Chromosome T2T genome assembling method based on artificial intelligence technology

Abstract

The invention relates to the technical field of bioinformatics genome assembly analysis, and particularly discloses a chromosome T2T genome assembly method based on an artificial intelligence technology, and the method comprises the following steps: carrying out a series of analysis and comparison on gene sequencing data, and based on two built deep learning models and a support vector machine model, carrying out a series of analysis and comparison on the gene sequencing data; gene data and telomere data which are supplemented with unknown basic groups in the gene data with the determined position relation are obtained, chromosome T2T gene data are obtained through data analysis and processing, and a chromosome T2T genome is finally obtained through format arrangement. The method is not limited by research species, sequence fragment series connection, hole filling, telomere prediction and addition are automatically completed in the analysis process, and finally the chromosome T2T genome is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of bioinformatics genome assembly and analysis, and in particular to an artificial intelligence-based chromosome T2T genome assembly method. Background Art

[0002] With the development of sequencing technology, biological data has exploded. Bioinformatics provides methods to parse, store and analyze these big data, enabling scientists to extract useful information from complex biological data and help people study the structure, function and evolution of the entire genome. Bioinformatics methods can be used to compare the genomes of different species to study their evolutionary relationships and mechanisms. In the agricultural field, bioinformatics can help understand the genomes of crops, thereby improving varieties and increasing crop yields and disease resistance.

[0003] With the continuous development of long-read sequencing technology, more and more species have completed the assembly of telomere-to-telomere (T2T) genomes. T2T genomes are of great significance for the discovery of functional genes and the study of important biological mechanisms.

[0004] Nowadays, there are still some shortcomings in the research on the assembly method of chromosome T2T genome using artificial intelligence technology. Specifically, the traditional assembly method is inefficient and difficult to assemble to the chromosome T2T level. Most of the genomes currently published are at the scanfold level. The assembly method based on statistical analysis to assemble the chromosome T2T genome has the disadvantages of high error rate and difficulty in assembly. Artificial intelligence technology is now introduced to assemble a large number of chromosome T2T genomes to promote the development of bioinformatics technology. Therefore, it is necessary to optimize the assembly method of chromosome T2T genome, which is specifically reflected in the use of artificial intelligence to assist multiple steps, so as to achieve high-efficiency and high-quality assembly of chromosome T2T genomes and promote the development of bioinformatics technology. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a method for assembling a chromosome T2T genome based on artificial intelligence technology, which can effectively solve the problems involved in the above-mentioned background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: analyzing the quality of the gene sequencing data to be analyzed (second-generation sequencing data), and selecting the optimal assembly result based on the optimal value; selecting a reference genome based on the gene sequencing data to be analyzed, and constructing a sequence alignment deep learning model; based on the sequence alignment deep learning model based on the Transformer architecture, the gene sequencing data to be analyzed is input into the model to determine the positional relationship of the gene sequencing data to be analyzed; based on the gene data with the determined positional relationship, the long-read sequencing data is aligned to the gene data with the determined positional relationship using the BLAST alignment tool to determine the alignment result ; Based on the alignment results and long-read sequencing data, a deep learning model for gene data gap filling based on the Transformer architecture is constructed, and the gene data with determined positional relationships is input into the model to obtain the output of the model, that is, the gene data with unknown bases filled in the gene data with determined positional relationships; based on the quantitative characteristics of breakpoint information of the optimal assembly result and the assembly results of multiple genomes, an alternating test is performed to construct a support vector machine model to maximize the classification information, thereby determining the telomere data, that is, the contig of the telomere information, and based on the telomere data and the gene data with unknown bases filled, the chromosome T2T genome is obtained.

[0007] As a further method, survey data information of the gene sequencing data to be analyzed is obtained, and the survey data information is specifically: sequencing error rate of the gene data to be analyzed, number of repeated sequences of the gene data to be analyzed, sequence depth of the gene data to be analyzed, and sequence coverage of the gene data to be analyzed based on the survey data information of the gene sequencing data to be analyzed. The quality assessment characteristic value of the gene sequencing data to be analyzed is obtained by comprehensive analysis, and the quality assessment characteristic value of the gene sequencing data to be analyzed will be used as the main basis for selecting the optimal assembly result; the gene sequencing data to be analyzed is assembled multiple times, and the assembly results and the quality assessment characteristic value of the gene sequencing data to be analyzed are calculated to obtain the optimal value, which is mainly used to evaluate the optimal assembly result; the optimal values ​​of multiple assemblies are comprehensively analyzed, and the highest value among the optimal values ​​is used as the optimal assembly result, and the remaining optimal values ​​are used as reference.

[0008] As a further method, the survey data information of the gene sequencing data to be analyzed specifically includes: the sequencing error rate of the gene data to be analyzed, the number of repeated sequences of the gene data to be analyzed, the sequencing depth of the gene data to be analyzed, and the sequence coverage of the gene data to be analyzed.

[0009] As a further method, the quality assessment characteristic value of the gene data to be analyzed is specifically calculated as follows: ; In the formula, in the formula, is the quality assessment feature value of the gene sequencing data to be analyzed, is the sequencing error rate of the gene data to be analyzed, is the number of repeated sequences in the gene data to be analyzed, is the sequence depth of the gene data to be analyzed, is the sequence coverage of the gene data to be analyzed, is the compensation factor for the sequencing error rate of the gene data to be analyzed. is the compensation factor for the depth of the gene data sequence to be analyzed. is the compensation factor for the coverage of the gene data sequence to be analyzed. is the compensation factor for the number of repeated sequences in the gene data to be analyzed. is a natural constant, and max is the maximum value function.

[0010] As a further method, the optimal value is obtained, and the specific calculation method is: ; In the formula, is the optimal value, is the quality assessment feature value of the gene data to be analyzed, contig Count down from the longest one until the segment with a length that is half of the total length. The length of the last segment counted is the contig. The longer the length, the more long fragments there are and the better the quality of the final assembly.

[0011] As a further method, a sequence alignment deep learning model is constructed. The specific process is as follows: based on species information and gene sequencing data, the genome of closely related species is obtained as a reference genome, and the obtained reference genome is converted into a json format input data set through a program; the activation function ReLU and neural network structure Transformer pre-provided by PyTorch are selected, and the input layer is defined as contig data in json format, that is, the gene data to be aligned is formatted and converted into json format input; the hidden layer is defined as sequence vectorization through Transfomer technology, and each vectorized fragment is aligned to the vectorized data of the reference genome through the cosine similarity function, and the optimal result is selected. The output layer is defined as continuous gene data with a determined position relationship; the stochastic gradient descent method is used to update the sequence alignment deep learning model parameters to minimize the loss function, and the computer computing resources are adjusted by reducing or increasing the usage of video memory, and the sequence alignment deep learning model based on the Transformer architecture is verified. For example, the data randomly cut from the Agaricus bisporus reference genome can be used to input the model for verification. The reference genome mentioned in this paragraph is the Agaricus bisporus genome or the Ganoderma lucidum genome downloaded from NCBI; the program mentioned in this paragraph means completing the data format conversion by programming; the data set mentioned in this paragraph is a standard format data set, which randomly cuts off the sequence information, takes the previous one as 1, and the next one as 2 to make a json data set; the sequence vectorization mentioned in this paragraph is realized by Transformer technology, specifically, input embedding is realized by looking up the embedding matrix. When each word in the input sequence passes through the embedding layer, it will find its corresponding vector representation in the embedding matrix. The main goal of position encoding is to introduce position information into the embedding vector of sequence data to capture the order information of words or characters in the language. As a further method, the positional relationship of the gene sequencing data to be analyzed is determined. The specific analysis process is: based on the sequence alignment deep learning model based on the Transformer architecture, the gene sequencing data to be analyzed is input into the model, and the output result is converted into a common format, that is, a program is written in a programming language to convert json into fasta, and the positional relationship of the gene sequencing data to be analyzed is determined.

[0012] As a further method, the long-read sequencing data is aligned to the gene data with a determined positional relationship to determine the alignment result. The specific analysis process is as follows: the average coverage length of the long-read sequencing data is longer than that of the short-read sequencing data. Therefore, the long-read sequence data is used to cover the gaps in the short-read sequence data to determine the blank base sequence that is not determined by the short-read sequencing data; the blank base sequence position, blank base sequence, long-read sequencing data, and short-read sequencing data are converted into json format data for storage to obtain the alignment result in json format.

[0013] As a further method, genetic data with filled unknown bases in genetic data with determined positional relationships are obtained. The specific process is: a deep learning model Transformer is constructed according to the comparison results in json format and long-read sequencing data; PyTorch is used to define the input layer as genetic data in text format; the hidden layer is defined as sequence vectorization achieved through Transformer technology, and each vectorized fragment is then aligned to the vectorized data of the reference genome through the cosine similarity function, the optimal result is selected, and the optimal result is matched with the input layer data, that is, the selected optimal result string and the input string are concatenated to fill the blank sequence; the output layer is defined as genetic data with filled unknown bases in genetic data with determined positional relationships; based on the constructed genetic data gap filling deep learning model based on Transformer architecture, genetic data with filled unknown bases in genetic data with determined positional relationships are obtained.

[0014] As a further method, a support vector machine model is constructed to determine the telomere data. The specific process is as follows: based on the quantitative characteristics of the breakpoint information of the optimal assembly result and the results of multiple gene assembly as sample points, a support vector machine model is constructed; based on the constructed support vector machine model, the short-read sequencing data is classified and judged to determine the telomere sequence; based on the determined telomere sequence, the gene data with unknown bases filled in the gene data with a determined position relationship is compared to determine the telomere position; the telomere position and telomere sequence are combined into binary data to obtain telomere information. The binary data here is a tuple data type of Python, which consists of a telomere position and a telomere sequence. That is, a telomere position uniquely corresponds to a telomere sequence.

[0015] As a further method, the optimal assembly result breakpoint information quantity feature is calculated as follows: ; Wherein, T is the number of telomere sequences in the total number of data values ​​of each breakpoint of the long-read sequencing data aligned to the gene data with unknown bases filled in, R is the total number of data values ​​of each breakpoint end of the long-read sequencing data aligned to the gene data with unknown bases filled in, and Ratio is the breakpoint information quantity feature.

[0016] As a further method, a chromosome T2T genome is obtained. The specific process is: based on the genetic data and telomere information of the genetic data with the unknown bases filled in the genetic data with the determined positional relationship, the telomere sequence is inserted into the genetic data with the unknown bases filled in through the corresponding quantitative positional relationship to obtain the genetic data of chromosome T2T; the genetic data of chromosome T2T is formatted to obtain the chromosome T2T genome.

[0017] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: The present invention provides a method for assembling a chromosome T2T genome based on artificial intelligence technology. Through artificial intelligence technology, gene alignment is completed to obtain gene data with a determined positional relationship; blank base filling is completed to obtain complete and continuous gene sequencing data; telomere information is completed to confirm the chromosome T2T genome with telomere information.

[0018] Based on the two deep learning models and a support vector machine model, the traditional genome assembly efficiency can be improved, saving time for scientific researchers and technicians, and ensuring accuracy. The sequence alignment deep learning model based on the Transformer architecture has sufficient generalization ability. By adding genetic data of other species, it can be aligned without being restricted by the research species, thus completing the sequence alignment.

[0019] Transformer, a deep learning model for filling blank base sequences, can efficiently use artificial intelligence prediction methods to predict unknown bases. Specifically, it is a method of building a model by complementing the advantages of long-read data and short-read data. It is not limited by the research species and only needs to add the corresponding data to the model to complete the filling of blank bases in the genetic data.

[0020] The support vector machine model corresponding to telomere information judgment can efficiently use artificial intelligence methods to judge whether it is a telomere sequence. If it is a telomere sequence, the position information and sequence information are retained, and the telomere sequence is added to the chromosome genome to form a chromosome T2T genome. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0022] Figure 1 It is a schematic diagram of the method flow of the present invention.

[0023] Figure 2The results of chromosome assembly and T2T gene assembly (where: the first column is the chromosome name, the second column is the telomere start coordinate, the third column is the telomere end coordinate, and the fourth column is the chromosome length) DETAILED DESCRIPTION

[0024] The following will be combined with the attached embodiment of the present invention Figure 1-2 , the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0025] Agaricus bisporus Agaricus bisporus (Large) Sing. ) is a fungus of the Agaricus family and the genus Agaricus. The fruiting body of Agaricus bisporus is medium-sized, with a cap width of 5-12 cm, which is hemispherical at first, then flat, white, smooth, slightly dry and gradually yellow, and the edges are initially rolled inward. The flesh is white and thick, and turns slightly light red after being injured, with a unique mushroom smell. Agaricus bisporus is edible and delicious. It is an edible fungus with a large cultivation scale and a wide cultivation range. Its protein content is as high as 42%, and it is rich in amino acids, nucleotides and vitamins. Agaricus bisporus can also be used as medicine. It contains a large amount of tyrosinase, which is effective in lowering blood pressure. It can also be made into an auxiliary therapeutic agent for pneumonia. Due to the successful research on submerged culture, people can also use mushroom mycelium to produce substances such as protein, oxalic acid and bacterial sugar. This example uses a chromosome T2T genome assembly method based on artificial intelligence technology provided by the present invention to assemble the chromosome T2T genome based on short-read sequencing data and long-read sequencing data of Agaricus bisporus to explain the specific steps and methods of the present invention. The specific steps are as follows: Reference Figure 1-2 As shown, the present invention provides a method for assembling a chromosome T2T genome based on artificial intelligence technology.

[0026] (1) Based on the second-generation short-read sequencing data of Agaricus bisporus, the quality assessment characteristic value is calculated by the surveyed data sequencing error rate, the number of repeated sequences in the gene data to be analyzed, the sequence depth of the gene data to be analyzed, and the sequence coverage of the gene data to be analyzed. Multiple gene assemblies are performed, and the optimal value is calculated to select the optimal assembly result. The specific implementation process is: based on the second-generation short-read sequencing data of Agaricus bisporus and the improved gene data sequencing error rate, the number of repeated sequences in the gene sequencing data to be analyzed, the sequence depth of the gene data to be analyzed, and the sequence coverage of the gene sequencing data to be analyzed, the quality assessment characteristic value of the Agaricus bisporus gene data to be analyzed is calculated. The calculation formula is: ; In the formula, in the formula, is the quality assessment feature value of the gene sequencing data to be analyzed, is the sequencing error rate of the gene data to be analyzed, is the number of repeated sequences in the gene data to be analyzed, is the sequence depth of the gene data to be analyzed, is the sequence coverage of the gene data to be analyzed, is the compensation factor for the sequencing error rate of the gene data to be analyzed. is the compensation factor for the depth of the gene data sequence to be analyzed. is the compensation factor for the coverage of the gene data sequence to be analyzed. is the compensation factor for the number of repeated sequences in the gene data to be analyzed. is a natural constant, and max is a maximum value function. After ten calculations, the quality assessment characteristic values ​​are 0.89, 0.87, 0.98, 0.93, 0.67, 0.74, 0.77, 0.82, 0.84, and 0.98. The Agaricus bisporus genome was assembled multiple times to obtain the optimal value. The formula is used:

[0027] As a further method, the optimal value is obtained, and the specific calculation method is: ; In the formula, is the optimal value, is the quality assessment feature value of the gene data to be analyzed, contig To arrange the assembled results in descending order of length and add them up in order, when the length just exceeds 50% of the total length of the genome, the length of this contig is contig . Among them, contig The optimal values ​​calculated for 262.5 are 233.625, 228.375, 257.25, 244.125, 175.875, 194.25, 202.125, 215.25, 220.5, and 257.25, respectively. The value with the highest optimal value obtained by the bubble sort method is 257.25. This assembly result is selected as the optimal assembly result. (2) Select a reference genome based on the gene sequencing data to be analyzed, and construct a sequence alignment deep learning model based on the Transformer architecture. The specific implementation process is as follows: Based on the second-generation short-read sequencing data of Agaricus bisporus, obtain the genome of the same or closely related species as the reference genome, select the activation function and neural network structure pre-provided by PyTorch, define the input layer as contig data in json format, that is, format the gene data to be compared and convert it into json format input; define the hidden layer as sequence vectorization through Transformer technology, and then use the cosine similarity function to align each vectorized fragment to the vectorized data of the reference genome, and select the optimal result. Define the output layer as continuous gene data with a certain position relationship; use the stochastic gradient descent method to update the model parameters to minimize the loss function, adjust the computer computing resources, and verify the model. For example, the data randomly cut from the Agaricus bisporus reference genome can be input into the model for verification; (3) Based on the sequence alignment deep learning model, the gene sequencing data to be analyzed is input into the model to determine the positional relationship of the gene sequencing data to be analyzed. The specific implementation process is as follows: the reference genome is used to convert the input data set into json format through the program; the activation function ReLU and neural network structure Transformer pre-provided by PyTorch are selected, and the input layer is defined as contig data in json format, which is converted into json format input; the hidden layer is defined as sequence vectorization through Transfomer technology, and then each vectorized fragment is aligned to the vectorized data of the reference genome, and the cosine similarity function is used to select the optimal result. The output layer is defined as the continuous gene data with a determined positional relationship; the stochastic gradient descent method is used to update the parameters of the sequence alignment deep learning model based on the Transformer architecture to minimize the loss function, adjust the computer computing resources, and train and verify the sequence alignment deep learning model based on the Transformer architecture.

[0028] (4) Based on the genetic data with a determined positional relationship, the long-read sequencing data is aligned to the genetic data sequence fragment to determine the alignment result. The specific implementation process is as follows: Since the average coverage length of the long-read sequencing data is longer than that of the short-read sequencing data, the long-read sequence data is used to cover the gaps in the short-read sequencing data to determine the blank base sequences that are not determined by the short-read sequencing data; the blank base sequence position, blank base sequence, long-read sequencing data, and short-read sequencing data are converted into json format data for storage to obtain the alignment result in json format.

[0029] (5) Based on the alignment results and the number of long-read sequencing, a deep learning model Transformer is constructed to obtain the genetic data with unknown bases filled in the genetic data with a determined positional relationship. The specific implementation process is as follows: based on the alignment results in the json format of Agaricus bisporus optimized in step (4) and the third-generation long-read sequencing data, a genetic data gap filling deep learning model based on the Transformer architecture is constructed and trained as a tool for filling blank base sequences through the method provided by PyTorch. The model is used to take the genetic data with a determined positional relationship in step (3) as input data, define the input layer as genetic data in text format, define the hidden layer as sequence vectorization achieved through Transformer technology, and then use the cosine similarity function to align each vectorized fragment to the vectorized data of the reference genome, select the optimal result, match the optimal result with the input layer data, that is, concatenate the selected optimal result string and the input string to fill the blank sequence; define the output layer as the genetic data with unknown bases filled in the genetic data with a determined positional relationship, and obtain the genetic data with unknown bases filled; (6) Based on the optimal assembly result breakpoint information quantity characteristics and multiple gene assembly results, alternate tests are performed to build a support vector machine model to maximize the classification information model, determine whether the contig is a telomere or contains a telomere, and determine the telomere data. The specific implementation process is: Use the optimal assembly result of Agaricus bisporus in step (1) to calculate the breakpoint information quantity characteristics. The specific characteristic value is: Ratio ; Ratio is the feature of the number of breakpoint information, where T is the number of telomere sequences in the total number of data values ​​of each breakpoint of the long-read sequencing data aligned to the gene data with unknown bases filled in, and R is the total number of data values ​​of each breakpoint end of the long-read sequencing data aligned to the gene data with unknown bases filled in. The other gene assembly results are alternately aligned, and the obtained data is used as the data basis of the support vector machine to build a support vector machine model. The support vector machine model is used to judge the telomere sequence of the second-generation long-read sequencing data to obtain telomere information; (7) Based on the telomere data and the genetic data with the unknown bases filled in, a chromosome T2T genome is obtained. The specific implementation process is: based on the telomere information in step (6) and the genetic data with the unknown bases filled in step (5), the telomere information in step (6) is inserted into the genetic data with the unknown bases filled in step (5). After the genetic data is obtained, the format is optimized to finally obtain the chromosome T2T genome of Agaricus bisporus, which has a length of 3630890.

[0030] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

[0031] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0032] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for assembling a chromosome T2T genome based on artificial intelligence technology, characterized in that: The following steps are involved: Perform quality analysis on the gene sequencing data to be analyzed and select the optimal assembly result; Select a reference genome based on the gene sequencing data to be analyzed, and build a sequence alignment deep learning model based on the Transformer architecture; Based on the sequence alignment deep learning model based on the Transformer architecture, the gene sequencing data to be analyzed is input into the model to determine the positional relationship of the gene sequencing data to be analyzed, and the gene data with the determined positional relationship is obtained; Based on the gene data with determined positional relationships, the long-read sequencing data is aligned to the gene sequencing data with determined positional relationships to determine an alignment result; Based on the alignment results and long-read sequencing data, a deep learning model for filling gaps in genetic data based on the Transformer architecture was constructed to obtain genetic data with filled unknown bases in genetic data with determined positional relationships; Based on the quantitative characteristics of the breakpoint information of the optimal assembly result and the assembly results of multiple gene sequencing data, an alternate test was performed to build a support vector machine model to determine the telomere data; Based on the telomere data and the gene data with unknown bases filled in, the chromosome T2T genome was obtained.

2. The method for assembling a chromosome T2T genome based on artificial intelligence technology according to claim 1, characterized in that: The specific process of the step of selecting the optimal assembly result is: Acquire survey data information of the gene sequencing data to be analyzed, obtain quality assessment characteristic values ​​of the gene sequencing data to be analyzed based on the acquired survey data information of the gene sequencing data to be analyzed, and use the quality assessment characteristic values ​​of the gene sequencing data to be analyzed as the main basis for selecting the optimal assembly result; The gene sequencing data to be analyzed are assembled multiple times using the k-mer graph method to obtain assembly results, and the gene assembly results are calculated with the quality assessment characteristic values ​​of the gene sequencing data to be analyzed to obtain the optimal value; A comprehensive analysis is performed on the optimal values, the highest optimal value is selected as the optimal assembly result, and the remaining optimal values ​​are used as references.

3. The method for assembling a chromosome T2T genome based on artificial intelligence technology according to claim 2, characterized in that: The survey data information of the gene sequencing data to be analyzed specifically includes the sequencing error rate of the gene data to be analyzed, the number of repeated sequences of the gene data to be analyzed, the sequencing depth of the gene data to be analyzed, and the sequence coverage of the gene data to be analyzed.

4. The method for assembling a chromosome T2T genome based on artificial intelligence technology according to claim 2, characterized in that: The specific calculation method of the quality assessment characteristic value of the gene sequencing data to be analyzed is: ； In the formula, is the quality assessment feature value of the gene sequencing data to be analyzed, is the sequencing error rate of the gene data to be analyzed, is the number of repeated sequences in the gene data to be analyzed, is the sequence depth of the gene data to be analyzed, is the sequence coverage of the gene data to be analyzed, is the compensation factor for the sequencing error rate of the gene data to be analyzed. is the compensation factor for the depth of the gene data sequence to be analyzed. is the compensation factor for the coverage of the gene data sequence to be analyzed. is the compensation factor for the number of repeated sequences in the gene data to be analyzed. is a natural constant, and max is the maximum value function.

5. The method for assembling a chromosome T2T genome based on artificial intelligence technology according to claim 2, characterized in that: The specific calculation method for obtaining the optimal value is: ； In the formula, is the optimal value, It is the quality assessment characteristic value of the genetic data to be analyzed. Contig N50 is to arrange the assembled results from large to small in length and add them in order. When the length just exceeds 50% of the total length of the genome, the length of this contig is contig N50.

6. The method for assembling a chromosome T2T genome based on artificial intelligence technology according to claim 1, characterized in that: The specific process of constructing the sequence alignment deep learning model based on the Transformer architecture is as follows: Based on the species information and the gene sequencing data to be analyzed, the genome of the closely related species is obtained as the reference genome, and the obtained reference genome is converted into a json format input data set through a program; Select the activation function ReLU and neural network structure Transformer pre-provided by PyTorch, define the input layer as contig data in json format, format the gene data to be compared, and convert it into json format input; The hidden layer is defined as the sequence vectorization achieved through Transformer technology, and then each vectorized fragment is compared to the vectorized data of the reference genome, and the cosine similarity function is used to select the best result. The output layer is defined as the continuous gene data with a certain position relationship; The stochastic gradient descent method is used to update the parameters of the sequence alignment deep learning model based on the Transformer architecture to minimize the loss function, adjust the computer computing resources, and conduct training and verification of the sequence alignment deep learning model based on the Transformer architecture.

7. The method for assembling a chromosome T2T genome based on artificial intelligence technology according to claim 1, characterized in that: The specific process of the step of determining the positional relationship of the gene sequencing data to be analyzed is: Based on the sequence alignment deep learning model based on the Transformer architecture, the gene sequencing data to be analyzed is input into the model, and the output results are converted into a common format to determine the positional relationship of the gene sequencing data to be analyzed.

8. The method for assembling a chromosome T2T genome based on artificial intelligence technology according to claim 1, characterized in that: The specific process of aligning the long-read sequencing data to the gene data sequence fragment and determining the alignment result is as follows: Since the average coverage length of long-read sequencing data is longer than that of short-read sequencing data, the scoring matrix method is used to determine the blank base sequences that are not determined in the short-read sequencing data; The blank base sequence position, blank base sequence, long read sequencing data, and short read sequencing data are converted into json format data for storage to obtain the comparison results in json format.

9. The method for assembling a chromosome T2T genome based on artificial intelligence technology according to claim 1, characterized in that: The specific process of the step of obtaining the gene data with unknown bases filled in the gene data with determined positional relationship is as follows: Constructing a gene data gap filling deep learning model based on the Transformer architecture according to the json format comparison results and long-read sequencing data described in claim 8; Using PyTorch, define the input layer as genetic data in text format; Define the hidden layer to vectorize the sequence using Transformer technology, then align each vectorized fragment to the vectorized data of the reference genome, select the best result through the cosine similarity function, match the best result with the input layer data, and fill in the blank sequence; The output layer is defined as the gene data with unknown bases filled in the gene data with determined positional relationships; Based on the constructed Transformer-based deep learning model for genetic data gap filling, genetic data with filled unknown bases in genetic data with determined positional relationships are obtained.

10. The method for assembling a chromosome T2T genome based on artificial intelligence technology according to claim 1, characterized in that: The specific process of constructing the support vector machine model and determining the telomere data step is: Based on the quantitative characteristics of the optimal assembly result breakpoint information and the multiple gene assembly results as described in claim 1 as sample points, a support vector machine model is constructed; Based on the constructed support vector machine model, the long-read sequencing data is classified and judged to determine the telomere sequence; Based on the determined telomere sequence and the gene data with unknown bases filled in the gene data with determined positional relationship as claimed in claim 9, comparison is performed to determine the telomere position; The telomere position and telomere sequence are combined into binary data to obtain telomere information.

11. The method for assembling a chromosome T2T genome based on artificial intelligence technology according to claim 10, characterized in that: The specific calculation method of the optimal assembly result breakpoint information quantity feature is: ； Ratio is the feature of the number of breakpoint information, where T is the number of data with telomere sequences in the total number of data values ​​of each breakpoint of the long-read sequencing data aligned to the gene data with unknown bases filled in, and R is the total number of data values ​​of each breakpoint end of the long-read sequencing data aligned to the gene data with unknown bases filled in.

12. The method for assembling a chromosome T2T genome based on artificial intelligence technology according to claim 1, characterized in that: The specific process of the step of obtaining the chromosome T2T genome is as follows: Based on the genetic data with filled unknown bases in the genetic data with determined positional relationship as described in claim 9 and the telomere information as described in claim 10, the telomere sequence is inserted into the genetic data with filled unknown bases through the corresponding quantitative positional relationship to obtain chromosome T2T genetic data; The chromosome T2T genome is obtained by formatting the gene data of chromosome T2T.

Images