Method for centromere sequence extraction and chromosome classification and related device

Abstract

The invention discloses a centromere sequence extraction and chromosome classification method and a related device, and belongs to the technical field of bioinformatics. The method comprises the following steps: acquiring a second-generation sequencing sequence data file and a corresponding reference genome sequence file, and segmenting the second-generation sequencing sequence data file into N sub-files according to a set parallel thread count; creating N parallel thread units, extracting feature vectors of the sequencing sequences in the N sub-files by adopting a DNA sequence feature extraction model, inputting the feature vectors into a pre-constructed centromere sequence recognition model, and screening out candidate centromere sequences; and converting each candidate centromere sequence into a feature vector by using the DNA sequence feature extraction model, and inputting the feature vector into a pre-constructed chromosome classification model to obtain a chromosome attribution result. According to the method, by combining data parallel preprocessing, machine learning and a deep learning model, the centromere region sequence can be identified in large-scale massive next-generation sequencing data, and the chromosome to which the centromere region sequence belongs can be further predicted.

Description

Technical Field

[0001] This invention belongs to the field of bioinformatics technology and relates to a method and related apparatus for centromere sequence extraction and chromosome classification. Background Technology

[0002] Centromeres are crucial core functional elements of eukaryotic chromosomes, playing a key role as microtubule attachment points during mitosis and meiosis, ensuring the accurate and equal segregation of chromosomes into daughter cells. The structural integrity and functional stability of centromeres are the cornerstones of maintaining genome stability and ensuring normal heredity in a species. Therefore, abnormalities in the centromere region are closely related to the occurrence and development of many major human diseases, including chromosomal instability in cancer, aneuploidy such as Down syndrome, and various developmental disorders. In-depth research on the human centromere region has significant scientific and clinical value.

[0003] However, centromeres are typically composed of highly tandemly repeated satellite DNA sequences consisting of hundreds of thousands to millions of base pairs (megabases) of varying lengths. The uneven length of centromeres, the presence of complex higher-order repetitive structures and variations, and the evolutionary changes in their internal structure all contribute to significant challenges in the sequencing, assembly, and analysis of genes in centromeres. In recent years, high-throughput sequencing technologies, represented by next-generation sequencing (NGS), have made revolutionary progress, generating massive amounts of NGS human genome sequencing data globally. This data contains a wealth of undiscovered valuable information about centromere sequence variations and epigenetic states. However, accurately extracting centromere sequence fragments from these standard sequencing data consisting of short reads (only 150 bases in length) and determining their chromosomal origin remains a major bottleneck in the field of bioinformatics.

[0004] Currently, there is a lack of dedicated bioinformatics tools on the market that can directly access second-generation sequencing data and automate centromere sequence extraction and analysis without requiring other data support. Researchers often either rely heavily on third-generation sequencing data or require complex manual verification, leading to inefficiency and underutilization of the vast amount of second-generation sequencing data. Summary of the Invention

[0005] The purpose of this invention is to provide a method and related apparatus for centromere sequence extraction and chromosome classification, so as to solve the technical problem of low efficiency and accuracy in identifying centromere region sequences from second-generation sequencing data in the prior art.

[0006] To achieve the above objectives, the present invention employs the following technical solution: In a first aspect, the present invention provides a method for centromere sequence extraction and chromosome classification, comprising the following steps: Obtain the next-generation sequencing sequence data file and the corresponding reference genome sequence file, and divide the next-generation sequencing sequence data file into N sub-files according to the set number of parallel threads; N parallel thread units are created, and the feature vectors of the sequencing sequences in the N sub-files are extracted using a DNA sequence feature extraction model. The feature vectors are then input into a pre-constructed centromere sequence recognition model to screen out candidate centromere sequences. After converting each candidate centromere sequence into a feature vector using a DNA sequence feature extraction model, the vector is input into a pre-constructed chromosome classification model to obtain the chromosome assignment result.

[0007] Furthermore, the step of obtaining the next-generation sequencing sequence data file and the corresponding reference genome sequence file, and dividing the next-generation sequencing sequence data file into N sub-files according to the set number of parallel threads, specifically includes: Obtain the second-generation sequencing sequence data file and the corresponding reference genome sequence FASTA file, and verify the validity of the file; the second-generation sequencing sequence data file is a CRAM format file containing second-generation sequencing short read sequence information. Based on the coordinate information of the reference genome sequence and the set number of parallel threads, the second-generation sequencing sequence data file is partitioned into N sub-files.

[0008] Furthermore, the DNA sequence feature extraction model is the dna2vec model, which is used to convert k-mer fragments in the DNA sequence into high-dimensional feature vectors.

[0009] Furthermore, the centromere sequence recognition model is an XGBoost binary classification model; the chromosome classification model is a multi-class deep learning model.

[0010] Furthermore, the step of creating N parallel thread units, using a DNA sequence feature extraction model to extract feature vectors from the sequencing sequences in N sub-files, and inputting the feature vectors into a pre-constructed centromere sequence recognition model to screen out candidate centromere sequences specifically includes: Create N parallel thread units, each of which loads a pre-trained DNA sequence feature extraction model and a centromere sequence recognition model; The DNA sequence feature extraction model is used to extract each 5-mer sequence in a 150bp sequence and convert it into a 100-dimensional feature vector and store it in a set. Finally, the average of the feature vectors in the set is used as the feature vector of the entire sequence. The feature vector of each sequence is classified using a centromere sequence recognition model, and finally it is determined whether the sequence is a centromere sequence. When a sequence is determined to be a centromere sequence, all relevant information of the sequence is extracted and stored in a temporary local candidate list; if a sequence is determined to be non-centromere, the sequence is discarded.

[0011] Furthermore, the step of converting each candidate centromere sequence into a feature vector using a DNA sequence feature extraction model and then inputting it into a pre-constructed chromosome classification model to obtain the chromosome assignment result specifically includes: Each candidate centromere sequence is converted into a feature vector using a DNA sequence feature extraction model. The feature vector is input into the chromosome classification model, and the forward propagation process of the model outputs a probability vector containing 24 elements; each element corresponds to the probability that the sequence belongs to a certain chromosome; the chromosome corresponding to the element with the highest probability value in the feature vector is selected as the final label of the sequence.

[0012] Furthermore, the method also includes: Create 24 data containers, each corresponding to one of the 24 chromosomes. Store the complete information of each sequence into the corresponding container based on the chromosome affiliation tag of each sequence. The data in the 24 containers are serialized and saved in chr*.pkl format in the specified path.

[0013] Secondly, the present invention provides a system for centromere sequence extraction and chromosome classification, comprising: The data preprocessing module is used to obtain the second-generation sequencing sequence data file and the corresponding reference genome sequence file, and to divide the second-generation sequencing sequence data file into N sub-files according to the set number of parallel threads; The centromere identification module is used to create N parallel thread units, use a DNA sequence feature extraction model to extract feature vectors from the sequencing sequences in N sub-files, and input the feature vectors into a pre-constructed centromere sequence identification model to screen out candidate centromere sequences. The chromosome classification module is used to convert each candidate centromere sequence into a feature vector using a DNA sequence feature extraction model, and then input it into a pre-built chromosome classification model to obtain the chromosome assignment result.

[0014] Thirdly, the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method for centromere sequence extraction and chromosome classification as described above.

[0015] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method for centromere sequence extraction and chromosome classification as described above.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention discloses a method and related apparatus for centromere sequence extraction and chromosome classification. By splitting second-generation sequencing data files into multiple sub-files and combining a centromere sequence recognition model for first-stage centromere sequence identification, the identified sequences are simultaneously input into a deep learning-based chromosome classification model for chromosome attribution determination. This enables the automated and efficient extraction and classification of centromere regions, which are previously difficult to locate quickly in large-scale second-generation sequencing data. Furthermore, this invention introduces a parallel processing mechanism in the first stage, improving the efficiency of processing large-scale data. Simultaneously, the second-stage deep learning model accurately determines chromosome attribution, significantly improving classification accuracy. This invention, through its technical approach of "data fragmentation preprocessing + binary classification recognition + multi-class chromosome determination," solves the problems of low efficiency, limited recognition accuracy, and difficulty in handling large-scale data in existing methods. It achieves the unification of automated centromere sequence extraction and chromosome-level classification, providing an efficient and accurate new method for analyzing complex regions of the human genome. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the system of the present invention; Figure 3 This is a flowchart of the overall program of an embodiment of the present invention; Figure 4 This is a flowchart of the centromere recognition module according to an embodiment of the present invention; Figure 5 This is a flow control diagram of the chromosome classification module in an embodiment of the present invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0020] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this invention is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0021] In existing technologies, detailed analysis of highly repetitive regions such as centromeres often heavily relies on the long read lengths of third-generation sequencing (NGS) to traverse repetitive units. However, this approach presents two problems: First, the cost of NGS is significantly higher than that of NGS, resulting in limited data volume and restricting its large-scale application. Second, global research institutions and sequencing centers store massive amounts (petabytes) of historical NGS data, and the centromere information contained within this data remains largely unused due to a lack of effective analysis tools, leading to a waste of data resources. This invention, through innovative algorithm design, can directly mine and analyze centromere sequences from short-read NGS data without the need for assistance or error correction from long-read NGS data. This not only significantly reduces research costs and lowers the technical application threshold but also allows for the utilization of massive existing data resources.

[0022] See Figure 1 This invention specifically discloses a method for centromere sequence extraction and chromosome classification, comprising the following steps: S1, obtain the second-generation sequencing sequence data file and the corresponding reference genome sequence file, and divide the second-generation sequencing sequence data file into N sub-files according to the set number of parallel threads; This step receives two input data sets: a CRAM format file containing short read sequence information from next-generation sequencing, and a corresponding FASTA file of the reference genome sequence. The CRAM file is a compressed binary format containing a large amount of sequencing data and its alignment information. First, the method verifies the validity of the input file. Then, based on the coordinates of the reference genome and the user-defined number of parallel computing resources (e.g., N threads), the method partitions the CRAM file. The file is divided into N sub-files, each assigning a unique, non-contiguous parallel processing task to the subsequent N parallel threads. Thus, the original single data stream is decomposed into N independent, parallelizable task units, providing a foundation for accelerating the next step of centromere sequence identification.

[0023] S2, create N parallel thread units, use the DNA sequence feature extraction model to extract feature vectors of sequencing sequences in N sub-files, input the feature vectors into the pre-constructed centromere sequence recognition model, and screen out candidate centromere sequences; The goal of this step is to quickly filter candidate centromere sequences from massive background data. This stage employs a parallel processing architecture to accelerate the process. First, multiple parallel computing threads are created based on the number of available CPU cores or a user-specified number. Each parallel thread handles one task unit generated in the data preprocessing module. Each parallel thread first loads a pre-trained binary classification machine learning model (centromere sequence recognition model) to distinguish between centromere and non-centromere sequences. This model is trained on a simulated second-generation dataset generated using ngsSim on the complete human genome CHM13 released by the T2T consortium, and its performance is evaluated on a simulated second-generation dataset generated using ngsSim on the hg002 dataset. In addition, each instance also loads a pre-trained dna2vec model (DNA sequence feature extraction model), which can transform a single kmer into a 100-dimensional feature vector, thereby achieving sequence feature extraction and sequence encoding.

[0024] Subsequently, each parallel thread unit will traverse all sequencing sequences in its assigned task unit and execute the following steps: (Note: The sequencing sequence length processed in this embodiment is 150bp (base pairs)) First, the feature vector of each sequencing sequence is extracted. Specifically, a pre-trained dna2vec model is invoked, which maps DNA base sequences (text strings composed of A, T, C, and G) into a high-dimensional vector space. In this embodiment, the program calls the dna2vec model to sequentially extract each 5-mer sequence from a 150bp sequence and converts it into a 100-dimensional feature vector, storing it in a set. Finally, the average of the feature vectors in the set is used as the feature vector for the entire sequence. This transforms each 150bp base sequence into a 100-dimensional numerical feature vector that characterizes its intrinsic biological information.

[0025] Next, the feature vector of each sequence is classified. Specifically, the 100-dimensional feature vector generated in the previous step is used as input and fed into the XGBoost binary classification model. The model calculates the input feature vector based on its internal decision tree set, and finally determines whether the sequence is a centromere sequence.

[0026] Finally, the model's output is used for filtering. When a sequence is determined by the model to be a "centromere," all relevant information about the sequence (such as sequence ID, original base sequence, quality value, etc.) is extracted and stored in a temporary local candidate list. Conversely, if a sequence is determined to be non-centromere, it is discarded.

[0027] Through the above operations, each parallel thread unit can independently and efficiently complete the screening of its assigned data subset, obtaining a batch of high-quality candidate centromere sequences.

[0028] S3: After converting each candidate centromere sequence into a feature vector using a DNA sequence feature extraction model, the vector is input into a pre-constructed chromosome classification model to obtain the chromosome assignment result.

[0029] The goal of this step is to accurately determine the chromosome origin of each sequence in the candidate set. The method loads a pre-trained, more complex multi-class deep learning model (chromosome classification model). This model is also trained on simulated second-generation data of the CHM13 reference genome and validated on the hg002 dataset using simulated second-generation datasets generated by the ngsSim tool, demonstrating its ability to accurately identify the twenty-four chromosome classes in humans.

[0030] This embodiment iterates through the candidate centromere sequence set generated in S2. For each candidate sequence, necessary preprocessing is performed first. The 150bp base sequence is converted into a 100-dimensional feature vector using the dna2vec model, similar to step 2. This feature vector is then input into the deep learning model. The model's forward propagation outputs a probability vector containing 24 elements, where each element corresponds to the probability that the sequence belongs to a certain chromosome. The method selects the chromosome corresponding to the element with the highest probability value in this vector as the final attribution label for the sequence.

[0031] S4, Final Result Output After all candidate sequences have been classified for chromosomes, this step will process and output the final results. The method creates 24 data containers, each corresponding to one of the 24 chromosome types. Then, it iterates through the labeled sequence set again, storing the complete information of each sequence into the corresponding container based on its chromosome affiliation tag. Finally, the method serializes the data in these 24 containers and saves them in independent, easily readable file formats (e.g., chr1.pkl, chr2.pkl, ..., chrY.pkl) in the user-specified output path. Each file contains all centromere sequence information identified in this analysis and belonging to the corresponding chromosome.

[0032] See Figure 2This invention discloses a system for centromere sequence extraction and chromosome classification, including a data preprocessing module, a centromere identification module, and a chromosome classification module. Specifically, the data preprocessing module acquires second-generation sequencing sequence data files and corresponding reference genome sequence files, and divides the second-generation sequencing sequence data files into N sub-files according to a set number of parallel threads. The centromere identification module creates N parallel thread units, uses a DNA sequence feature extraction model to extract feature vectors from the sequencing sequences in the N sub-files, and inputs these feature vectors into a pre-constructed centromere sequence identification model to screen out candidate centromere sequences. The chromosome classification module uses the DNA sequence feature extraction model to convert each candidate centromere sequence into a feature vector, and then inputs it into a pre-constructed chromosome classification model to obtain the chromosome attribution result. This invention constructs a two-stage cascaded classification structure model. The first stage adopts a parallel computing architecture and uses the XGBoost binary classification machine learning model, which has fast computation speed and a simple model structure, to ensure the most efficient acquisition of potential centromere region sequences. The second stage uses a deep learning 24-class classification model with strong feature representation capabilities and high classification accuracy, but with a complex model, to ensure the most accurate location of the chromosome origin of the sequence. By combining two-stage models, this invention addresses the challenge of single models simultaneously handling massive data screening and high-precision, complex multi-class classification. Furthermore, it provides an end-to-end automated analysis workflow. Users only need to provide standard next-generation sequencing data files (CRAM format files) and the corresponding reference genome; this method automatically completes the entire process, including data preprocessing, parallel initial screening, candidate sequence aggregation, fine chromosome classification, and structured result output. The entire workflow requires no complex manual intervention, significantly lowering the barrier to entry and improving analytical efficiency and result reproducibility. In addition, this invention strictly adheres to a modular design, encapsulating the three modules—data preprocessing, centromere identification, and chromosome classification—into separate Python programs with interfaces. If a downstream task only requires the classification results of one stage, then only the interfaces for data preprocessing and centromere identification can be called, providing options for diverse downstream analytical tasks. The following is a specific implementation example, combined with... Figure 3 Describe the overall process.

[0033] 1. First, call the data preprocessing module interface we provide, and provide the interface with the CRAM file location information of the input data, the corresponding reference genome location information, and the thread number parameter information for parallel processing; 2. Subsequently, after receiving the number of threads for parallel processing, the program will verify the validity of the file, and then divide the entire genome alignment task into multiple independent sub-task units based on the coordinates of the CHM13 reference genome, and prepare to distribute them to multiple parallel processes. 3. Subsequently, the program launches multiple parallel processes, each of which is ready to enter... Figure 3 The centromere identification module is described in the code. For each parallel thread, the program first loads the pre-trained model dna2vec, iterates through the task units it is responsible for, and transforms the sequence into a 100-dimensional feature vector. Then, the program loads our pre-trained binary classification machine learning model, iterates through the feature vectors of all sequences, and feeds them into the model for calculation. Finally, the sequence is determined to be either a centromere sequence or a non-centromere sequence. Once determined to be a centromere sequence, all information about this sequence in the original file is temporarily stored. The flowchart of the entire centromere identification module is shown below. Figure 4 As shown; 4. After the centromere identification module is completed, the program will organize the output of this module. Each parallel task unit will generate a pkl file for the sub-file it processes. Each file contains all the centromere sequences in this sub-file and their detailed information (including feature vectors). 5. Subsequently, the program passes the output of the centromere recognition module to the chromosome classification module. In this module, the method pre-loads our trained 24-classification (humans have 24 chromosomes) deep learning model. It sequentially reads the output of the centromere recognition module, iterating through each sequence. For each sequence, the model extracts the feature vector generated by the centromere recognition module as input, performs forward propagation inference, and finally outputs a 24-dimensional probability vector, representing the probability that the sequence is classified as a chromosome. Finally, we take the category with the highest probability as the classification result. The flowchart of the entire process is as follows: Figure 5 As shown; 6. Next, a dictionary is created, with chromosome names as keys and lists as values. In the chromosome classification module, once a sequence is identified as belonging to a particular chromosome, the sequence and its complete information are saved and added to the corresponding key-value list in the dictionary. Finally, the method saves the list data of `chr*` as `chr*.pkl` in the specified path. The program then terminates. This completes the identification of centromere sequences and the classification of their corresponding chromosomes.

[0034] The performance of the core model in this embodiment is evaluated below: The one-stage binary classification XGBoost model and the two-stage 24-class chromosome classification model used in this embodiment were both trained on a simulated second-generation dataset generated using the ngsSim tool on the CHM13 dataset. To verify the accuracy and reliability of the core models of this software, we used the industry-recognized standard dataset HG002 (GIAD Consortium) to independently evaluate the performance of the two core models. The evaluation environment was isolated from the training environment to ensure the objectivity of the test results.

[0035] 1. Performance of Centromere Sequence Recognition Model The goal of this model is to accurately screen centromere sequences from massive sequencing data. We used a binary classification model to evaluate its performance, and its performance on the HG002 test set is shown in Table 1 below: Table 1: Performance of the one-stage classification model on hg002

[0036] 2. Performance of the chromosome classification model (deep learning 24-class classification) The goal of this model is to accurately locate the identified centromere sequences to their respective chromosomes (a total of 24 classes). We used a multi-class classification task evaluation metric, and its performance on the HG002 test set is shown in Table 2 below: Table 2: Performance of the two-stage classification model on hg002

[0037] In one embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used in the operation of a method for centromere sequence extraction and chromosome classification.

[0038] This invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the method for centromere sequence extraction and chromosome classification in the above embodiments.

[0039] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0040] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0041] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0042] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for centromere sequence extraction and chromosome classification, characterized by, The method comprises the following steps: obtaining a second-generation sequencing sequence data file and a corresponding reference genome sequence file, and dividing the second-generation sequencing sequence data file into N sub-files according to a set number of parallel threads; creating N parallel thread units, extracting feature vectors of sequencing sequences in the N sub-files by using a DNA sequence feature extraction model, inputting the feature vectors into a pre-constructed centromere sequence identification model, and screening candidate centromere sequences; after converting each candidate centromere sequence into a feature vector by using the DNA sequence feature extraction model, inputting the feature vector into a pre-constructed chromosome classification model, and obtaining a chromosome attribution result.

2. The method of centromere sequence extraction and chromosome classification according to claim 1, wherein, The step of obtaining a second-generation sequencing sequence data file and a corresponding reference genome sequence file, and dividing the second-generation sequencing sequence data file into N sub-files according to a set number of parallel threads specifically comprises: obtaining a second-generation sequencing sequence data file and a corresponding reference genome sequence FASTA file, and checking the validity of the files; the second-generation sequencing sequence data file is a CRAM format file containing second-generation sequencing short read sequence information; dividing the second-generation sequencing sequence data file into N sub-files according to the coordinate information of the reference genome sequence and the set number of parallel threads.

3. The method of centromere sequence extraction and chromosome classification according to claim 1, wherein, The DNA sequence feature extraction model is a dna2vec model, which is used to convert k-mer fragments in a DNA sequence into a high-dimensional feature vector.

4. The method of centromere sequence extraction and chromosome classification according to claim 1, wherein, The centromere sequence identification model is an XGBoost binary classification model, and the chromosome classification model is a multi-classification deep learning model.

5. The method of centromere sequence extraction and chromosome classification according to claim 1, wherein, The step of creating N parallel thread units, extracting feature vectors of sequencing sequences in the N sub-files by using a DNA sequence feature extraction model, inputting the feature vectors into a pre-constructed centromere sequence identification model, and screening candidate centromere sequences specifically comprises: creating N parallel thread units, each of which loads a pre-trained DNA sequence feature extraction model and a centromere sequence identification model; extracting each group of 5-mer sequences in a sequence with a length of 150 bp by using the DNA sequence feature extraction model, converting them into a 100-dimensional feature vector, storing the feature vector in a set, and finally averaging the feature vectors in the set as the feature vector of the entire sequence; classifying the feature vector of each sequence by using the centromere sequence identification model, and finally determining whether the sequence is a centromere sequence; when the sequence is determined to be a centromere sequence, all related information of the sequence is extracted and stored in a temporary local candidate set list; if it is determined to be a non-centromere, the sequence is discarded.

6. The method of centromere sequence extraction and chromosome classification according to claim 1, wherein, The step of converting each candidate centromere sequence into a feature vector by using the DNA sequence feature extraction model, and inputting the feature vector into a pre-constructed chromosome classification model to obtain a chromosome attribution result specifically comprises: converting each candidate centromere sequence into a feature vector by using the DNA sequence feature extraction model; The feature vector is input into a chromosome classification model, and the forward propagation process of the model outputs a probability vector containing 24 elements; each element corresponds to the probability that the sequence belongs to a certain chromosome; the chromosome corresponding to the element with the highest probability value in the feature vector is selected as the final attribution label of the sequence.

7. The method of centromere sequence extraction and chromosome classification according to claim 1, wherein, Also includes: 24 data containers are created, corresponding to 24 chromosomes respectively, and the complete information of the sequence is stored in the corresponding container according to the chromosome attribution label of each sequence; The data in the 24 containers is serialized respectively, and saved in the specified path in the format of chr*.pkl.

8. A system for centromere sequence extraction and chromosome classification, characterized by, Including: The data preprocessing module is used to obtain the second generation sequencing sequence data file and the corresponding reference genome sequence file, and to divide the second generation sequencing sequence data file into N sub-files according to the set number of parallel threads; The centromere recognition module is used to create N parallel thread units, extract the feature vector of the sequencing sequence in the N sub-files by using the DNA sequence feature extraction model, and input the feature vector into the pre-constructed centromere sequence recognition model to screen out the candidate centromere sequence; The chromosome classification module is used to convert each candidate centromere sequence into a feature vector by using the DNA sequence feature extraction model, and input the feature vector into the pre-constructed chromosome classification model to obtain the chromosome attribution result.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the steps of the method for extracting centromere sequence and classifying chromosomes according to any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. The computer program is executed by the processor to realize the steps of the method for extracting centromere sequence and classifying chromosomes according to any one of claims 1-7.

Images