Model construction method and device, storage medium and program product

Abstract

A model construction method and apparatus, a storage medium and a program product, the method comprising: obtaining sequencing data of a plurality of training samples, the plurality of training samples at least comprising a first type of samples and a second type of samples, the sequencing data comprising a plurality of sequencing sequences; the sequencing data of each training sample are compared to the first genome and the second genome respectively, first genome segment data and second genome segment data are obtained, the first genome segment data comprise a plurality of sequencing sequences of the compared first genome, and the second genome segment data comprise a plurality of sequencing sequences of the compared second genome; the second genome fragment data comprises a plurality of sequencing sequences for comparing second genomes; generating sample feature data according to the first genome fragment data and the second genome fragment data; and performing classification model construction by using the sample feature data.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to, but are not limited to, the field of bioinformatics, and in particular to a model building method and device, a storage medium, and a program product. Background Art

[0002] Chronic active Epstein-Barr virus infection (CAEBV) is a chronic, non-self-healing disease caused by primary infection of immune cells such as NK / T cells by the Epstein-Barr virus. During its development, it may transform into EBV-associated lymphoma. Its incidence is low, and its pathogenesis is currently unclear. Clinically, quantitative real-time PCR (qPCR) is mainly used to detect the EBV nucleic acid load as an auxiliary diagnosis. Its detection rate is limited by the DNA load in the sample to be tested. Summary of the Invention

[0003] The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims.

[0004] The present disclosure provides a model building method, including: Acquire sequencing data of a plurality of training samples, where the plurality of training samples include at least a first type of sample and a second type of sample, and the sequencing data includes a plurality of sequencing sequences; Aligning the sequencing data of each training sample to the first genome and the second genome, respectively, to obtain first genome segment data and second genome segment data, wherein the first genome segment data includes a plurality of sequencing sequences aligned to the first genome, and the second genome segment data includes a plurality of sequencing sequences aligned to the second genome; generating sample feature data according to the first genome segment data and the second genome segment data; A classification model is constructed using the sample feature data.

[0005] An embodiment of the present disclosure also provides a model building device, comprising a memory; and a processor connected to the memory, wherein the memory is used to store instructions, and the processor is configured to execute the steps of the model building method described in any embodiment of the present disclosure based on the instructions stored in the memory.

[0006] An embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the model building method described in any embodiment of the present disclosure.

[0007] An embodiment of the present disclosure further provides a program product, comprising instructions. When the computer program product is executed by a computer, the instructions execute the model building method as described in any embodiment of the present disclosure.

[0008] The present disclosure also provides a model building device, including a data acquisition module, a comparison module, a feature generation module, and a model building module, wherein: The data acquisition module is configured to acquire sequencing data of a plurality of training samples, wherein the plurality of training samples include at least a first type of sample and a second type of sample, and the sequencing data includes a plurality of sequencing sequences; The alignment module is configured to align the sequencing data of each training sample to the first genome and the second genome, respectively, to obtain first genome segment data and second genome segment data, wherein the first genome segment data includes a plurality of sequencing sequences aligned to the first genome, and the second genome segment data includes a plurality of sequencing sequences aligned to the second genome; The feature generation module is configured to generate sample feature data based on the first genome segment data and the second genome segment data; The model building module is configured to use the sample feature data to build a classification model.

[0009] The model construction method and apparatus, storage medium and program product of the embodiments of the present disclosure, by acquiring sequencing data of multiple training samples, aligning the sequencing data of each training sample to the first genome and the second genome respectively, obtaining first genome fragment data and second genome fragment data, generating sample feature data based on the first genome fragment data and the second genome fragment data, and using the sample feature data to construct a classification model, are not limited by the DNA load limit in the sample, have low requirements for the DNA load in the sample, have a high tolerance for low sample quality, can obtain a lower detection limit at a higher sequencing depth, and can be tested at very low viral loads relative to PCR. When the first type of sample is cfDNA from a CAEBV patient and the second type of sample is cfDNA from one or more EBV-related tumor patients, the embodiments of the present disclosure can simultaneously acquire and compare the cfDNA fragment genomic level changes of CAEBV patients and EBV-related tumor patients, and have higher reliability in assessing the probability of CAEBV transforming into a tumor.

[0010] Other features and advantages of the present disclosure will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present disclosure. Other advantages of the present disclosure can be realized and obtained through the solutions described in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings are used to provide an understanding of the technical solution of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the technical solution of the present disclosure and do not constitute a limitation to the technical solution of the present disclosure.

[0012] Figure 1 A flowchart of a model building method provided by an exemplary embodiment of the present disclosure; Figure 2 is a schematic diagram of sample feature data of an exemplary embodiment of the present disclosure; Figure 3 A schematic structural diagram of a model building device provided by an exemplary embodiment of the present disclosure; Figure 4 A schematic structural diagram of another model building device provided by an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0013] The present disclosure describes a plurality of embodiments, but this description is exemplary rather than restrictive, and it will be apparent to those skilled in the art that there may be more embodiments and implementations within the scope of the embodiments described in the present disclosure. Although many possible feature combinations are shown in the drawings and discussed in the detailed description, many other combinations of the disclosed features are also possible. Unless specifically limited, any feature or element of any embodiment may be used in combination with any other feature or element in any other embodiment, or may replace any other feature or element in any other embodiment.

[0014] The present disclosure includes and contemplates combinations of features and elements known to those of ordinary skill in the art. The disclosed embodiments, features, and elements of the present disclosure may also be combined with any conventional features or elements to form a unique inventive solution defined by the claims. Any features or elements of any embodiment may also be combined with features or elements from other inventive solutions to form another unique inventive solution defined by the claims. Therefore, it should be understood that any feature shown and / or discussed in this disclosure may be implemented individually or in any appropriate combination. Therefore, the embodiments are not subject to other limitations except for the limitations set forth in the appended claims and their equivalents. In addition, various modifications and changes may be made within the scope of protection of the appended claims.

[0015] In addition, when describing representative embodiments, the specification may have presented the method and / or process as a specific sequence of steps. However, to the extent that the method or process does not rely on the specific order of the steps described herein, the method or process should not be limited to the steps in the specific order described. As will be understood by those skilled in the art, other orders of steps are also possible. Therefore, the specific order of the steps set forth in the specification should not be interpreted as a limitation on the claims. In addition, the claims to the method and / or process should not be limited to performing their steps in the order written, and those skilled in the art can readily understand that these orders can be changed and still remain within the spirit and scope of the disclosed embodiments.

[0016] like Figure 1 As shown, the embodiment of the present disclosure provides a model building method, including: Step 101: Acquire sequencing data of a plurality of training samples, where the plurality of training samples include at least a first type of sample and a second type of sample, and the sequencing data includes a plurality of sequencing sequences; Step 102: Align the sequencing data of each training sample to the first genome and the second genome, respectively, to obtain first genome segment data and second genome segment data, where the first genome segment data includes multiple sequencing sequences aligned to the first genome, and the second genome segment data includes multiple sequencing sequences aligned to the second genome; Step 103: Generate sample feature data based on the first genome segment data and the second genome segment data; Step 104: Use the sample feature data to build a classification model.

[0017] The model construction method provided by the embodiments of the present disclosure obtains sequencing data of multiple training samples, aligns the sequencing data of each training sample to the first genome and the second genome, obtains first genome fragment data and second genome fragment data, generates sample feature data based on the first genome fragment data and the second genome fragment data, and uses the sample feature data to construct a classification model. The method is not limited by the DNA load limit in the sample, has low requirements for the DNA load in the sample, has a high tolerance for low sample quality, can obtain a lower detection limit at a higher sequencing depth, and can be tested at very low viral loads compared to PCR.

[0018] The classification model constructed using the model construction method of the embodiment of the present disclosure can be used as a classification model to distinguish whether the sample to be tested is a first type of sample or a second type of sample.

[0019] For example, in step 101, the first type of sample can be circulating free DNA (cfDNA) from a CAEBV patient, and the second type of sample can be cfDNA from one or more EBV-related tumor patients, however, this disclosure is not limited thereto. The disclosed embodiments can simultaneously acquire and compare omics-level changes in cfDNA fragments from CAEBV patients and EBV-related tumor patients, providing greater reliability in assessing the probability of CAEBV transforming into a tumor.

[0020] In the embodiment of the present disclosure, when the second type of sample is the free DNA of a patient with an EB virus-related tumor, the classification model constructed using the model construction method of the embodiment of the present disclosure is a binary classification model; when the second type of sample is the free DNA of patients with multiple EB virus-related tumors, the classification model constructed using the model construction method of the embodiment of the present disclosure is a multi-classification model.

[0021] cfDNA refers to highly fragmented DNA that circulates outside cells. Clinically, analysis of cfDNA fragmentation data can be used for non-invasive disease diagnosis, disease monitoring, and treatment response assessment. The study of cfDNA fragmentation is crucial for understanding disease biology and developing relevant biomarkers.

[0022] In the disclosed embodiments, the cfDNA sample source can be peripheral blood plasma, and the sequencing type can be whole genome sequencing (WGS) or whole exome sequencing (WES). WGS performs high-throughput sequencing of the entire genome, while WES uses sequence capture technology to capture and enrich DNA in the exon regions of the entire genome for high-throughput sequencing.

[0023] In the disclosed embodiments, sequencing data includes multiple sequencing sequences (reads). Reads are short sequence fragments obtained in high-throughput sequencing. Sequencing the entire genome will generate millions of reads.

[0024] In step 101, cfDNA is extracted from the plasma of patients with CAEBV and EBV-related tumors, and WES or WGS sequencing is performed to obtain off-machine sequencing data. The sequencing type can be paired-end sequencing of 2*100bp or 2*150bp.

[0025] For example, using the second type of sample, cell-free DNA from patients with peripheral T-cell lymphoma (PTCL), as an example, cfDNA is extracted from the plasma of CAEBV and PTCL patients. After library construction, WGS or WES sequencing is performed. Because WGS sequencing covers the entire genome, it requires a relatively high initial cfDNA input. Therefore, for WGS, the cfDNA input per sample is at least 50 to 100 nanograms (ng); for WES, the cfDNA input per sample is between 40 and 50 ng. The completed WGS or WES library is then sequenced on a sequencing machine to generate raw sequencing data.

[0026] In some exemplary embodiments, in step 102, the first genome may be an Epstein-Barr virus genome, and the second genome may be a human genome.

[0027] In the embodiments of the present disclosure, the human genome may be version hg19 (Human Genome Version 19) or version hg38 (Human Genome Version 38). Both hg19 and hg38 are human genome version numbers, representing human genome reference sequences released at different time points and widely used as standards in scientific research. This disclosure uses version hg38 as an example. The EBV viral genome can be obtained from the National Center of Biotechnology Information (NCBI) database, and the human hg38 genome can be obtained from the Ensembl database.

[0028] In some exemplary embodiments, before step 102, the method further comprises: preprocessing the acquired sequencing data.

[0029] In the disclosed embodiment, the obtained sequencing data is preprocessed, including performing data quality control and data filtering on the sequencing data.

[0030] In the disclosed embodiments, data filtering is primarily performed on low-quality data. For example, low-quality data includes at least one of the following: sequences with an average base quality below a preset quality threshold Q, sequences containing linkers, sequences with an excessively high proportion of N bases, and sequences with a length below a preset length threshold.

[0031] In step 102, the following operations are performed on the sequencing data of CAEBV patients and PTCL patients, respectively: the sequencing data are quality controlled and filtered to remove low-quality sequences and redundant connectors, and then the sequencing data are aligned to the EBV genome and the human genome, respectively, to obtain the aligned genomic fragment data, which are divided into EBV genome fragment data and human genome fragment data.

[0032] In some exemplary embodiments, in step 103, the sample characteristic data may include: first genome segment coverage, wherein the first genome segment coverage may be calculated according to the following formula: ; Where C is the coverage of the first genome fragment, is the number of sequencing sequences of the first genome compared, is the number of sequencing sequences aligned to the second genome.

[0033] In some exemplary embodiments, in step 103 , the sample characteristic data may include: a first genomic short fragment ratio and a second genomic short fragment ratio.

[0034] Among them, the first genome short fragment ratio can be generated by the following method: dividing the first genome into multiple first genome intervals; determining the first genome interval to which each sequencing sequence in the first genome fragment data belongs; determining the number of short fragments and the number of long fragments in each first genome interval based on the length of the sequencing sequence; and calculating the short fragment ratio corresponding to each first genome interval based on the number of short fragments and the number of long fragments in each first genome interval.

[0035] The short fragment ratio of the second genome can be generated by the following method: dividing the second genome into multiple second genome intervals; determining the second genome interval to which each sequencing sequence in the second genome fragment data belongs; determining the number of short fragments and the number of long fragments in each second genome interval based on the length of the sequencing sequence; and calculating the short fragment ratio corresponding to each second genome interval based on the number of short fragments and the number of long fragments in each second genome interval.

[0036] In the disclosed embodiments, when aligning the sequencing data of each training sample to the EBV genome, the EBV genome can be divided into m1 first genomic intervals according to the first window length. For example, as shown in Table 1, the first window length can be 1 kb. In this case, m1 is 172, meaning the EBV genome is divided into 172 first genomic intervals of 1 kb. In Table 1, the first column indicates the EBV species number, the second column indicates the starting position of each first genomic interval, and the third column indicates the ending position of each first genomic interval.

[0037] NC_007605.1 0 1000 NC_007605.1 1000 2000 NC_007605.1 2000 3000 NC_007605.1 3000 4000 NC_007605.1 4000 5000 NC_007605.1 5000 6000 NC_007605.1 6000 7000 NC_007605.1 7000 8000 NC_007605.1 8000 9000 NC_007605.1 9000 10000 NC_007605.1 10000 11000 NC_007605.1 11000 12000 NC_007605.1 12000 13000 NC_007605.1 13000 14000 NC_007605.1 14000 15000 NC_007605.1 15000 16000 NC_007605.1 16000 17000 NC_007605.1 17000 18000 NC_007605.1 18000 19000 NC_007605.1 19000 20000 …… …… …… Table 1 In the disclosed embodiment, when aligning the sequencing data of each training sample to the hg38 genome, the hg38 genome can be divided into m2 second genomic intervals according to the second window length. For example, as shown in Table 2, the second window length can be 5 Mb. In this case, m2 is 587, that is, the hg38 genome is divided into 587 second genomic intervals of 5 Mb. In Table 2, the first column is the chromosome number, the second column is the starting position of each second genomic interval, and the third column is the ending position of each second genomic interval.

[0038] 1 0 5000000 1 5000000 10000000 1 10000000 15000000 1 15000000 20000000 1 20000000 25000000 1 25000000 30000000 1 30000000 35000000 1 35000000 40000000 1 40000000 45000000 1 45000000 50000000 1 50000000 55000000 1 55000000 60000000 1 60000000 65000000 1 65000000 70000000 1 70000000 75000000 1 75000000 80000000 1 80000000 85000000 1 85000000 90000000 1 90000000 95000000 1 95000000 100000000 …… …… …… Table 2 In the disclosed embodiment, when the sequencing data of each training sample is aligned to the EBV genome or the hg38 genome, the start and end information of each sequencing sequence can be intersected with the previously divided genome interval information, and the genome interval to which each sequencing sequence belongs can be determined based on the intersection ratio of each sequencing sequence with one or more genome intervals.

[0039] In the disclosed embodiment, determining the genomic interval to which each sequencing sequence belongs based on the intersection ratio between each sequencing sequence and one or more genomic intervals includes: When the intersection ratio between a sequencing sequence and a genomic interval is greater than 50%, the sequencing sequence is determined to fall within the genomic interval.

[0040] In the disclosed embodiment, when the intersection ratio is greater than 50% of the sequencing sequence length of the sequencing sequence, the sequencing sequence is considered to fall within the corresponding genomic interval, so the distribution of sequencing sequences in all genomic intervals can be obtained.

[0041] In some exemplary embodiments, determining the number of short fragments and the number of long fragments in each first genomic interval (or second genomic interval) based on the length of the sequencing sequence includes: Setting the short fragment length range and the long fragment length range corresponding to the first genomic interval (or the second genomic interval); According to the length of each sequencing sequence, each sequencing sequence is judged to be a short fragment or a long fragment; The number of short fragments and the number of long fragments in each first genomic interval (or second genomic interval) are counted.

[0042] In the embodiment of the present disclosure, the short fragment length range and the long fragment length range can be set as needed.

[0043] In some exemplary embodiments, the length of the short fragment may range from n1 bp to n2 bp, and the length of the long fragment may range from (n2+1) bp to n3 bp, where n1 <n2<n3。

[0044] In some exemplary embodiments, n1≧100, 150≦n2≦160, and n3≦250.

[0045] For example, the short fragment length ranges from 100 bp to 150 bp, and the long fragment length ranges from 151 bp to 220 bp. That is, sequencing sequences with a length of 100 bp to 150 bp are considered short fragments, and sequencing sequences with a length of 151 bp to 220 bp are considered long fragments. The number of short fragments and the number of long fragments within each first genomic region (or second genomic region) are counted.

[0046] In some exemplary embodiments, the short fragment ratio corresponding to the first genomic interval or the second genomic interval is calculated according to the following formula: short fragment ratio=number of short fragments / number of long fragments.

[0047] According to the number of long and short fragments in the sequencing sequences distributed in each interval, the ratio of the number of short fragments to the number of long fragments can be calculated, and this ratio is used as the short fragment ratio corresponding to the genomic interval. The short fragment ratio of the first genomic interval is recorded as Fe, and the short fragment ratio of the second genomic interval is recorded as Fh.

[0048] In other exemplary embodiments, determining the number of short fragments and the number of long fragments in each first genomic interval based on the length of the sequencing sequence may further include: Comparing and counting the lengths of all sequencing sequences of the first genome, and determining the length of the sequencing sequence with the largest number of counts as a first threshold length; Determine whether each sequencing sequence is a short fragment or a long fragment based on its length, wherein a sequencing sequence with a length less than a first threshold length is a short fragment, and a sequencing sequence with a length greater than or equal to the first threshold length is a long fragment; The number of short fragments and long fragments in each first genomic interval was counted.

[0049] In the disclosed embodiment, the lengths of the fragments of the aligned EBV genome can be proportionally counted in a statistical manner to draw a histogram, where the fragments on the left side of the peak are short fragments and the fragments on the right side of the peak are long fragments.

[0050] In some exemplary embodiments, the model construction method further includes: performing dimensionality reduction processing on the short fragment ratios of multiple first genomic intervals and the short fragment ratios of multiple second genomic intervals respectively; or, splicing the short fragment ratios of multiple first genomic intervals and the short fragment ratios of multiple second genomic intervals by column, and performing dimensionality reduction processing on the spliced ​​short fragment ratios of the multiple first genomic intervals and the short fragment ratios of the multiple second genomic intervals.

[0051] In some exemplary embodiments, in step 103 , the sample characteristic data may include: a first genome end motif ratio and a second genome end motif ratio.

[0052] Among them, the first genome terminal motif ratio and the second genome terminal motif ratio are generated by the following method: for each sequencing sequence in the first genome fragment data and the second genome fragment data, the arrangement of the first k bases at the 5' end of each sequencing sequence is determined, and the ratios of the occurrence of multiple terminal motifs in the first genome and the second genome are counted. Each terminal motif corresponds to an arrangement of the first k bases at the 5' end of the sequencing sequence, where k is a natural number greater than or equal to 4.

[0053] In the embodiment of the present disclosure, the k bp starting from the 5' end of all the fragments of the aligned genome are called the terminal motif. For example, when k is 4, the 4 bp base arrangement sequence starting from the 5' end of all the fragments of the aligned EBV genome and the human genome is taken out. Since there are four possible bases A / T / C / G at each position, there are There are 256 possible terminal motifs in total. The ratio of each terminal motif is counted. The ratio of the first genome terminal motif is recorded as Re, and the ratio of the second genome terminal motif is recorded as Rh.

[0054] In some exemplary embodiments, the model building method further comprises: performing dimensionality reduction processing on the ratios of occurrence of the plurality of terminal motifs in the first genome and the second genome.

[0055] In the embodiment of the present disclosure, when the short fragment ratios of multiple first genomic intervals, the short fragment ratios of multiple second genomic intervals, and the ratios of the occurrence of multiple terminal motifs of the first genome and the second genome are subjected to dimensionality reduction processing, as shown in Table 3, the short fragment ratios Fe of multiple first genomic intervals, the short fragment ratios Fh of multiple second genomic intervals, the multiple terminal motif ratios Re of the first genome, and the multiple terminal motif ratios Rh of the second genome can be spliced ​​by column (in the embodiment of the present disclosure, the short fragment ratio Fe of the first genomic interval contains 172 columns, the short fragment ratio Fh of the second genomic interval contains 587 columns, the multiple terminal motif ratios Re of the first genome contains 256 columns, and the multiple terminal motif ratios Rh of the second genome contains 256 columns), and the dimensionality reduction processing is performed on all the spliced ​​columns. In other examples, dimensionality reduction can be performed separately for the short fragment ratios Fe of multiple first genomic intervals, the short fragment ratios Fh of multiple second genomic intervals, the multiple terminal motif ratios Re of the first genome, and the multiple terminal motif ratios Rh of the second genome. The dimensions of Fe, Fh, Re, and Rh after dimensionality reduction can be the same or different, and this disclosure is not limited thereto. Dimensionality reduction methods can include principal component analysis (PCA) and linear discriminant analysis (LDA). Taking PCA as an example, the number of dimensions after dimensionality reduction can be manually set to X (X greater than 1, for example, X = 10), or the variance explained ratio can be set to Y% (Y greater than 90, for example, Y% = 95%) to automatically select the number of dimensions after dimensionality reduction. The new eigenvalues ​​after dimensionality reduction and the first genomic segment coverage (EBV segment coverage) C are concatenated column by column to form the original dataset, and a classification model is constructed on this original dataset.

[0056]

[0057] Table 3 The disclosed embodiments perform dimensionality reduction on the short fragment ratios Fe of multiple first genomic intervals, the short fragment ratios Fh of multiple second genomic intervals, the ratios Re of multiple terminal motifs in the first genome, and the ratios Rh of multiple terminal motifs in the second genome. This removes redundant information from the data, reduces computing resource overhead, accelerates model calculations, and improves model accuracy. Since the first genome segment coverage (EBV segment coverage) C only has one column, dimensionality reduction is not required.

[0058] In some exemplary embodiments, in step 104 , the classification model may be any one of the following models: a logistic regression (LR) classification model, a random forest (RF) classification model, a support vector machine (SVM) classification model, etc.

[0059] When building a classification model, the training samples can be divided into a training set and a validation set according to a preset ratio, using a K-fold cross-validation method, where K is greater than or equal to 3. For example, on the original data set, the cross-validation method is used to set the fold number K to 5, and the training set and validation set are split in a ratio of 7:3. Classification algorithms such as logistic regression, random forest, and support vector machine can be selected to train the model on the training set, and the model performance is evaluated on the validation set. The average AUC (Area Under Curve) of different classifiers under 5 cross-validations is calculated, and the classifier model with the highest AUC average is selected as the final prediction classification model.

[0060] The classification model of the disclosed embodiment can give the predicted probabilities of CAEBV patients and EBV-related tumor patients, and thus can evaluate the likelihood of CAEBV patients transforming into tumor patients. When predicting the sample to be tested, the sequencing data of the sample to be tested is first obtained, and the sequencing data of the sample to be tested is respectively aligned to the first genome and the second genome to obtain the first genome fragment data and the second genome fragment data. The sample characteristic data of the sample to be tested is generated based on the first genome fragment data and the second genome fragment data. The sample characteristic data of the sample to be tested is input into the classification model, and the predicted probabilities of CAEBV patients and EBV-related tumor patients are output. The likelihood of CAEBV transforming into EBV-related tumors can be judged based on the probability of belonging to an EBV-related tumor patient.

[0061] The model construction method of the embodiment of the present disclosure has higher reliability in assessing the probability of CAEBV transforming into tumors by simultaneously acquiring and comparing the cfDNA fragment omics-level changes in CAEBV patients and EBV-related tumor patients. In addition, the model construction method is not limited by the DNA load limit in the sample, has low requirements for the DNA load in the sample, has a high tolerance for low sample quality, and can obtain lower detection limits at higher sequencing depths. Compared with PCR, it can be tested at very low viral loads.

[0062] like Figure 3 As shown, the embodiment of the present disclosure further provides a model building device, including a data acquisition module 310, a comparison module 320, a feature generation module 330 and a model building module 340, wherein: The data acquisition module 310 is configured to acquire sequencing data of a plurality of training samples, where the plurality of training samples include at least a first type of sample and a second type of sample, and the sequencing data includes a plurality of sequencing sequences; an alignment module 320 configured to align the sequencing data of each training sample to the first genome and the second genome, respectively, to obtain first genome segment data and second genome segment data, wherein the first genome segment data includes a plurality of sequencing sequences aligned to the first genome, and the second genome segment data includes a plurality of sequencing sequences aligned to the second genome; A feature generation module 330 is configured to generate sample feature data based on the first genome segment data and the second genome segment data; The model building module 340 is configured to build a classification model using the sample feature data.

[0063] In some exemplary embodiments, the first type of sample is cell-free DNA from patients with chronic active EBV disease, and the second type of sample is cell-free DNA from patients with one or more EBV-associated tumors.

[0064] In some exemplary embodiments, the first genome is an Epstein-Barr virus genome and the second genome is a human genome.

[0065] In some exemplary embodiments, the sample feature data includes: first genome segment coverage, wherein the feature generation module 330 calculates the first genome segment coverage C according to the following formula: ; in, is the number of sequencing sequences aligned with the first genome, is the number of sequencing sequences aligned with the second genome.

[0066] In some exemplary embodiments, the sample characteristic data includes: a first genomic short fragment ratio and a second genomic short fragment ratio; The feature generation module 330 generates the first genome short fragment ratio by the following method: dividing the first genome into multiple first genome intervals; determining the first genome interval to which each sequencing sequence in the first genome segment data belongs; determining the number of short fragments and the number of long fragments in each first genome interval based on the length of the sequencing sequence; and calculating the short fragment ratio corresponding to each first genome interval based on the number of short fragments and the number of long fragments in each first genome interval. The feature generation module 330 generates the second genome short fragment ratio by the following method: dividing the second genome into multiple second genome intervals; determining the second genome interval to which each sequencing sequence in the second genome fragment data belongs; determining the number of short fragments and the number of long fragments in each second genome interval based on the length of the sequencing sequence; and calculating the short fragment ratio corresponding to each second genome interval based on the number of short fragments and the number of long fragments in each second genome interval.

[0067] In some exemplary embodiments, the feature generation module 330 calculates the short fragment ratio corresponding to the first genomic interval or the second genomic interval according to the following formula: short fragment ratio=number of short fragments / number of long fragments.

[0068] In some exemplary embodiments, the feature generation module 330 is further configured to: perform dimensionality reduction processing on the short fragment ratios of multiple first genomic intervals and the short fragment ratios of multiple second genomic intervals respectively; or, splice the short fragment ratios of multiple first genomic intervals and the short fragment ratios of multiple second genomic intervals by column, and perform dimensionality reduction processing on the spliced ​​short fragment ratios of the multiple first genomic intervals and the short fragment ratios of the multiple second genomic intervals.

[0069] In some exemplary embodiments, the sample characteristic data includes: a first genome end motif ratio and a second genome end motif ratio; Among them, the feature generation module 330 generates the first genome terminal motif ratio and the second genome terminal motif ratio by the following method: for each sequencing sequence in the first genome fragment data and the second genome fragment data, determine the arrangement of the first k bases at the 5' end of each sequencing sequence, and count the ratios of the occurrence of multiple terminal motifs of the first genome and the second genome, each terminal motif corresponds to an arrangement of the first k bases at the 5' end of the sequencing sequence, where k is a natural number greater than or equal to 4.

[0070] In some exemplary embodiments, the feature generation module 330 is further configured to perform dimensionality reduction processing on the ratios of occurrence of the various terminal motifs in the first genome and the second genome.

[0071] In some exemplary embodiments, the classification model is any one of the following: a logistic regression classification model, a random forest classification model, and a support vector machine classification model.

[0072] An embodiment of the present disclosure also provides a model building device, comprising a memory; and a processor connected to the memory, wherein the memory is used to store instructions, and the processor is configured to execute the steps of the model building method described in any embodiment of the present disclosure based on the instructions stored in the memory.

[0073] like Figure 4 As shown, in one example, the model building device may include: a processor 410, a memory 420, a bus system 430, and a transceiver 440, wherein the processor 410, the memory 420, and the transceiver 440 are connected via the bus system 430, the memory 420 is used to store instructions, and the processor 410 is used to execute the instructions stored in the memory 420 to control the transceiver 440 to transmit and receive signals. Specifically, under the control of the processor 410, the transceiver 440 may obtain sequencing data of multiple training samples, the multiple training samples including at least a first type of sample and a second type of sample, the sequencing data including a plurality of sequencing sequences, the processor 410 aligns the sequencing data of each training sample to the first genome and the second genome, respectively, to obtain first genome segment data and second genome segment data, the first genome segment data including a plurality of sequencing sequences aligned to the first genome, and the second genome segment data including a plurality of sequencing sequences aligned to the second genome; generates sample feature data based on the first genome segment data and the second genome segment data; and uses the sample feature data to build a classification model.

[0074] It should be understood that the processor 410 may be a central processing unit (CPU), or may be 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. The general-purpose processor may be a microprocessor or any conventional processor.

[0075] The memory 420 may include a read-only memory and a random access memory, and provides instructions and data to the processor 410. A portion of the memory 420 may also include a non-volatile random access memory. For example, the memory 420 may also store information about the device type.

[0076] In addition to the data bus, the bus system 430 may also include a power bus, a control bus, and a status signal bus. Figure 4 Various buses are labeled as bus system 430 .

[0077] During implementation, the processing performed by the processing device can be completed by the hardware integrated logic circuit in the processor 410 or by instructions in the form of software. That is, the method steps of the embodiment of the present disclosure can be embodied as being executed by a hardware processor, or by a combination of hardware and software modules in the processor. The software module can be located in a storage medium such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 420, and the processor 410 reads the information in the memory 420 and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

[0078] The present disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the model building method described in any of the embodiments of the present disclosure. The model building method driven by executing executable instructions is substantially the same as the model building method described in the above embodiments of the present disclosure and is not further described here.

[0079] In some possible implementations, various aspects of the model building method provided by the present disclosure may also be implemented in the form of a program product, which includes program code. When the program product is run on a computer device, the program code is used to enable the computer device to execute the steps of the model building method according to various exemplary embodiments of the present disclosure described above in this specification. For example, the computer device may execute the model building method described in the embodiments of the present disclosure.

[0080] The program product may utilize any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or component, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0081] Those skilled in the art will appreciate that all or some of the steps, systems, and functional modules / units in the methods, systems, and devices disclosed above may be implemented as software, firmware, hardware, or any combination thereof. In hardware implementations, the division between functional modules / units described above does not necessarily correspond to the division between physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on computer-readable media, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is well known to those skilled in the art, the term computer storage media encompasses both volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those skilled in the art, communication media typically embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0082] It should be noted that the above-described embodiments or implementations are merely illustrative and not restrictive. Therefore, the present disclosure is not limited to what is specifically shown and described herein. Various modifications, substitutions, or omissions may be made to the forms and details of the implementations without departing from the scope of the present disclosure.

Claims

1. A model building method, characterized in that: include: Acquire sequencing data of a plurality of training samples, wherein the plurality of training samples include at least a first type of sample and a second type of sample, and the sequencing data includes a plurality of sequencing sequences; Aligning the sequencing data of each training sample to the first genome and the second genome, respectively, to obtain first genome segment data and second genome segment data, wherein the first genome segment data includes a plurality of sequencing sequences aligned to the first genome, and the second genome segment data includes a plurality of sequencing sequences aligned to the second genome; Generate sample characteristic data according to the first genome segment data and the second genome segment data; The sample feature data is used to construct a classification model.

2. The method according to claim 1, characterized in that The first type of sample is free DNA from patients with chronic active EB virus disease, and the second type of sample is free DNA from patients with one or more EB virus-related tumors; The first genome is an Epstein-Barr virus genome, and the second genome is a human genome.

3. The method according to claim 1, characterized in that The sample characteristic data includes: a first genome segment coverage rate, and the first genome segment coverage rate is calculated according to the following formula: ； Wherein, C is the coverage rate of the first genome fragment, is the number of sequencing sequences aligned with the first genome, is the number of sequencing sequences aligned with the second genome.

4. The method according to claim 1, characterized in that: The sample characteristic data includes: a first genome short fragment ratio and a second genome short fragment ratio; The first genome short fragment ratio is generated by the following method: dividing the first genome into a plurality of first genome intervals; determining the first genome interval to which each sequencing sequence in the first genome fragment data belongs; determining the number of short fragments and the number of long fragments in each first genome interval according to the length of the sequencing sequence; and calculating the short fragment ratio corresponding to each first genome interval according to the number of short fragments and the number of long fragments in each first genome interval; The second genome short fragment ratio is generated by the following method: dividing the second genome into multiple second genome intervals; determining the second genome interval to which each sequencing sequence in the second genome fragment data belongs; determining the number of short fragments and the number of long fragments in each second genome interval according to the length of the sequencing sequence; calculating the short fragment ratio corresponding to each second genome interval according to the number of short fragments and the number of long fragments in each second genome interval, wherein the short fragment ratio = the number of short fragments / the number of long fragments.

5. The method according to claim 4, characterized in that The method further comprises: performing dimensionality reduction processing on the short fragment ratios of the plurality of first genomic intervals and the short fragment ratios of the plurality of second genomic intervals respectively; or, The short fragment ratios of the plurality of first genomic intervals and the short fragment ratios of the plurality of second genomic intervals are spliced ​​in columns, and dimensionality reduction processing is performed on the spliced ​​short fragment ratios of the plurality of first genomic intervals and the short fragment ratios of the plurality of second genomic intervals.

6. The method according to claim 1, characterized in that The sample characteristic data includes: a first genome terminal motif ratio and a second genome terminal motif ratio; Among them, the first genome terminal motif ratio and the second genome terminal motif ratio are generated by the following method: for each sequencing sequence in the first genome fragment data and the second genome fragment data, the arrangement of the first k bases at the 5' end of each sequencing sequence is determined, and the ratios of the occurrence of multiple terminal motifs of the first genome and the second genome are counted, and each of the terminal motifs corresponds to an arrangement of the first k bases at the 5' end of the sequencing sequence, and k is a natural number greater than or equal to 4.

7. The method according to claim 6, characterized in that The method further comprises: performing dimensionality reduction processing on the ratios of occurrence of a plurality of terminal motifs of the first genome and the second genome.

8. The method according to claim 1, characterized in that The classification model is any one of the following: a logistic regression classification model, a random forest classification model, and a support vector machine classification model.

9. A model building device, characterized in that: The method comprises a memory; and a processor connected to the memory, wherein the memory is used to store instructions, and the processor is configured to execute the steps of the model building method according to any one of claims 1 to 8 based on the instructions stored in the memory.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the model building method according to any one of claims 1 to 8 is implemented.

11. A computer program product, characterized in that The computer program product comprises instructions, which, when executed by a computer, execute the model building method according to any one of claims 1 to 8.

12. A model building device, characterized in that: It includes data acquisition module, comparison module, feature generation module and model building module, among which: The data acquisition module is configured to acquire sequencing data of a plurality of training samples, wherein the plurality of training samples at least include a first type of sample and a second type of sample, and the sequencing data includes a plurality of sequencing sequences; The comparison module is configured to compare the sequencing data of each training sample to the first genome and the second genome respectively, to obtain first genome segment data and second genome segment data, wherein the first genome segment data includes a plurality of sequencing sequences compared to the first genome, and the second genome segment data includes a plurality of sequencing sequences compared to the second genome; The feature generation module is configured to generate sample feature data according to the first genome segment data and the second genome segment data; The model building module is configured to use the sample feature data to build a classification model.

Images