sequencing biological noise factors, respectively, in the target plant. Simply put, sequencing technical noise can arise from errors or improper operation of the sequencing instrument, while sequencing biological noise can be related to biological processes such as sample collection and storage.

[0132] Next, the system performs linked noise localization. For the sequencing technical noise detection vector, the system generates a first linked noise localization vector based on the relevant portion of the sequencing biological noise detection vector. This vector is a comprehensive representation of the sequencing technical noise, taking into account the sequencing biological noise factor. Similarly, for the sequencing biological noise detection vector, the system also generates a second linked noise localization vector based on the relevant portion of the sequencing technical noise detection vector.

[0133] After obtaining these two linked noise localization vectors, the system merges them to produce a global noise localization vector. This global noise localization vector comprehensively considers noise factors from both sequencing technology and biology, providing precise guidance for subsequent noise removal.

[0134] Based on this global noise localization vector, the system determines a noise removal perspective for the target plant object. This noise removal perspective is actually a comprehensive noise removal strategy that takes into account the source, nature, and impact of various noises to ensure targeted and effective noise removal.

[0135] Finally, the system denoises the first and second initial genome sequencing datasets based on this denoising perspective. This process may include data cleaning, correction, and reconstruction to ensure that the denoised datasets more accurately reflect the genomic information of the target plant.

[0136] Through this noise removal process, the AI ​​analysis system not only improves the accuracy and reliability of genome sequencing data but also provides a higher-quality data foundation for subsequent genomic research and applications. The implementation of this noise removal technology is crucial for advancing genomic research, improving the accuracy of medical diagnoses, and promoting the development of personalized medicine.

[0137] In some other possible embodiments, the method further includes: updating the genome sequencing system of the target plant object.

[0138] In other possible embodiments, the AI ​​analysis system not only focuses on denoising the genome sequencing data but also further updates the genome sequencing system of the target plant. This step is critical to ensure that the sequencing system continues to maintain optimal performance and accuracy.

[0139] After the system completes the noise removal process on the first and second initial genome sequencing datasets, it will have accumulated a wealth of processing experience and data insights. This experience and insights are crucial for optimizing the genome sequencing system. Therefore, the AI ​​analysis system uses this information to update and improve the sequencing system.

[0140] Specifically, the system first analyzes the various noise types and sources identified during the noise removal process. This noise may include instrument noise, environmental noise, reagent noise, and more. By gaining a deeper understanding of the characteristics of this noise, the system can identify potential problems and deficiencies in the sequencing system.

[0141] Next, the AI ​​analysis system addresses these issues and deficiencies by providing specific optimization suggestions and improvement measures. For example, if instrument noise is identified as the primary source of noise, the system may recommend calibration or upgrading the sequencing instrument to reduce instrument errors. If environmental noise is a significant factor, the system may suggest improving the sequencing environment, such as reducing external interference or optimizing temperature control.

[0142] Furthermore, the system optimizes the sequencing system's algorithms and parameters based on data insights from the noise removal process. For example, by adjusting parameters such as sequencing depth and read length, sequencing coverage and accuracy can be improved. Furthermore, the system utilizes advanced machine learning algorithms to refine sequencing data analysis methods and processing procedures, further enhancing data accuracy and reliability.

[0143] Finally, the AI ​​analysis system integrates these optimization recommendations and improvement measures into the target plant's genome sequencing system. This update process may involve software upgrades, hardware adjustments, or algorithm optimization. Through this series of updates, the sequencing system will be able to better adapt to various complex sequencing environments and requirements, providing more accurate and reliable genome sequencing data.

[0144] In summary, by updating the genome sequencing system of target plants, the AI ​​analysis system can not only improve the accuracy and reliability of sequencing data, but also enhance the overall performance and adaptability of the sequencing system. This will provide stronger data support and technical assurance for subsequent genomic research, medical diagnosis, and drug development.

[0145] It is worth mentioning that the genome sequencing system of the target plant object is updated, including: using an artificial intelligence analysis system to deeply analyze the denoised first initial genome sequencing data set and the second initial genome sequencing data set to identify systematic errors and random errors in the sequencing process; based on the error analysis results, adjusting the optical detection system of the sequencing system, optimizing the parameters of laser excitation and fluorescence signal capture to reduce the impact of light signal attenuation and background noise; at the same time, improving the fluid control module of the sequencing system to ensure the precise delivery and mixing of reaction reagents and reduce the noise introduced by reagent concentration fluctuations; further, updating the base recognition algorithm of the sequencing system, introducing a deep learning model to improve the accuracy of base interpretation, and updating the algorithm model in real time through the cloud platform to adapt to the ever-changing sequencing needs; finally, comprehensively upgrading the hardware and software of the sequencing system to ensure the stability and efficiency of the system, thereby achieving comprehensive optimization and update of the genome sequencing system of the target plant object.

[0146] Specifically, AI analysis systems play a crucial role in updating the genome sequencing system of target plants. This update process is not only complex but also highly sophisticated, requiring the system to possess the capabilities of in-depth analysis and precise adjustments.

[0147] First, the AI ​​analysis system conducts an in-depth analysis of the first and second initial genome sequencing datasets, which have undergone noise removal. This step aims to identify systematic and random errors that may exist in the sequencing process. Systematic errors are often caused by inherent biases in the sequencing system or improper operation, while random errors can arise from a variety of unpredictable external factors. By deeply analyzing these datasets, the system can more accurately pinpoint the source of errors, providing strong support for subsequent optimization efforts.

[0148] Next, based on the results of the error analysis, the AI ​​analysis system adjusts the sequencing system's optical detection system. This involves optimizing parameters for laser excitation and fluorescence signal capture. Laser excitation is a critical step in the sequencing process, responsible for stimulating the fluorescently labeled nucleotides to emit a fluorescent signal. By adjusting parameters such as laser intensity, wavelength, and excitation time, the system can more effectively stimulate the fluorescent signal while minimizing the effects of optical signal attenuation and background noise. This helps improve the clarity and accuracy of the sequencing signal.

[0149] At the same time, the system will also improve the sequencing system's fluid control module. The fluid control module is responsible for the precise control and delivery of reagents in the sequencing reaction, such as nucleotides, enzymes, and buffers. By optimizing fluid control parameters such as flow rate, concentration, and mixing ratio, the system ensures precise delivery and mixing of reaction reagents. This not only reduces noise introduced by reagent concentration fluctuations, but also improves the stability and reproducibility of the sequencing reaction.

[0150] To further improve sequencing accuracy, the AI ​​analysis system also updates the sequencing system's base calling algorithm. Traditional base calling algorithms can be limited by fixed patterns and parameters, making them difficult to adapt to evolving sequencing needs. Therefore, the system incorporates a deep learning model to improve the base calling process. Deep learning models possess powerful feature extraction and classification capabilities, enabling more accurate identification and interpretation of base information in sequencing signals. Furthermore, by enabling real-time updates to the algorithm model via the cloud platform, the system continuously learns and adapts to new sequencing data and environmental changes, thereby maintaining optimal recognition performance.

[0151] Finally, to ensure the overall performance and stability of the sequencing system, the AI ​​analysis system comprehensively upgrades both the system's hardware and software. Hardware upgrades may include replacing more advanced sequencing instruments, increasing computer processing power and storage capacity, etc.; software upgrades may involve optimizing the operating system, updating data analysis software, and introducing new sequencing algorithms. These upgrades ensure that the sequencing system performs well in the face of large-scale, highly complex genome sequencing tasks, providing researchers with high-quality and efficient sequencing services.

[0152] In summary, by comprehensively optimizing and updating the genome sequencing system of target plant objects, the artificial intelligence analysis system can significantly improve the accuracy and reliability of sequencing data.

[0153] Furthermore, Figure 2 This is a structural diagram of an artificial intelligence analysis system 200 provided in an embodiment of the present application. Figure 2 The artificial intelligence analysis system 200 shown includes a processor 210, which can call and run a computer program from a memory to implement the method in the embodiment of the present application.

[0154] Alternatively, as Figure 2 As shown, the artificial intelligence analysis system 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method in the embodiment of the present application.

[0155] The memory 230 may be a separate device independent of the processor 210 , or may be integrated into the processor 210 .

[0156] Alternatively, as Figure 2 As shown, the artificial intelligence analysis system 200 may further include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices. Specifically, it may send information or data to other devices, or receive information or data sent by other devices.

[0157] Optionally, the artificial intelligence analysis system 200 can implement the corresponding processes corresponding to the storage engine or components in the storage engine (such as a processing module) or the device deployed with the storage engine in each method of the embodiments of the present application. For the sake of brevity, they will not be repeated here.

[0158] It should be understood that the processor of the embodiments of the present application may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method embodiment can be completed by hardware integrated logic circuits in the processor or software instructions. The above processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The various methods, steps, and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of the present application can be directly implemented as a hardware decoding processor, or can be implemented by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0159] It is understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM bus random access memory (DR RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0160] It should be understood that the above-mentioned memories are exemplary but not restrictive. For example, the memories in the embodiments of the present application may also be static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus random access memory (DRRAM). In other words, the memories in the embodiments of the present application are intended to include, but are not limited to, these and any other suitable types of memories.

[0161] Based on the above, a computer-readable storage medium is provided, on which a computer program is stored, and the computer program implements the above method when running.

[0162] The above are merely examples of the present application and are not intended to limit the present application. For those skilled in the art, the present application may be subject to various modifications and variations.

Claims

1. A plant genome sequencing data analysis method based on artificial intelligence, characterized in that: Applied to an artificial intelligence analysis system, the method comprises: Determine a first genome expression description vector relationship network of a first initial genome sequencing data set and a second genome expression description vector relationship network of a second initial genome sequencing data set using a long short-term memory network, wherein the first initial genome sequencing data set and the second initial genome sequencing data set correspond to a target plant object, and the first genome expression description vector relationship network and the second genome expression description vector relationship network are any one of a genome expression feature set, a transcript structure feature set, or an epigenetic marker feature set; Determine at least one local description vector relationship network from the first genome expression description vector relationship network and the second genome expression description vector relationship network based on the AI ​​sliding sampling kernel, and determine at least one vector relationship network binary based on the data distribution characteristics of each local description vector relationship network in the corresponding genome expression description vector relationship network, wherein the first local description vector relationship network and the second local description vector relationship network in each vector relationship network binary belong to different genome expression description vector relationship networks and have the same data distribution characteristics in the corresponding genome expression description vector relationship network; Integrate the first local description vector relationship network and the second local description vector relationship network in each of the vector relationship network binary groups to obtain a genome sequencing integrated vector set, and determine the sequencing noise perturbation results of the first local description vector relationship network and the second local description vector relationship network in the corresponding vector relationship network binary group based on each of the genome sequencing integrated vector sets; Determine the sequencing noise perturbation results of the first initial genome sequencing data set and the second initial genome sequencing data set based on the sequencing noise perturbation results of each of the vector relationship network binary groups, wherein the sequencing noise perturbation result is the presence of sequencing noise perturbation or the absence of sequencing noise perturbation; For the first local description vector relationship network and the second local description vector relationship network in each of the vector relationship network tuples, the first local description vector relationship network and the second local description vector relationship network are integrated to obtain a genome sequencing integrated vector set, including: generating a genome sequencing integrated vector set based on sequencing attention variables of each characteristic element of the first local description vector relationship network and the second local description vector relationship network, wherein each characteristic element of the genome sequencing integrated vector set includes at least one attention dimension, and the sequencing attention variable of each attention dimension is determined by at least one of the sequencing attention variables of the first local description vector relationship network corresponding to the characteristic element or the sequencing attention variables of the second local description vector relationship network corresponding to the characteristic element; Among them, the sequencing attention variable of each characteristic element of the genome expression feature set is the genome expression mapping parameter, the sequencing attention variable of each characteristic element of the transcript structure feature set is the transcript structure mapping parameter, and each characteristic element of the epigenetic mark feature set corresponds to multiple sequencing attention variables and each sequencing attention variable is an epigenetic mapping parameter of an epigenetic attention dimension; The method of determining at least one local description vector relationship network from the first genome expression description vector relationship network and the second genome expression description vector relationship network based on the AI ​​sliding sampling kernel respectively includes: determining at least one first target data distribution feature in the first genome expression description vector relationship network, determining the data distribution feature in the second genome expression description vector relationship network that is the same as each of the first target data distribution features as a second target data distribution feature, and each of the first target data distribution features is used to reflect the core sequencing embedding vector of the first genome expression description vector relationship network; and determining at least one local description vector relationship network from each of the first target data distribution features and each of the second target data distribution features based on the AI ​​sliding sampling kernel and the selected sampling strategy.

2. The method according to claim 1, characterized in that The determining the sequencing noise perturbation results of the first initial genome sequencing data set and the second initial genome sequencing data set based on the sequencing noise perturbation results of each of the vector relationship network binary groups includes: Determine at least one target data distribution feature cluster, each of the target data distribution feature clusters including one of the first target data distribution features and a second target data distribution feature corresponding to the first target data distribution feature; Determine the first number of vector relationship network tuples corresponding to each of the target data distribution feature clusters, and the second number of vector relationship network tuples in which the sequencing noise disturbance result in the corresponding vector relationship network tuple is that there is no sequencing noise disturbance; For each of the target data distribution feature clusters, determining sequencing noise perturbation results of the first target data distribution feature and the second target data distribution feature in the target data distribution feature cluster based on the first number and the second number corresponding to the target data distribution feature cluster; Based on the sequencing noise perturbation results of each of the target data distribution feature clusters, the sequencing noise perturbation results of the first initial genome sequencing data set and the second initial genome sequencing data set are determined.

3. The method according to claim 2, characterized in that For each of the target data distribution feature clusters, determining the sequencing noise perturbation results of the first target data distribution feature and the second target data distribution feature in the target data distribution feature cluster based on the first number and the second number corresponding to the target data distribution feature cluster, comprises: Determine a first ratio of a second number corresponding to the target data distribution feature cluster to a first number corresponding to the first number; If the first proportion corresponding to the target data distribution feature cluster is less than the first threshold value, determine the noise disturbance analysis weight of the target data distribution feature cluster based on the first proportion corresponding to the target data distribution feature cluster and the second number, and determine the sequencing noise disturbance results of the first target data distribution feature and the second target data distribution feature in the target data distribution feature cluster based on the noise disturbance analysis weight; If the first proportion corresponding to the target data distribution feature cluster is not less than the first threshold value, it is determined that the sequencing noise disturbance results of the first target data distribution feature and the second target data distribution feature in the target data distribution feature cluster are that there is no sequencing noise disturbance.

4. The method according to claim 2, characterized in that: The step of determining the sequencing noise perturbation results of the first initial genome sequencing data set and the second initial genome sequencing data set based on the sequencing noise perturbation results of each of the target data distribution feature clusters comprises: If the sequencing noise perturbation result of at least one of the target data distribution feature clusters is that there is no sequencing noise perturbation, determining that the sequencing noise perturbation results of the first initial genome sequencing data set and the second initial genome sequencing data set are that there is no sequencing noise perturbation; If the sequencing noise perturbation results of each of the target data distribution feature clusters are that sequencing noise perturbations exist, it is determined that the sequencing noise perturbation results of the first initial genome sequencing data set and the second initial genome sequencing data set are that sequencing noise perturbations exist.

5. The method according to claim 1, characterized in that The sequencing noise perturbation results of the first local description vector relationship network and the second local description vector relationship network in the corresponding vector relationship network tuple determined based on each genome sequencing integrated vector set are achieved through a long short-term memory network, and the long short-term memory network is debugged by the following steps: Determine a plurality of genome expression vector relationship network sample tuples, each of the genome expression vector relationship network sample tuples comprising a first genome expression description vector relationship network sample of a first initial genome sequencing data set sample and a second genome expression description vector relationship network sample of a second initial genome sequencing data set sample, the first initial genome sequencing data set sample and the second initial genome sequencing data set sample correspond to a target plant object, and the first genome expression description vector relationship network sample and the second genome expression description vector relationship network sample are respectively any one of a genome expression feature set, a transcript structure feature set or an epigenetic marker feature set; Integrate the first genome expression description vector relationship network sample and the second genome expression description vector relationship network sample in each genome expression vector relationship network sample binary to obtain a genome sequencing integrated vector set sample, input each genome sequencing integrated vector set sample into an initial recurrent neural network to obtain a discriminant viewpoint corresponding to each genome sequencing integrated vector set sample, and determine the sequencing noise perturbation prediction result of the first genome expression description vector relationship network sample and the second genome expression description vector relationship network sample in the genome expression vector relationship network sample binary corresponding to each genome sequencing integrated vector set sample based on the discriminant viewpoint; Based on the actual sequencing noise perturbation results and the sequencing noise perturbation prediction results of the first genome expression description vector relationship network sample and the second genome expression description vector relationship network sample in each genome expression vector relationship network sample binary group, a debugging error function is determined, and the initial recurrent neural network is debugged based on the debugging error function and each genome expression vector relationship network sample binary group until the debugging is completed when the debugging error function meets the debugging termination requirements, and the recurrent neural network obtained after the debugging is determined as the long short-term memory network.

6. The method according to claim 1, characterized in that When the sequencing noise disturbance result is that sequencing noise disturbance exists, the method further includes: Based on the sequencing noise disturbance result, the first initial genome sequencing data set and the second initial genome sequencing data set are subjected to denoising processing.

7. The method according to claim 6, characterized in that The method further comprises: The genome sequencing system of the target plant object is updated.

8. An artificial intelligence analysis system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 7.

Images