WO2022262569A1

WO2022262569A1 - Method for distinguishing somatic mutation and germline mutation

Info

Publication number: WO2022262569A1
Application number: PCT/CN2022/096125
Authority: WO
Inventors: 刘成林; 王俊; 张周; 揣少坤; 汉雨生
Original assignee: 广州燃石医学检验所有限公司
Priority date: 2021-06-18
Filing date: 2022-05-31
Publication date: 2022-12-22
Also published as: CN115497556A

Abstract

A method for distinguishing a somatic mutation and a germline mutation. The method comprises: obtaining at least one mutation site derived from a sample of a subject; obtaining a wild-type support fragment and a mutant-type support fragment, wherein the wild-type support fragment is a cfDNA fragment containing a wild-type base sequence, the mutant-type support fragment is a cfDNA fragment containing a mutant-type base sequence, the wild-type base sequence is the same as a nucleotide sequence at a corresponding position of the mutation site of a human reference genome, and the mutant-type base sequence is different from the nucleotide sequence; obtaining the number of the wild-type support fragments with at least one length, obtaining the number of the corresponding mutant-type support fragments with the same length, and calculating the difference value between the ratio of the number of the wild-type support fragments with the same length to the total number of the corresponding support fragments and the ratio of the number of the mutant-type support fragments with the same length to the total number of the corresponding support fragments; and using the difference value as a distinguishing index. Provided are a method and a device for identifying ctDNA from cfDNA. The method is used for tumor family management and TMB detection.

Description

A method for distinguishing between somatic and germline mutations

technical field

The present application relates to the field of biological information, in particular to a method for distinguishing somatic mutations and germline mutations.

Background technique

In the plasma of tumor patients, cfDNA widely exists, including a small amount of tumor-specific ctDNA. These ctDNAs differ from other normal cfDNAs in the way of shearing during cell senescence and apoptosis. In other words, the fragmentation patterns of ctDNA and other conventional cfDNA in cell-free DNA in plasma are different. Therefore, differences in this distribution pattern can serve as markers for ctDNA recognition.

Somatic mutations are non-genetic variations that are distinct from germline mutations (also known as: germline mutations) that gradually accumulate during the human life cycle. Somatic mutation is an important marker of tumor formation because it is closely related to the molecular signaling pathway of tumorigenesis. Germline mutations are heritable mutations that occur in germ cells, and are of great significance to the study of genetic diseases and genome evolution. In the "Tumor Mutation Burden Detection and Clinical Application Chinese Expert Consensus (2020 Edition)", it is mentioned that in the standardization requirements of the Tumor Mutation Burden (TMB) algorithm, the core element is the detection and calculation of somatic mutations that can affect protein coding . Since the currently public population databases are dominated by European and American populations, they are not suitable for TMB detection in the Chinese population. Therefore, it is recommended that control samples (peripheral blood or paracancerous tissues) should be used to remove germline mutations in patients, or use Construction of a large-sample hereditary mutation database in the Chinese population background library filtering germline variation. Therefore, correctly distinguishing the type and source of mutations in cells plays an important role in the classification, treatment, and prognosis of tumors.

However, the current methods for identifying somatic mutations mainly rely on the detection of paired samples. Parallel sequencing of paired samples can accurately determine the source of mutations. However, it is often very difficult to re-collect paired samples for samples that have not been collected for the first time. In addition, performing high-throughput sequencing at the same depth as tumor samples will consume a lot of funds and computing resources. At the same time, this method has high requirements on the integrity of sample collection and computing and storage resources, and will significantly increase the cost of mutation detection. In addition, the methods of mutation frequency filtering and mutation annotation database comparison still cannot meet the requirements in terms of accuracy.

Contents of the invention

The present application provides a method for distinguishing somatic mutations and germline mutations, a method for identifying ctDNA in cfDNA, and devices and applications corresponding to the methods. The method and/or device described in this application has at least one of the following characteristics: (1) only need to use a single sample, that is, a sample from a subject; (2) has a wide range of applications and can be applied to different cancers Identification of somatic mutations in species, and/or identification of ctDNA; (3) high sensitivity; (4) high accuracy, for example, multiple factors can be combined on the basis of mutation database, population frequency, and mutation abundance at the same time Participate in the method described in this application to improve the reliability of the discrimination results; (5) easy to implement, no limit to the number of mutation sites; (6) fast operation, for example, the plasma of the subject can be used as a sample; (7) A new dimension of distinction is introduced.

In one aspect, the application provides a method for distinguishing between a somatic mutation and a germline mutation, comprising the steps of:

(1) Obtain at least one mutation site derived from a subject sample; wherein, the mutation site is obtained by gene sequencing, (2) for each of the mutation sites, obtain a wild-type supporting fragment and a mutation type support fragment; wherein, the wild-type support fragment is a cfDNA fragment comprising a wild-type base sequence, and the mutant support fragment is a cfDNA fragment comprising a mutant base sequence, wherein the wild-type base sequence is , compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome, the same sequence, wherein the mutant base sequence is, at the corresponding position of the mutation site in the reference genome Compared with the nucleotide sequence of different sequences, wherein the reference genome is the human reference genome in the gene sequencing; (3) for each mutation site, obtain at least one length of the wild-type support fragment Quantity, obtaining the corresponding number of mutant-type supporting fragments of the same length, calculating the ratio WC of the number of wild-type supporting fragments of this length to the total number of wild-type supporting fragments; calculating the mutation of the same length The ratio MC of the quantity of type support fragment and the total quantity of described mutant type support fragment; Calculate the difference value of described ratio WC and described ratio MC under the same length; (4) use described difference as distinguishing described mutation position Points are indicators of somatic or germline mutations.

In one aspect, the application provides a method for identifying ctDNA in cfDNA, comprising the following steps:

(1) Obtain at least one mutation site derived from a subject sample; wherein, the mutation site is obtained by gene sequencing, (2) for each of the mutation sites, obtain a wild-type supporting fragment and a mutation type support fragment; wherein, the wild-type support fragment is a cfDNA fragment comprising a wild-type base sequence, and the mutant support fragment is a cfDNA fragment comprising a mutant base sequence, wherein the wild-type base sequence is , compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome, the same sequence, wherein the mutant base sequence is, at the corresponding position of the mutation site in the reference genome Compared with the nucleotide sequence of different sequences, wherein the reference genome is the human reference genome in the gene sequencing; (3) for each mutation site, obtain at least one length of the wild-type support fragment Quantity, obtaining the corresponding number of mutant-type supporting fragments of the same length, calculating the ratio WC of the number of wild-type supporting fragments of this length to the total number of wild-type supporting fragments; calculating the mutation of the same length The ratio MC of the quantity of type support fragment and the total quantity of described mutant type support fragment; Calculate the difference of described ratio WC and described ratio MC under the same length; (4) use described difference as identification described mutation position Indicator of whether spots are located on ctDNA.

On the one hand, the application provides a kind of training method of machine learning model, it comprises the following steps:

(1) Obtain at least one mutation site derived from a subject sample; wherein, the mutation site is obtained by gene sequencing, (2) for each of the mutation sites, obtain a wild-type supporting fragment and a mutation type support fragment; wherein, the wild-type support fragment is a cfDNA fragment comprising a wild-type base sequence, and the mutant support fragment is a cfDNA fragment comprising a mutant base sequence, wherein the wild-type base sequence is , compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome, the same sequence, wherein the mutant base sequence is, at the corresponding position of the mutation site in the reference genome Compared with the nucleotide sequence of different sequences, wherein the reference genome is the human reference genome in the gene sequencing; (3) for each mutation site, obtain at least one length of the wild-type support fragment Quantity, obtaining the corresponding number of mutant-type supporting fragments of the same length, calculating the ratio WC of the number of wild-type supporting fragments of this length to the total number of wild-type supporting fragments; calculating the mutation of the same length The ratio MC of the quantity of type support fragment and the total quantity of described mutation type support fragment; Calculate the difference value of described ratio WC and described ratio MC under the same length; (4) input described difference value as training index to The machine learning model is used for machine learning training.

On the one hand, the application provides a method for establishing a database, which includes the following steps:

(1) Obtain at least one mutation site derived from a subject sample; wherein, the mutation site is obtained by gene sequencing, (2) for each of the mutation sites, obtain a wild-type supporting fragment and a mutation type support fragment; wherein, the wild-type support fragment is a cfDNA fragment comprising a wild-type base sequence, and the mutant support fragment is a cfDNA fragment comprising a mutant base sequence, wherein the wild-type base sequence is , compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome, the same sequence, wherein the mutant base sequence is, at the corresponding position of the mutation site in the reference genome Compared with the nucleotide sequence of different sequences, wherein the reference genome is the human reference genome in the gene sequencing; (3) for each mutation site, obtain at least one length of the wild-type support fragment Quantity, obtaining the corresponding number of mutant-type supporting fragments of the same length, calculating the ratio WC of the number of wild-type supporting fragments of this length to the total number of wild-type supporting fragments; calculating the mutation of the same length The ratio MC of the quantity of type support fragments and the total quantity of said mutant support fragments; calculate the difference between said ratio WC and said ratio MC under the same length; (4) store said difference in the database, so that Distinguish between somatic and germline mutations, and/or identify ctDNA in cfDNA.

In some embodiments, the gene sequencing includes next generation gene sequencing (NGS).

In certain embodiments, the methods use only samples derived from the subject.

In certain embodiments, the sample comprises a blood sample.

In certain embodiments, the method further comprises the step of obtaining a sample from the subject.

In certain embodiments, the mutation site comprises a single nucleotide variation (SNV).

In certain embodiments, the mutation site comprises more than two nucleotide variations.

In certain embodiments, the wild-type supporting fragment and/or the mutant supporting fragment range in length from about 1 nucleotide to about 550 nucleotides.

In certain embodiments, the wild-type supporting fragment and/or the mutant supporting fragment range in length from about 1 nucleotide to about 400 nucleotides.

In certain embodiments, the wild-type supporting fragment and/or the mutant supporting fragment range in length from about 1 nucleotide to about 200 nucleotides.

In some embodiments, the method includes the following steps: (4') obtaining the distribution of the difference in step (3), selecting the maximum value in the distribution as Dev(Max), and using the Dev( Max) as an indicator of the distinction and/or as the training sample.

In some embodiments, the method includes the following steps: (4') obtaining the distribution of the difference in step (3), which is referred to as the first distribution.

In some embodiments, the method includes the following steps: (5) within the length range of the effective segment interval, each difference in the first distribution is sequentially accumulated to obtain an added value, wherein, The length of the effective fragment interval covers the length of the nucleic acid sequence wound around the nucleosome.

In some embodiments, the nucleic acid sequence is capable of wrapping around a nucleosome for more than 2 weeks, or capable of wrapping around a nucleosome for less than 1 week.

In some embodiments, the length of the effective fragment interval is about 1 to about 167 nucleotides, and/or, about 200 or more nucleotides.

In certain embodiments, the length of the effective fragment interval is about 1 to about 167 nucleotides, and/or, about 250 to about 400 nucleotides.

In some embodiments, the method includes the following steps: (6) Obtaining the second distribution of the added value in step (5), and calculating the maximum value of the added value in the second distribution. In some embodiments, the maximum value of the added value is used as Dev(Max), and the Dev(Max) is used as the indicator of the distinction and/or as the training sample.

In some embodiments, the difference is smoothed, wherein the smoothing includes the following steps:

(a) determine the smoothing window value; wherein the smoothing window value is about an integer in 1-10; (b) determine a number of smoothing sampling length ranges whose length values are equal to the smoothing window value, wherein each smoothing The minimum value of the sampling length range is the initial length, wherein the initial length range is the length range of the wild-type support fragment and/or the mutant support fragment; (c) obtain any smoothed sampling length In the range, the number of wild-type support fragments of at least one smoothed sampling length is obtained to obtain the corresponding number of mutant support fragments of the same length,

calculating the ratio WC of the number of wild-type support fragments of the length to the total number of wild-type support fragments;

calculating the ratio MC of the number of mutant supporting fragments of the same length to the total number of mutant supporting fragments;

calculating the difference between said ratio WC and said ratio MC at the same length; (d) calculating an average difference over a range of smoothed sample lengths based on said difference of said at least one smoothed sample length; (e) The obtained average difference is used as a representative value of the smoothed sampling length range.

In some embodiments, the smoothing window value is an integer between about 2-6.

In some embodiments, the smoothing window value is 3.

In some embodiments, the smoothing process includes the following steps: (f) obtaining the first distribution of the average difference values in step (e).

In some embodiments, the smoothing process includes the following steps: (g) within the length range of the effective segment interval, each average difference in the first distribution is sequentially accumulated to obtain an added value, Wherein, the length of the effective fragment interval is the length of the nucleic acid sequence wound around the nucleosome.

In some embodiments, the smoothing process includes the following steps: (h) obtaining a second distribution of the added value in step (g), and calculating the maximum value of the added value in the second distribution.

In some embodiments, the maximum value of the added value is used as Dev(Max), and the Dev(Max) is used as the indicator of the distinction and/or as the training sample.

In some embodiments, the index also includes one or more parameters selected from the following group: the chromosomal position where the mutation site is located, the base substitution pattern of the mutation site, the mutation site The count value of the nucleic acid fragments of each length in the wild type of the point and/or the count value of the nucleic acid fragments of each length in the mutant type of the mutation site, the allelic variation of the mutation site, the age and The mutation type of the mutation site.

In some embodiments, the index also includes one or more parameters selected from the following group: the chromosome position where the SNV site is located, the base substitution pattern of the SNV site, the SNV site The count value of the nucleic acid fragment of each length in the wild type of the point and/or the count value of the nucleic acid fragment of each length in the mutant type of the SNV site, the allelic variation of the SNV site, the age and The mutation type of the SNV site.

In some embodiments, detecting the mutation site comprises the following steps:

(1) obtaining data from the sample; (2) performing variation identification on the data obtained in step (1); (3) performing variation annotation on the variation identified in step (2); and, (4) performing variation annotation on step (3) ) to filter the annotated variation to obtain the mutation site; optionally, perform quality control on the mutation site.

On the other hand, the present application provides a device for distinguishing somatic mutations and germline mutations, which includes: a calculation module for calculating the difference between the ratio WC and the ratio MC of the same length; wherein, for each mutation site , according to the number of wild-type support fragments of at least one length, and the corresponding number of mutant-type support fragments of the same length; the ratio WC is the number of wild-type support fragments of one length and the number of wild-type support fragments The ratio of the total number; wherein the ratio MC is the ratio of the number of corresponding mutant-type support fragments of the same length to the total number of the mutant-type support fragments; wherein, the wild-type support fragment is a base containing wild-type A cfDNA fragment of a base sequence, the mutant supporting fragment is a cfDNA fragment comprising a mutant base sequence, wherein the wild-type base sequence is the nucleoside corresponding to the position of the mutation site in the reference genome Compared with the acid sequence, the same sequence, wherein the mutant base sequence is, compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome, a different sequence, wherein the reference The genome is the human reference genome in the gene sequencing; the mutation site is derived from the subject sample, wherein the mutation site is obtained by gene sequencing, and the judgment module is used for machine learning training The machine learning model obtains the identification result of identifying the somatic mutation, wherein the machine learning training includes inputting the difference as a training sample into the machine learning model to perform machine learning training.

On the other hand, the present application provides a device for identifying ctDNA in cfDNA, which includes:

A calculation module, configured to calculate the difference between the ratio WC and the ratio MC of the same length; wherein, for each mutation site, according to the number of wild-type support fragments of at least one length, and the corresponding mutant support fragments of the same length Quantity; the ratio WC is the ratio of the number of wild-type support fragments of a length to the total number of wild-type support fragments; wherein the ratio MC is the corresponding number of mutant-type support fragments of the same length Ratio to the total number of mutant support fragments; wherein, the wild-type support fragment is a cfDNA fragment comprising a wild-type base sequence, and the mutant support fragment is a cfDNA fragment comprising a mutant base sequence, wherein , the wild-type base sequence is, compared with the nucleotide sequence of the reference genome at the corresponding position of the mutation site, the same sequence, wherein, the mutant base sequence is, and the reference genome at the Compared with the nucleotide sequence at the corresponding position of the mutation site, a different sequence, wherein the reference genome is the human reference genome in the gene sequencing; the mutation site is derived from a subject sample, Wherein, the mutation site is obtained by gene sequencing, and the judgment module is used to obtain the judgment result of identifying ctDNA in the cfDNA according to the machine learning model that has been trained by machine learning, wherein the machine learning training includes The difference is input to the machine learning model as a training sample for machine learning training.

On the other hand, the application provides a training device for a machine learning model, which includes:

A calculation module, configured to calculate the difference between the ratio WC and the ratio MC of the same length; wherein, for each mutation site, according to the number of wild-type support fragments of at least one length, and the corresponding mutant support fragments of the same length Quantity; the ratio WC is the ratio of the number of wild-type support fragments of a length to the total number of wild-type support fragments; wherein the ratio MC is the corresponding number of mutant-type support fragments of the same length Ratio to the total number of mutant support fragments; wherein, the wild-type support fragment is a cfDNA fragment comprising a wild-type base sequence, and the mutant support fragment is a cfDNA fragment comprising a mutant base sequence, wherein , the wild-type base sequence is, compared with the nucleotide sequence of the reference genome at the corresponding position of the mutation site, the same sequence, wherein, the mutant base sequence is, and the reference genome at the Compared with the nucleotide sequence at the corresponding position of the mutation site, a different sequence, wherein the reference genome is the human reference genome in the gene sequencing; the mutation site is derived from a subject sample, Wherein, the mutation site is obtained by gene sequencing, and the training module is used to input the difference as a training sample into the machine learning model for machine learning training.

In certain embodiments, the device uses only samples derived from the subject.

In some embodiments, the device further includes: an output module, used to display the identification result of the somatic mutation and/or the identification result of ctDNA in the cfDNA generated by the determination module.

In some embodiments, the device further includes a sample obtaining module, configured to obtain the sample from the subject.

In certain embodiments, the sample comprises a blood sample.

In certain embodiments, the sample obtaining module includes reagents and/or instruments for obtaining the sample.

In some embodiments, the device further includes a data receiving module, configured to obtain the mutation site in the sample.

In some embodiments, detecting the mutation site in the device comprises the following steps:

In some embodiments, the data receiving module includes reagents and/or instruments required for the gene sequencing.

In some embodiments, the device further includes an input module, configured to obtain the number of the wild-type supporting fragment of the at least one length, and/or the corresponding number of the mutant supporting fragment of the same length. quantity.

In certain embodiments, the import module is capable of distinguishing between the wild-type supporting fragment and the mutant supporting fragment.

In some embodiments, the input module counts the number of wild-type support fragments of different lengths; and counts the number of wild-type support fragments of different lengths.

In some embodiments, in the calculation module: obtain the distribution of the difference, select the maximum value in the distribution as Dev(Max), use the Dev(Max) as the distinguishing index and /or as the training sample. In some embodiments, the difference is smoothed in the calculation module, wherein the smoothing process includes the following steps: (a) determining a smoothing window value, wherein the smoothing window value is about 1 Integer in -10; (b) determine the smoothing sampling length range of several length values equal to the smoothing window value, wherein the minimum value of each smoothing sampling length range is the starting length, wherein the range of the starting length is the range of the length of the wild-type support fragment and/or the mutant-type support fragment; (c) obtaining the number of the wild-type support fragment of at least one smoothed sampling length in any smoothed sampling length range, obtaining corresponding to the number of mutant support fragments of the same length,

In some embodiments, the smoothing window value is 3.

In some embodiments, the smoothing process includes the following steps: (g) within the length range of the effective segment interval, each average difference in the first distribution is sequentially accumulated to obtain an added value, Wherein, the length of the effective fragment interval covers the length of the nucleic acid sequence wound around the nucleosome.

In some embodiments, the calculation module outputs the Dev(Max).

In some embodiments, the index and/or training samples also include one or more parameters selected from the following group: the chromosomal position where the mutation site is located, the base substitution pattern of the mutation site , the count value of nucleic acid fragments of various lengths in the wild type of the mutation site and/or the count value of nucleic acid fragments of various lengths in the mutant type of the mutation site, the allelic variation of the mutation site, the affected The age of the test subject and the mutation type of the mutation site.

In some embodiments, the index and/or training samples also include one or more parameters selected from the following group: the chromosome position where the SNV site is located, the base substitution pattern of the SNV site , the count value of nucleic acid fragments of various lengths in the wild type of the SNV site and/or the count value of nucleic acid fragments of various lengths in the mutant type of the SNV site, the allelic variation of the SNV site, the affected The age of the test subject and the mutation type of the SNV site.

In another aspect, the present application provides an electronic device, including a memory; and a processor coupled to the memory, the processor configured to execute based on instructions stored in the memory to implement the instructions described in the present application. A method for distinguishing somatic mutations from germline mutations; a method for identifying ctDNA in cfDNA as described herein, or a method for training a machine learning model as described herein.

In another aspect, the present application provides a non-volatile computer-readable storage medium, on which a computer program is stored, and the program is executed by a processor to implement the method for distinguishing somatic mutations and germline mutations described in the present application. ; the method for identifying ctDNA in cfDNA described in the present application, or the training method of the machine learning model described in the present application.

In another aspect, the present application provides a database system, which includes a memory; and a processor coupled to the memory, the processor configured to execute based on instructions stored in the memory to implement the The method for distinguishing somatic mutations and germline mutations as described above; the method for identifying ctDNA in cfDNA as described in this application, or the method for building a database as described in this application.

In another aspect, the present application provides an application of the method for distinguishing somatic mutations and germline mutations described in the present application in the management of tumor families.

In another aspect, the present application provides an application of the method for distinguishing somatic mutations and germline mutations described in the present application in the detection of tumor mutation burden (TMB).

Those skilled in the art can easily perceive other aspects and advantages of the present application from the following detailed description. In the following detailed description, only exemplary embodiments of the present application are shown and described. As those skilled in the art will appreciate, the content of the present application enables those skilled in the art to make changes to the specific embodiments which are disclosed without departing from the spirit and scope of the invention to which this application relates. Correspondingly, the drawings and descriptions in the specification of the present application are only exemplary rather than restrictive.

Description of drawings

The particular features of the invention to which this application relates are set forth in the appended claims. The features and advantages of the invention to which this application relates can be better understood with reference to the exemplary embodiments described in detail hereinafter and the accompanying drawings. A brief description of the accompanying drawings is as follows:

Figure 1 shows the training set used by the method described in this application for the machine learning model, and the verification set required to verify that the method for distinguishing somatic mutations and germline mutations described in this application can distinguish between somatic mutations and germline mutations Case.

Figure 2 shows the machine training results of the machine learning model obtained using the method described in this application.

Figure 3 shows how the machine learning model obtained by the method described in this application distinguishes between somatic mutations and germline mutations in validation set 1.

FIG. 4 shows how the machine learning model obtained by the method described in this application distinguishes between somatic mutations and germline mutations in the verification set 2.

Figure 5 shows that using the method described in this application can distinguish between somatic mutations and germline mutations for different tumor types.

Figure 6 shows the AUC results of the method described in this application for distinguishing somatic mutations from germline mutations.

Figure 7 shows the AUC results of the method described in this application for distinguishing somatic mutations from germline mutations.

Figure 8 shows the distribution of the lengths of the wild-type supporting fragment and the mutant supporting fragment for a mutation site.

Figure 9 shows the distribution of the lengths of the wild-type supporting fragment and the mutant supporting fragment for a mutation site.

Figure 10 shows the distribution of the lengths of the wild-type supporting fragment and the mutant supporting fragment for a mutation site.

detailed description

The implementation of the invention of the present application will be described in the following specific examples, and those skilled in the art can easily understand other advantages and effects of the invention of the present application from the content disclosed in this specification.

Definition of Terms

In this application, the term "somatic mutation" generally refers to an acquired class of mutations that occur in non-embryonic cells. In the present application, the somatic mutation may include a genetic change occurring in a somatic tissue (eg, a cell outside the germline). In this application, the somatic mutations may include point mutations (for example, the exchange of a single nucleotide for another nucleotide (for example, silent mutations, missense mutations, and nonsense mutations)), insertions, and deletions (for example, , addition and/or removal of one or more nucleotides (eg, indels), amplifications, gene duplications, copy number alterations (CNAs), rearrangements, and splice variants. The somatic mutations may be closely related to the processes of cell growth, programming, senescence and apoptosis. For example, the somatic mutations may be associated with alterations in signaling pathways in tumorigenesis, angiogenesis and/or tumor metastasis.

In this application, the term "germline mutation" generally refers to a heritable mutation that occurs in a germ cell (eg, egg or sperm). The germline mutation can be passed on to progeny, eg, can be incorporated into the DNA of every cell (eg, germline and somatic) in the progeny. The germline mutation may be less associated with tumorigenesis. For example, the germline mutation can serve as a "baseline" in TMB analysis.

In this application, the term "gene sequencing" generally refers to the technique used to determine the order of the nucleotide bases adenine, guanine, cytosine and thymine in a DNA molecule. The gene sequencing may include first-generation gene sequencing, second-generation gene sequencing, third-generation gene sequencing or single-molecule sequencing (SMS). Second-generation or next-generation gene sequencing may refer to techniques that use advanced techniques (optical) to detect base position methods while generating many sequences (see, for example, Metzker, 2009 for a review). The term "next-generation sequencing" or "next-generation sequencing" (Next-generation sequencing, NGS) is a high-throughput sequencing technology (High-throughput sequencing), which can parallelize hundreds of thousands to millions of DNA at a time. Molecules are sequenced, generally with short read lengths. According to the development history, influence, sequencing principle and technology, there are mainly the following types: massively parallel signature sequencing (Massively Parallel Signature Sequencing, MPSS), polymerase cloning (Polony Sequencing), 454 pyrosequencing (454pyro sequencing) , Illumina (Solexa) sequencing, ion semiconductor sequencing (Ion semi conductor sequencing), DNA nanoball sequencing (DNA nano-ball sequencing), Complete Genomics' DNA nanoarray and combined probe anchor ligation sequencing, etc. The second-generation gene sequencing can make it possible to analyze the transcriptome and genome of a species in detail, so it is also called deep sequencing (deep sequencing).

In the present application, the term "mutation site" generally refers to the site where there is a nucleotide difference compared with the nucleotide sequence of the control sequence. For example, the control sequence may be a reference sequence used in gene sequencing (for example, it may be a human reference genome). In the present application, the mutation site may include at least one (for example, 1, 2, 3, 4 or more) difference in the nucleotide sequence at the site (for example, the difference Nucleotide substitutions, duplications, deletions and/or additions may be included). For example, the mutation site may include a nucleotide mutation at at least one nucleotide position. The nucleotide mutation can be a natural mutation or an artificial mutation. The mutation site may comprise a single nucleotide variation (SNV).

In the present application, the term "wild-type base sequence" generally refers to the same sequence compared with the nucleotide sequence at the corresponding position of the mutation site in a reference genome (for example, a human reference genome). In some cases, the wild-type base sequence may be the nucleotide sequence at the corresponding position of the mutation site in the human reference genome. In some cases, for a specific mutation site described in this application, the wild-type base sequence may not contain the mutation site.

In the present application, the term "mutant base sequence" generally refers to a sequence that is different from the nucleotide sequence at the corresponding position of the mutation site in a reference genome (for example, it may be a human reference genome). In some cases, for a specific mutation site described in this application, the mutant base sequence may contain the mutation site.

In this application, the term "wild-type supporting fragment" generally refers to a cfDNA fragment comprising the wild-type base sequence described in this application. In this application, for a specific mutation site described in this application, the wild-type supporting fragments may have different sequence lengths. In some cases, for a specific mutation site described in this application, the wild-type supporting fragment may not contain the mutation site. In some cases, for a specific mutation site described in the application, the wild-type support fragment may not contain the mutation site, but for another mutation site described in the application , the wild-type supporting fragment may or may not contain the other mutation site. The term "the length of the wild-type supporting fragment" refers to the length of the wild-type supporting fragment described in this application, and the unit is the number of "nucleotides".

In the present application, the term "mutant supporting fragment" generally refers to a cfDNA fragment comprising the mutated base sequence described in the present application. In some cases, for a specific mutation site described in this application, the mutant support fragment may contain the mutation site. In some cases, for a specific mutation site described in the present application, the mutant support fragment may contain the mutation site, but for another mutation site described in the application, The mutant supporting fragment may or may not contain the other mutation site. The term "length of the mutant supporting fragment" refers to the length of the mutant supporting fragment described in this application, and the unit is the number of "nucleotides".

In this application, the term "human reference genome" generally refers to the human genome that can function as a reference in gene sequencing. The information of the human reference genome can refer to UCSC (http://genome.ucsc.edu/index.html). The human reference genome can have different versions, for example, it can be hg19, GRCH37 or ensembl 75.

In the present application, the term "at the corresponding position" generally refers to the position of at least one specific base in one sequence, and the position of the specific base in the other sequence. For example, the corresponding position can be the nucleotide position at the mutation site in the wild-type base sequence or the mutant base sequence described in the application, and the mutation in the reference genome described in the application location of the site. For example, if the mutation site is the 100th nucleotide in the mutant base sequence, then the corresponding position in the reference genome can be the 100th nucleotide of the corresponding sequence in the reference genome .

In this application, the term "cfDNA" usually refers to the abbreviation of Cell free DNA, and may refer to plasma free DNA. For example, the cfDNA can be an extracellular DNA fragment located in the peripheral circulation.

In this application, the term "ctDNA" generally refers to circulating tumor DNA. ctDNA is a fragment of tumor-derived DNA that is not associated with cells in the blood. The ctDNA can be produced by the entry of genomes in apoptotic or necrotic tumor cells into the blood. The ctDNA may carry specific gene characteristics of primary tumor or metastatic tumor. The ctDNA can be considered as a special kind of the cfDNA.

In this application, the term "machine learning model" generally refers to a system or collection of program instructions and/or data configured to implement an algorithm, process, or mathematical model. In the present application, the algorithm, process or mathematical model can predict and provide a desired output based on a given input. In the present application, the parameters of the machine learning model may not be explicitly programmed, and in the traditional sense, the machine learning model may not be explicitly designed to follow specific rules in order to provide the desired output. For example, the use of the machine learning model may mean that the machine learning model and/or the data structure/set of rules being the machine learning model are trained by a machine learning algorithm.

In this application, the term "database" generally refers to an organized entity of related data, regardless of the manner in which the data or the organized entity is represented. For example, the organized entity of related data may take the form of a table, map, grid, group, datagram, file, document, list, or any other form. In this application, the database may include any data collected and stored in a computer-accessible manner.

In this application, the term "single nucleotide variation (SNV)" generally refers to a variation in a single nucleotide that occurs at a specific location in a genome that is identical to a reference genome (such as the one described in this application). The nucleotides at corresponding positions in the human reference genome differ (for example, substitutions, duplications, deletions, or additions of one nucleotide).

In this application, the term "smoothing" generally refers to a method of data processing that reduces the deviation between one or more of the differences described herein. For example, the smoothing process may include obtaining an average value of a certain number of difference values described in this application. For example, the smoothing process may include selecting different lengths (for example, the smoothing sampling length described in this application) corresponding to a certain interval length (for example, it may be the smoothing window value described in this application). The number of the wild-type support fragment and/or the mutant-type support fragment, calculate the ratio of the two numbers to the total number of the wild-type support ratio and the ratio to the total number of the mutant-type support fragment difference. For example, the smoothing process may include dividing the accumulated value of the difference within a certain length range by the interval length to obtain a ratio. For example, said ratio may be considered as the average difference of said differences over the length range.

In the present application, the term "smoothing window value" generally refers to the interval between the selected wild-type support fragments and/or mutant support fragments of different lengths in the smoothing process described in the present application. Nucleotide length value. For example, in the smoothing process, the length of the selected wild-type support fragment and/or the mutant support fragment can be 1, 4, 7, 10, 13 ... nucleotides in sequence, Then the smoothing window value may be 3. The smoothing window value may be an integer of about 1-30, for example, may be 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10. For example, it can be 1, 2, 3, 4, 5 or 6.

In the present application, the term "smoothed sampling length" generally refers to the length value of the wild-type support fragment selected for counting in the smoothing process described in the present application, and/or, selected for counting Count the length values of the mutant support fragments. For example, the smoothed sampling length may be the length value of each support fragment within the smoothed sampling length range within the length range of the wild-type support fragment and/or the mutant support fragment described in the present application. For example, in each smoothing sampling length range, it can be from the initial length (for example, it can start from a length of 1 nucleotide) to the maximum value of the smoothing sampling length range (for example, it can be the initial length+ (smoothing window value - 1)), where each supports the length value of the segment. For example, if the smoothing window value can be 3, if the initial length is 1 nucleotide, then the smoothing sampling length range can be 1-3, 4-6, 7-9...; For example, if the smoothing window value can be 3, if the initial length is 1 nucleotide, then the smoothing sampling length range can also be 1-3, 2-4, 3-5... . In the present application, the initial length may also be other lengths than 1 (for example, it may start from a length of 2 nucleotides). For example, if the initial length is 2 nucleotides, the smoothing sampling length range can be 2-4, 5-7, 8-10...; for example, if the smoothing window value can be 3. If the initial length is 2 nucleotides, the smoothed sampling length range may also be 2-4, 3-5, 4-6....

In the present application, the term "first distribution" generally refers to the distribution of the average difference of each smoothed sampling length range described in the present application. In some cases, the first distribution may be a collection of average differences described in the present application.

In the present application, the term "the length of a nucleic acid sequence that winds around a nucleosome" generally refers to the length required for a nucleic acid sequence to wind around a nucleosome. For example, the nucleic acid sequence may wrap around the nucleosome at a certain multiple (eg, may wrap within one time, or may wrap more than twice).

In the present application, the term "the length of the effective fragment interval" generally refers to the range of the length corresponding to the wild-type supporting fragment and/or the mutant supporting fragment required for calculating the addition value described in the present application.

In the present application, the term "second distribution" generally refers to the distribution of addition values described in the present application. In some cases, the second distribution may be a collection of the added values described in this application.

In this application, the term "calculation module" generally refers to a functional module for calculating the difference between the number of wild-type support fragments described in this application and the number of mutant support fragments described in this application with the same length. The calculation module can input the number of wild-type supporting fragments described in the present application, and correspondingly the number of mutant-type supporting fragments of the same length. The calculation module can output the difference value described in this application. For example, Dev(Max) described in this application may be output. In the calculation module, the smoothing process described in this application can be performed.

In this application, the term "judgment module" generally refers to a machine learning model that has been trained by machine learning to obtain relevant judgment results (for example, the judgment results may include the recognition results of somatic mutations described in this application, And/or the judgment result of recognizing ctDNA in the cfDNA described in this application). In this application, the judging module may input the difference described in this application (for example, the Dev(Max)). The judging module can output the related judging result. In the judging module, the machine learning model can be used for judging.

In the present application, the term "training module" generally refers to a functional module for inputting the difference described in the present application (such as the Dev(Max)) as a training sample into the machine learning model for machine learning training . The "machine learning" may refer to artificial intelligence systems configured to learn from data without being explicitly programmed. The "machine learning model" can be a collection of parameters and functions that can train parameters on a set of training samples. Parameters and functions can be collections of linear algebraic, nonlinear algebraic, and tensor algebraic operations. Parameters and functions can contain statistical functions, tests, and probability models. In the present application, the training module may input the difference value described in the present application (for example, the Dev(Max)). The training module can output a machine learning model that has been trained by machine learning.

In the present application, the term "output module" generally refers to a functional module for displaying the recognition result of the somatic mutation generated by the judgment module of the present application and/or the judgment result of recognition of ctDNA in the cfDNA. For example, the output module may include a display, which may display (for example, in the form of graphs and/or text) the recognition result of the somatic mutation generated by the judgment module of the present application and/or the recognition of ctDNA in the cfDNA judgment result.

In this application, the term "sample obtaining module" generally refers to a functional module for obtaining said sample of a subject. For example, the sample obtaining module may include reagents and/or instruments needed to obtain the sample (eg, blood sample). For example, lancets, blood collection tubes, and/or blood sample transport boxes may be included. The sample acquisition module can output the samples described herein.

In this application, the term "data receiving module" generally refers to a functional module for obtaining the mutation site in the sample. In this application, the data receiving module may input the sample described in this application (such as a blood sample). The data receiving module can output the mutation site. The data receiving module can detect the mutation site of the sample. For example, the data receiving module may perform the gene sequencing described in this application (eg, next-generation gene sequencing) on the sample. For example, the data receiving module may include reagents and/or instruments required for the gene sequencing. The data receiving module can detect the single nucleotide variation.

In the present application, the term "input module" generally refers to the number of wild-type support fragments used to obtain the at least one length, and/or the corresponding number of mutant support fragments of the same length. functional module. In this application, the input module can input the mutation site described in this application. The input module may output (eg, may display) the number of the wild-type supporting fragments of the at least one length, and/or the corresponding number of the mutant supporting fragments of the same length. Said input module may comprise reagents and/or instruments capable of enumerating said wild-type support fragments of a specified length. The input module may comprise reagents and/or instruments capable of counting the mutant supporting fragments of a specified length. The input module can identify the lengths of the wild-type supporting fragments and count them respectively; the input module can identify the lengths of the mutant-type supporting fragments and count them respectively. The input module can determine whether the length of the wild-type supporting fragment is the same as that of the mutant-type fragment.

In this application, the term "cancer family management" generally refers to providing help for cancer-related matters to patients with family hereditary tumors, their relatives and/or high-risk groups. For example, the cancer family management may include providing genetic counseling for the above-mentioned population, performing detection of tumor-related genes and interpretation of results, risk assessment of cancer, consultation and/or implementation of preventive interventions.

In this application, the term "tumor mutation burden (TMB)" refers to tumor mutation burden. According to the definition in "Tumor mutation burden detection and clinical application Chinese expert consensus (2020 edition)", TMB generally refers to the tumor mutation burden per megabase in a specific genomic region. The number of somatic non-synonymous mutations in base pairs (Mb), usually expressed in terms of mutations per megabase (XX mutations/Mb). The TMB can be used as a biomarker related to the response to immunotherapy. The TMB can indirectly reflect the ability and degree of neoantigen production by the tumor, and has been proven to predict the response to immunotherapy. For example, the first edition of the NSCLC guidelines in 2019 pointed out that TMB is used to identify candidates for "Nivolumab+Ipilimumab" immune combination therapy and "Nivolumab" monotherapy in patients with lung cancer. The expression level of TMB may be related to many factors, such as microsatellite instability (MSI-H) and the existence of certain driver genes.

In this application, the term "comprising" generally means including specifically specified features, but not excluding other elements.

In this application, the term "about" generally refers to a range of 0.5%-10% above or below the specified value, such as 0.5%, 1%, 1.5%, 2%, 2.5%, above or below the specified value. 3%, 3.5%, 4%, 4.5%, 5%, 5.5%, 6%, 6.5%, 7%, 7.5%, 8%, 8.5%, 9%, 9.5%, or 10%.

Detailed description of the invention

method

In the present application, the gene sequencing may include next generation gene sequencing (NGS). In the present application, the NGS can be selected from the following group: Solexa sequencing technology, 454 sequencing technology, SOLiD sequencing technology, Complete Genomics sequencing method and semiconductor (Ion Torrent) sequencing technology. The gene sequencing can be high-throughput, for example, hundreds of thousands or millions of DNA molecules can be sequenced at one time. The gene sequencing can be short-segment, for example, the read length of NGS can be no more than 500bp.

In the present application, the gene sequencing may include the following steps: (1) library construction; for example, it may include modifying the ends of the DNA molecules, and adding adapters (for example, Y-shaped adapters may be formed), and then perform PCR Amplification; (2) Sequencing; for example, it may include DNA replication using oligonucleotides as primers and library fragments as templates; then "bridge" amplification, and sequencing while synthesizing. Then add the sequencing primer Index primer to read the Index sequence in the linker to determine which library the DNA at each site belongs to.

In the present application, the method may only use samples derived from a subject. In the present application, the method may not require the use of paired samples. Therefore the method described in this application can greatly reduce the requirement of the subject's sample.

In the present application, the sample may include a blood sample.

In the present application, the method may further include the following step: obtaining a sample from a subject. For example, the step of obtaining a blood sample from said subject using a lancet system may be included. The method for obtaining samples may include a vacuum blood collection tube blood collection method.

In the present application, the mutation site may include a single nucleotide variation (SNV). In the present application, the mutation site may contain more than two nucleotide variations. For example, the mutation site described in the present application may include one of the SNVs, or two or more (for example, it may be 2, 3, 4, 5, 6, 7, 8, 9, 10 or more) SNVs (eg, may include more than two nucleotide variations). In the present application, for a specific mutation site, the nucleotide sequence at the position of the mutation site is different between the wild-type support fragment and the mutant support fragment. The mutation sites may include nucleotide substitutions, and may also include nucleotide deletions and/or insertions in some cases. In the present application, the mutation site may include nucleotide substitutions.

In the present application, the division of the wild type supporting fragment and/or the mutant supporting fragment may be specific to a specific mutation site. For example, if the nucleotide sequence at the mutation site is the same as the nucleotide sequence at the corresponding position of the mutation site in the reference genome, it can be considered as the reference genome for the mutation site. Wild-type supporting fragment; if the nucleotide sequence at the mutation site is different from the nucleotide sequence at the corresponding position of the reference genome at the mutation site, it can be considered when targeting the mutation site is the mutant support fragment.

In the present application, the wild-type support fragment and/or the mutant support fragment can range in length from about 1 nucleotide to about 550 nucleotides (for example, it can be about 1-500, about 1-450, about 1-400, about 1-350, about 1-300, about 1-250, about 1-200, or about 1-100). For example, it can be from about 1 nucleotide to about 400 nucleotides. For example, it can be from about 1 nucleotide to about 200 nucleotides.

In the present application, the method may include the following steps: (4') obtain the distribution of the difference in step (3), select the maximum value in the distribution as Dev(Max), and use the Dev(Max ) as the distinguishing index and/or as the training sample.

In this application, the distribution may be a collection of the differences. The Dev(Max) may be the maximum value of the difference in the set.

In this application, the difference may be smoothed. In the present application, after the smoothing process, the difference can more intuitively and accurately reflect the difference between the number of wild-type supporting fragments and the mutant-type supporting fragments of the same length. Further, the smoothed difference can more accurately, specifically and/or more sensitively distinguish the somatic mutation from the systemic mutation, and/or identify cfDNA in the ctDNA.

In the present application, the smoothing process may include the following steps:

(a) determining the smoothing window value; wherein the smoothing window value is an integer of about 1-10; (b) determining the smoothing sampling length range, wherein the minimum value of each smoothing sampling length range is the initial length , where the maximum value of each smoothing sampling length range is the initial length+(smoothing window value-1), that is: the length value of each smoothing sampling length range is equal to the determined smoothing window value; wherein the initial length The range is the range of the length of the wild-type support fragment and/or the mutant-type support fragment; (c) obtain at least one smoothed sampling length of the wild-type support fragment in any smoothed sampling length range Quantity, to obtain the corresponding quantity of the mutant support fragments of the same length,

Calculate the ratio WC of the number of wild-type support fragments of the length to the total number of wild-type support fragments; calculate the ratio of the number of mutant support fragments of the same length to the total number of mutant support fragments MC;

Calculate the difference between the ratio WC and the ratio MC under the same length; (d) accumulate each of the differences obtained in step (c), divide by the smoothing window value, and obtain the smoothing sampling length (e) using the obtained average difference as a representative value of the smoothed sampling length range.

In this application, the smoothing window value can be adjusted according to different subjects, different gene sequencing methods and/or different differentiation purposes, as long as the selected smoothing window value can make the smoothing window processing can be carried out. In the present application, the smoothing window value may be an integer between about 2-6 (for example, the smoothing window value may be 2, 3, 4, 5 or 6). For example, the smoothing window value may be 3.

In the present application, the smoothing process may include the following specific steps:

(a) determining a smoothing window value; wherein the smoothing window value is an integer between about 1-10 (for example, selecting the smoothing window value to be 3);

(b) Determine the smoothing sampling length range, wherein the minimum value of each smoothing sampling length range is the starting length, and wherein the maximum value of each smoothing sampling length range is the starting length+(smoothing window value-1) ; wherein the range of the starting length is the range of the length of the wild-type support fragment and/or the mutant support fragment (for example, may be about 1 nucleotide to about 400 nucleotides); in In the present application, the initial length may be any length within the range of the length of the wild-type supporting fragment and/or the mutant-type supporting fragment. In this application, the "length" can be measured by the number of nucleotides.

In this application, for each of the smoothing sampling length ranges, the minimum value in each smoothing sampling length range may be a starting length as the first item, and the smoothing The window value is a tolerance, the first item, the second item, the third item up to the Nth item in the arithmetic sequence within the range of the length of the wild-type supporting fragment and/or the mutant-type supporting fragment. For example, when the smoothing window value is 3 and the initial length is 1, then in the range of about 1 nucleotide to about 400 nucleotides, the smoothing minimum value may be 1, 4, 7, 10... 400.

For example, when the initial length is 1, if the smoothing window value is 3, and if the smoothing sampling length ranges do not overlap each other, the smoothing sampling length ranges can be 1-3, 4-6, 7 -10……. For example, when the initial length is 1, if the smoothing window value is 3, and if each of the smoothing sampling length ranges can overlap each other, then the smoothing sampling length ranges can be 1-3, 2-4, 3 -5..., or 1-3, 3-5, 5-7.... For another example, when the initial length is 2, if the smoothing window value is 3, and if the smoothing sampling length ranges do not overlap each other, the smoothing sampling length ranges can be 2-4, 5-7, 8-11.... For example, when the initial length is 2, if the smoothing window value is 3, and if each of the smoothing sampling length ranges can overlap each other, then the smoothing sampling length ranges can be 2-4, 3-5, 4 -6…….

(c) Obtain the number of wild-type support fragments of at least one (for example, at least 1, at least 2, at least 3 or more) smoothed sampling lengths in any smoothed sampling length range, and obtain the corresponding The number of the mutant support fragments of the same length, calculate the ratio WC of the number of the wild type support fragments of this length to the total number of the wild type support fragments; calculate the number of the mutant support fragments of the same length Ratio MC to the total number of mutant supporting fragments; calculate the difference between the ratio WC and the ratio MC at the same length.

For example, the number of wild-type supporting fragments with a length of 1 nucleotide is obtained, and the number is divided by the total number of wild-type supporting fragments W _total to obtain a ratio WC1; The number of the mutant supporting fragments is divided by the total number M _total of the mutant supporting fragments to obtain the ratio MC1, and the difference WC1-MC1 between the two is calculated; for example, the length of the acquisition is 4 nucleotides The number of the wild-type supporting fragments is divided by the total number W _total of the wild-type supporting fragments to obtain a ratio WC4; the number MC4 of the mutant supporting fragments with a length of 4 nucleotides is obtained, Divide this number by the total number M _total of the mutant support fragments, and calculate the difference WC4-MC4 between the two; thus obtain different smoothing sampling lengths (for example, 1, 4, 7, 10...400 ) respectively corresponding to the difference of the ratio; for example, within the range of each smoothed sampling length, under each smoothed sampling length, the number of wild-type support fragments and the total number of wild-type support fragments can be obtained The difference between the ratio of and the ratio of the number of mutant supporting fragments to the total number of mutant supporting fragments. For example, for the smoothing sampling length range 1-3, (WC1-MC1), (WC2-MC2) and (WC3-MC3) can be obtained respectively.

(d) calculating an average difference value over a range of smoothed sample lengths based on said difference values of said at least one smoothed sample length; and are divided by the smoothing window value to obtain the average difference. Optionally, it is also possible to calculate only partial differences in a single smoothing sampling length range, for example: (WC1-MC1) and (WC3-MC3), and then calculate their average value as the average difference;

(e) Taking the obtained average difference as a representative value of the average difference in the smoothed sampling length range. For example, by dividing the cumulative value of (WC1-MC1), (WC2-MC2) and (WC3-MC3) by the smoothing window value 3, the obtained average difference B1 can be used as the representative of the smoothing sampling length range value. For example, the accumulated value of (WC4-MC4), (WC5-MC5) and (WC6-MC6) is divided by the smoothing window value 3, and the obtained average difference B4 can be used as the representative of the smoothing sampling length range value.

In the present application, the smoothing process may include the following steps: (f) obtaining the first distribution of the average difference in step (e). For example, the respective accumulated values B1, B4, B7, etc. are formed into the first distribution D=[B1, B4, B7...B400].

In the present application, the smoothing process may further include the following steps: (g) within the length range of the effective segment interval, sequentially accumulating each average difference in the first distribution to obtain an added value, Wherein, the length of the effective fragment interval covers the length of the nucleic acid sequence wound around the nucleosome.

In the present application, the nucleic acid sequence can wrap around the nucleosome for more than 2 weeks, or can wrap around the nucleosome within 1 week. For example, the length of the effective fragment interval can be about 1 to about 180 nucleotides (for example, it can be about 1 to about 180, about 1 to about 179, about 1 to about 178, about 1 to about 177, about 1 to about 176, about 1 to about 175, about 1 to about 174, about 1 to about 173, about 1 to about 172, about 1 to about 171, about 1 to about 170 about 1-about 169, about 1-about 168, about 1-about 167, about 1-about 166 or about 1-about 165), and/or, about 200 or more nucleotides (For example, it can be about 200 or more, about 210 or more, about 220 or more, about 230 or more, about 240 or more, about 250 or more, about 260 or more, about 270 or more, about 280 or more, About 290 or more, about 300 or more, about 350 or more, or about 400 or more). For example, the length of the effective fragment interval may be about 1 to about 167 nucleotides, and/or, about 250 to about 400 nucleotides.

For example, B1 and B4 in the first distribution may be added up to obtain the added value D1; B1, B4 and B7 in the first distribution may be added up to obtain the added value D2.

In the present application, the smoothing process may include the following steps: (h) obtaining the second distribution of the added value in step (g), and calculating the maximum value of the added value in the second distribution. For example, each of the added values D1, D2, etc. can be formed into the second distribution A=[D1, D2...Di]. Wherein, i may be the length of the effective segment interval.

In this application, the maximum value in the second distribution may be taken as Dev(Max). In the present application, the Dev(Max) may be used as the distinguishing index and/or as the training sample.

In the present application, in order to further improve the accuracy, sensitivity and/or specificity of the method described in the application, other parameters can also be used as The distinguishing index and/or as the training sample. For example, the indicator may also include one or more parameters selected from the following group: the chromosomal position where the mutation site is located, the base substitution pattern of the mutation site, the wild type of the mutation site The count value of nucleic acid fragments of various lengths in and/or the count value of nucleic acid fragments of various lengths in the mutant type of the mutation site, the allelic variation of the mutation site, the age of the subject and the mutation site The point mutation type.

In this application, the indicator may also include one or more parameters selected from the following group: the chromosome position where the SNV site is located, the base substitution pattern of the SNV site, the SNV site The count value of nucleic acid fragments of various lengths in the wild type and/or the count value of nucleic acid fragments of various lengths in the mutant type of the SNV site, the allelic variation of the SNV site, the age of the subject and the The mutation type of the SNV locus.

In the present application, the method may further include the step of detecting the mutation site. The step of detecting the mutation site can be a routine step in the art. With reference to the gene sequencing, for example, detecting the mutation site can include the following steps: (1) obtaining data from the sample; The data obtained in step (1) is subjected to variation identification (for example, the variation identification can be carried out by base quality, mapping quality, number of mismatches, mutation frequency, reads supporting mutation and other factors); (3) step ( 2) The identified variants are annotated for variants (for example, ANNOVAR 20160201, 1000 Genomes database, ExAC database and/or gnomAD genome database can be used for annotation; for example, database annotation, hot site annotation, mutation type and/or population can be used frequency annotation); and, (4) filter the variation annotated in step (3) (for example, filtering of population mutation site frequency, filtering of hot spot mutation, filtering of clonal hematopoietic mutation, and/or maximum depth filter) to obtain the mutation site. For example, the step may also include performing quality control on the mutation site after step (4) (for example, the quality control may include removing repeated fragments, and/or filtering low-quality fragments).

device

In another aspect, the present application provides a device for distinguishing between somatic mutations and germline mutations, comprising:

A calculation module, configured to calculate the difference between the ratio WC and the ratio MC of the same length; wherein, for each mutation site, according to the number of wild-type support fragments of at least one length, and the corresponding mutant support fragments of the same length Quantity; the ratio WC is the ratio of the number of wild-type support fragments of a length to the total number of wild-type support fragments; wherein the ratio MC is the corresponding number of mutant-type support fragments of the same length Ratio to the total number of mutant support fragments; wherein, the wild-type support fragment is a cfDNA fragment comprising a wild-type base sequence, and the mutant support fragment is a cfDNA fragment comprising a mutant base sequence, wherein , the wild-type base sequence is, compared with the nucleotide sequence of the reference genome at the corresponding position of the mutation site, the same sequence, wherein, the mutant base sequence is, and the reference genome at the Compared with the nucleotide sequence at the corresponding position of the mutation site, a different sequence, wherein the reference genome is the human reference genome in the gene sequencing; the mutation site is derived from a subject sample, Wherein, the mutation site is obtained by gene sequencing,

A judging module, configured to obtain a recognition result for identifying the somatic mutation according to a machine learning model that has been trained by machine learning, wherein the machine learning training includes inputting the difference as a training sample into the machine learning model to Do machine learning training.

A judgment module, configured to obtain a judgment result of identifying ctDNA in the cfDNA according to a machine learning model that has been trained by machine learning, wherein the machine learning training includes inputting the difference as a training sample into the machine learning model for machine learning training.

A training module, configured to input the difference as a training sample into the machine learning model for machine learning training.

In the present application, the device may only use samples derived from the subject.

In the present application, the device may further include: an output module, configured to display the recognition result of the somatic mutation generated by the judgment module and/or the judgment result of recognition of ctDNA in the cfDNA.

In the present application, the output module may display the recognition result of the somatic mutation generated by the judgment module of the present application and/or the judgment result of recognition of ctDNA in the cfDNA. For example, the output module may include an output device (such as a display) and/or an output program (such as a mobile APP), so as to display the recognition result of the somatic mutation generated by the judgment module of the present application and/or the The judgment result of ctDNA recognition in cfDNA. In the present application, the output module inputs the identification result of the somatic mutation and/or the identification result of ctDNA in the cfDNA obtained by the determination module.

In the present application, the device may further include a sample obtaining module, configured to obtain the sample of the subject.

For example, the sample may comprise a blood sample. In the present application, the sample obtaining module may include reagents and/or instruments required for obtaining the sample. For example, the sample acquisition module may include blood collection needles, blood collection tubes and/or blood sample transport boxes. For example, the sample acquisition module can include an anticoagulant. In this application, the sample obtaining module can output the samples described in this application.

In the present application, the device may further include a data receiving module, configured to obtain the mutation site in the sample. For example, the data receiving module may input the samples. For example, the data receiving module can output the mutation sites described in this application. In the present application, the data receiving module may include reagents and/or instruments required for obtaining the mutation site. For example, the data receiving module may include reagents and/or instruments required for the gene sequencing. In the present application, the data receiving module may perform the gene sequencing described in the present application, for example, the gene sequencing may include next-generation gene sequencing (NGS).

For example, the data receiving module may include a next-generation gene sequencer (eg, Roche454 sequencer, Illumina sequencer). For example, the data receiving module can include an automated sample preparation system. For example, the data receiving module may include fluorescently labeled dNTP, end repair enzyme, end repair reaction buffer, DNA ligase, DNA ligation buffer and/or library amplification reaction solution.

In the present application, the mutation site may include a single nucleotide variation (SNV). In the present application, the mutation site may contain more than two nucleotide variations.

In the present application, detecting the mutation site in the device may include the following steps: (1) obtaining data from the sample; (2) performing mutation identification on the data obtained in step (1); (3) Annotate the variation identified in step (2); and, (4) filter the variation annotated in step (3) to obtain a mutation site; optionally, perform quality control on the mutation site.

In the present application, the device may further include an input module, configured to obtain the quantity of the wild-type support fragment of the at least one length, and/or the corresponding quantity of the mutant support fragment of the same length .

For example, the input module can input the mutation sites described in this application. The input module may output the number of the wild-type supporting fragments of the at least one length, and/or the corresponding number of the mutant supporting fragments of the same length. In the present application, the input module may include reagents and/or instruments capable of counting the wild-type support fragments of a specific length. The input module may comprise reagents and/or instruments capable of counting the mutant supporting fragments of a specified length. In the present application, the input module may include an instrument capable of displaying the quantity of the wild-type support fragment of the at least one length, and/or the quantity of the corresponding mutant support fragment of the same length ( Such as a display) and/or an output program (such as a mobile APP), so that the number of wild-type and/or mutant support fragments obtained by using the input module can be displayed. In the present application, the input module can distinguish between the wild type support fragment and the mutant support fragment. In this application, the input module can count the number of wild-type supporting fragments of different lengths; and, count the number of wild-type supporting fragments of different lengths.

In the present application, the length of the wild-type supporting fragment and/or the mutant supporting fragment may range from about 1 nucleotide to about 550 nucleotides. For example, it can be from about 1 nucleotide to about 400 nucleotides. For example, it can be from about 1 nucleotide to about 200 nucleotides.

In the present application, the calculation module can input (for example, can be obtained through the input module of the present application) the number of wild-type support fragments described in the present application, and the corresponding number of mutant support fragments of the same length . The calculation module may output the difference value described in this application, for example, the calculation module may output Dev(Max) described in this application. The calculation module may include calculation logic and/or a calculation program for calculating the difference value described in this application.

In the present application, the distribution of the difference can be obtained in the calculation module, the maximum value in the distribution is selected as Dev(Max), and the Dev(Max) is used as the indicator of the distinction and/or as the training sample.

In the present application, the difference value may be smoothed in the calculation module, wherein the smoothing process may include the following steps: (a) determining a smoothing window value, wherein the smoothing window value is about 1 Integers in -30; (b) determine several smoothing sampling length ranges whose length values are equal to the smoothing window value, wherein the minimum value of each smoothing sampling length range is the starting length, wherein the range of the starting length is the range of the length of the wild-type support fragment and/or the mutant-type support fragment; (c) obtaining the number of the wild-type support fragment of at least one smoothed sampling length in any smoothed sampling length range, Obtain the corresponding number of mutant support fragments of the same length, calculate the ratio WC of the number of wild-type support fragments of this length to the total number of wild-type support fragments; calculate the mutant support fragments of the same length the ratio MC of the number of fragments to the total number of mutant support fragments; calculating the difference between said ratio WC and said ratio MC at the same length; (d) said difference according to said at least one smoothed sampling length Calculate the average difference of the smoothed sampling length range; (e) use the average difference obtained in step (d) as the representative value of the smoothed sampling length range.

In the present application, the smoothing window value may be an integer between about 2-6. For example, the smoothing window value may be 3.

In the present application, the smoothing process may include the following steps: (f) obtaining the first distribution of the average difference in step (e).

In the present application, the smoothing process may include the following steps: (g) within the length range of the effective segment interval, each average difference in the first distribution is sequentially accumulated to obtain an added value, wherein , the length of the effective fragment interval covers the length of the nucleic acid sequence wound around the nucleosome.

In the present application, the nucleic acid sequence may be capable of wrapping around nucleosomes for more than 2 weeks, or capable of wrapping around nucleosomes for less than 1 week. In the present application, the length of the effective fragment interval may be about 1 to about 167 nucleotides, and/or, about 200 or more nucleotides. In the present application, the length of the effective fragment interval may be about 1 to about 167 nucleotides, and/or, about 250 to about 400 nucleotides.

In the present application, the smoothing process may include the following steps: (h) obtaining the second distribution of the added value in step (g), and calculating the maximum value of the added value in the second distribution. In the present application, the maximum value of the added value may be used as Dev(Max), and the Dev(Max) may be used as the distinguishing index and/or as the training sample.

In this application, the judgment module can obtain relevant judgment results according to the machine learning model that has been trained by machine learning (for example, the judgment results can include the recognition results of somatic mutations described in this application, and/or the results of this application The judgment result of recognizing ctDNA in the cfDNA). In this application, the judging module may input the difference described in this application (for example, the Dev(Max)). The judging module can output the related judging result. In this application, the judging module may include a machine learning model that has been trained by machine learning. Wherein, the machine learning model is obtained by using the verification set and the difference described in the present application (for example, using the parameters described in the present application), and using the training method of the machine learning model described in the present application.

In this application, the index and/or training samples may also include one or more of the following parameters: the chromosome position where the mutation site is located, the base substitution pattern of the mutation site, the mutation The count value of nucleic acid fragments of various lengths in the wild type of the site and/or the count value of nucleic acid fragments of various lengths in the mutant type of the mutation site, the allelic variation of the mutation site, the age of the subject and the mutation type of the mutation site.

In this application, the index and/or training samples may also include one or more of the following parameters: the chromosome position where the SNV site is located, the base substitution pattern of the SNV site, the SNV The count value of nucleic acid fragments of various lengths in the wild type of the site and/or the count value of nucleic acid fragments of various lengths in the mutant type of the SNV site, the allelic variation of the SNV site, the age of the subject and the mutation type of the SNV site.

In the present application, the device may include the calculating module and the judging module. The apparatus may include the computing module and the training module.

In the present application, the device may include the sample acquisition module, the data receiving module, the input module, the calculation module, the judgment module and the output module. In this application, the sample, and the information and/or calculation results derived from the sample can be obtained from the sample acquisition module, the data receiving module, the input module, the calculation module, the judgment The order of the modules and the output modules are transferred sequentially.

For example, the non-transitory computer readable storage medium may include a floppy disk, a flexible disk, a hard disk, a solid state storage (SSS) (such as a solid state drive (SSD)), a solid state card (SSC), a solid state module (SSM)), an enterprise high-grade flash drives, tape, or any other non-transitory magnetic media, etc. Non-transitory computer readable storage media may also include punched cards, paper tape, cursor sheets (or any other physical media having a pattern of holes or other optically identifiable markings), compact disc read only memory (CD-ROM) , Rewritable Disc (CD-RW), Digital Versatile Disc (DVD), Blu-ray Disc (BD) and/or any other non-transitory optical media.

For example, the database system may implement various mechanisms to ensure that the methods described herein performed on the database system produce correct results. In this application, the database system may use disks as permanent data storage. In this application, the database system can provide database storage and processing services for multiple database clients. The database client may store database data across multiple shared storage devices, and/or may utilize one or more execution platforms with multiple execution nodes. The database system can be organized such that storage and computing resources can be effectively scaled indefinitely.

application

In the present application, the method can be used to determine whether the subject has a germline mutation. Subjects carrying certain germline mutations may have a higher lifetime risk of developing cancer (eg, colorectal, endometrial, gastric, and/or ovarian cancer) than the general population. Therefore, the method can be used to screen out subjects with higher risk. The subject can receive individualized tumor monitoring, so as to achieve the purpose of early diagnosis and early treatment.

In this application, the method can be used to detect the TMB and can be used in clinical practice (for example, it can be speculated whether certain specific tumor treatment methods are suitable for the subject). In some cases, the TMB level detected by the method can be used in clinical practice in combination with other biomarkers such as immune checkpoints and T cell inflammatory markers.

Not intending to be limited by any theory, the following examples are only for explaining the fusion protein, preparation method and application of the application, and are not intended to limit the scope of the invention of the application.

Example

Example 1 Obtain the mutation site described in the application

1. Data preparation

a) Sequence reply: Use the mem module in the bwa 0.7.10 software to map the sequence to the human reference genome GRCh37/hg19 to form a .bam file of the alignment result.

2. Mutation Identification

Use vardict 1.5.1 to perform mutant calling (variant calling) on SNV, and the calling parameters are as follows:

a) removing bases with base quality (base quality)<30;

b) Remove reads with low mapping quality, such as <60 reads;

c) remove reads with too many mismatches, for example: more than 12, 10, 8 or 6 mismatches;

d) The mutation frequency should not be too small, for example: mutation frequency >=0.002, 0.001, 0.0005, 0.0002 or 0.0001;

e) reads (reads) >= 3, 2 or 1 supporting the mutation;

3. Variation Annotation

These include database annotations, hot spot mutation (hot) site annotations, mutation types, and population frequency annotations.

a) Use ANNOVAR 20160201 to annotate the variable sites;

b) Annotate the hotspot mutation (hot) site: if a mutation is in the hotspot mutation list, the mutation is a hotspot mutation, and in the subsequent mutation filtering, the hotspot mutation is not included in the prediction of the model;

c) Use SnpEff V4.3 to annotate the mutation type;

d) Annotation of population frequency: Given a mutation site, take the maximum value of the population frequency in various databases as the population frequency of the mutation site.

The databases used include but are not limited to: 1000Genomes database, ExAC database and ESP6500 database, etc. .

4. SNV mutation filtering

Annotate all the annotated mutation sites according to the following conditions:

a) Filtering of population mutation frequency: after filtering, retain mutations whose population mutation frequency is less than a specific value, for example: less than or equal to 0.005, 0.002 or 0.001;

b) filtering of hotspot mutations;

c) Clonal hematopoietic mutation filtering;

d) Maximum depth filtering: filter mutations greater than a specific sequencing depth, for example: sequencing depth greater than 20,000, etc.;

5. Quality control of SNV mutation site fragments

a) Repeat sequence removal: remove the repeat sequence generated during PCR amplification;

b) Filter low-quality fragments: filter fragments whose base quality median is less than Q20;

c) Filter fragments with sequencing errors: filter fragments that cannot be compared with the reference genome;

d) Mutation removal at low coverage depth: remove SNVs with less than 50 supporting fragments.

Embodiment 2 The method for obtaining the difference described in this application

2.1

According to the mutation site SNV obtained in Example 1, the difference value described in the application is calculated according to the following steps:

a) acquisition of wild-type support fragment and mutant support fragment: wherein, the wild-type support fragment is a cfDNA fragment comprising a wild-type base sequence, and the mutant support fragment is a cfDNA fragment comprising a mutant base sequence, Wherein, the wild-type base sequence is the same sequence as the nucleotide sequence at the corresponding position of the mutation site in the reference genome, wherein the mutant base sequence is the same sequence as the reference genome at the Compared with the nucleotide sequence at the corresponding position of the mutation site, the sequence is different, and the reference genome is the human reference genome in the gene sequencing.

b) within a specific length range, respectively construct the distribution patterns of the wild-type support fragment and the mutant support fragment:

Within the length range of 1 to 400 nucleotides, the distribution of the wild-type supporting fragment and the mutant supporting fragment is calculated.

c) where the difference in fragmentation patterns (Dev) between the two groups is quantified within a specific interval, calculated as follows:

D＝[B ₁ ,B ₄ ,B ₇ ...B ₃₉₇ ] (2)

WC _i and MC _i in the formula (1) respectively represent the number of the wild-type supporting fragments with a length of i nucleotides and the number of the wild-type support fragments with a length of i nucleotides at a certain mutation site. The mutant supports the number of fragments.

Wherein, 3 is the smoothing window value;

Wherein, j is the length value in the smoothing sampling length range, for example, j can be an integer in an arithmetic sequence such as 1, 4, 7, 10;

Wherein, 400 is the range of the length of the wild type supporting fragment and/or the mutant supporting fragment.

In other words, with 3 as the interval length, within the range of 1-400 nucleotide lengths, according to the formula (1), calculate the cumulative value of the ratio of different lengths, and the set of these ratios constitutes the first Distribution D (ie formula (2)).

Then, the length of the effective fragment interval is set to be about 1-about 167 nucleotides, and/or, about 250-about 400 nucleotides. In the present application, the length of the effective fragment interval may be the length of the nucleic acid sequence wound around the nucleosome. For example, the nucleic acid sequence can wrap around the nucleosome for more than 2 weeks, or, can wrap around the nucleosome within 1 week (for example, the length of the effective fragment interval is about 1 to about 167 nucleotides, and/or, about 250 to about 400 nucleotides).

Within the interval of the effective segment, the values of each B in the first distribution D (that is, the cumulative value of the ratio) are sequentially accumulated again to obtain the added value (that is, refer to formula (3)) .

For example, assuming that the length of the interval of the effective fragment is 100 (that is, i is 100), then in the range of 1-100 nucleotides in length, the values of each B in the first distribution D are calculated sequentially Bonus value.

The set of added values constitutes the second distribution A, and the largest added value in the second distribution is recorded as Dev(Max) (ie, refer to formula (4)).

Dev＝Max(A) (4)

For example, Figure 8 shows the distribution frequency of the lengths of the wild-type support fragment and the mutant support fragment of the present application obtained by using the method described in Example 2.1 for the mutation site C-T at No. 20525808 of human chromosome 4 .

For example, Figure 9 shows the distribution frequency of the lengths of the wild-type support fragment and the mutant support fragment of the present application obtained by using the method described in Example 2.1 for the mutation site G-T at No. 56189455 of human chromosome 5 .

For example, Figure 10 shows the distribution frequency of the lengths of the wild-type support fragment and the mutant support fragment of the present application obtained by using the method described in Example 2.1 for the mutation site C-A at No. 7577141 of human chromosome 17 .

2.2

D＝[B ₁ ,B ₂ ,B ₃ ...B ₄₀₀ ] (2)

WC _i and MC _i in the formula (1) respectively represent the number of the wild-type supporting fragments with a length of i nucleotides and the number of the wild-type support fragments with a length of i nucleotides at a certain mutation site The number of mutant support fragments.

Wherein, 3 is the smoothing window value;

Wherein, j is the length value in the smoothing sampling length range, for example, j can be an integer in an arithmetic sequence such as 1, 2, 3, 4;

Dev＝Max(A) (4)

Embodiment 3 Carry out the machine learning described in this application

(1) Input the indicators involved in Table 1 into the machine learning model described in this application for machine learning training.

These indicators can be divided into 7 types according to the types of different characteristics, and the indicators are all related to the mutation site.

Table 1

a) Location information: including the chromosome location where the SNV is located, for example, 68771372 on chromosome 16.

b) Base substitution pattern: In a single SNV locus, the wild-type base is transformed into a newly introduced mutant base pattern. For example, chr3, 178935093C>A, the base substitution mode is "CA". This feature uses the "one-hot encoding" method, taking into account the theoretical 12 replacement modes, namely: AT, AC, AG, TA, TC, TG, CA, CT, CG, GA, GT, GC.

c) Dev value obtained in Example 2 (that is, it can reflect the fragmentation mode of cfDNA): it can also characterize the characteristics W _ratio and M _ratio of the mutation shift direction.

To visualize the difference between two groups, the Delta _ratio can also be characterized. The calculation methods of the above three parameters are respectively shown in formula (5), formula (6) and formula (7).

Delta _ratio ＝W _ratio -M _ratio (7)

Wherein, 167 can also be any integer in 160-174.

In formula (5), C _1>167 and C _1<167 respectively represent the number of the wild-type support fragments with a length greater than 167 nucleotides, and the number of wild-type support fragments with a length less than 167 nucleotides. Quantity, W _ratio indicates the ratio of C _l>167 and C _l<167 .

In the formula (6), C _1>167 and C _1<167 respectively represent the number of the mutant support fragments with a length greater than 167 nucleotides, and the number of mutant support fragments with a length less than 167 nucleotides Quantity, M _ratio means the ratio of C _l>167 and C _l<167 .

Formula (7) represents the difference between W _ratio and M _ratio .

d) Fragment count: it includes all unmutated wild-type fragments in a certain mutation site, and the number of all supported fragments in which a single base mutation occurs at this site.

e) Allelic variation: This type of feature includes two types, namely sample frequency and population frequency. The sample frequency refers to the allele mutation frequency (Variant Allele Frequency) that occurs in a certain sample, and the population frequency refers to the frequency of the mutation in the population.

f) Age: the age of the sample that produced the mutation.

g) Mutation type: It is the result of mutation annotation, and the characteristics of this category include the following types:

splice_donor_variant, (splice donor mutation)

synonymous_variant, (synonymous mutation)

stop_gained, (terminator gain)

intron_variant, (intron variant)

stop_lost, (terminator missing)

missense_variant, (nonsense mutation)

splice_region_variant, (splice region mutation)

splice_acceptor_variant, (splice acceptor mutation)

promoter_region_variant, (promoter region mutation)

start_lost (start codon mutation)

After the encoding is completed, z-transform is performed on each feature type, that is, all values are converted into a standard normal distribution with a mean of 0 and a variance of 1.

(2) Model training

Use the ensemble module in the machine learning library sklearn v.0.23.2 in python during model training

parameter settings. Set the discriminant category separation purity method to "entropy", and the maximum decision tree depth is determined by the minimum score of the leaf node

The number of samples is determined to be None, the minimum number of samples that can be divided by a node is 10, and the final result consists of 40 decision trees

vote.

Example 4 Application of the method described in this application to specific tumors

The real data includes a total of 1309 lung cancer blood samples, which are divided into a training set containing 928 samples, and two sets of verification sets containing 191 and 190 samples respectively (that is, the training set, verification set in Figure 1, respectively). set 1 and validation set 2).

First, according to the steps of Examples 1-3, the remaining 12,173 germlines and 5,816 systemic mutations in the training set were modeled after being filtered by population frequency to obtain the machine learning model that has been trained by machine learning.

Then, use the machine learning model that has been trained by machine learning to perform model verification on the above two verification sets (see FIG. 1 ).

During training, 20% equivalent data of all 17989 mutations are divided into the test set. In the 80% training set, an internal 5-fold cross-validation is used to select the hyperparameters of all optimal models, and finally the results of each optimal model in the 20% test set are obtained. The machine training results of the model are shown in Figure 2. In FIG. 2 , the RF(+Dev) or RF(-Dev) respectively refer to the results of model verification of the above two verification sets by machine learning models that include the parameter Dev and do not include the parameter for machine learning training.

The results show that random forest performs best among all models, with an AUC value of 0.9975. In addition, in the above 2 validation sets (Figure 3-4). Figures 3-4 show the performances of the machine learning models in the verification set 1 and verification set 2 that have been trained by machine learning in this application, respectively.

It can be seen that the machine learning model that has been trained by machine learning described in this application also exhibits superior performance, with AUCs reaching 0.9973 and 0.9979 respectively, which proves the generalization ability of the method described in this application.

Example 5 Application of the method described in this application to different tumors

In order to confirm that the machine learning model that has been trained by machine learning described in this application can be comprehensively applied to the germline system discrimination of pan-cancer species, a total of 1008 samples from 11 cancer types were used (see Figure 5 for details of the samples). After filtering methods such as population frequency, 6,647 systemic mutations and 13,567 germline mutations were finally included in the assessment (Figure 5).

Overall, the machine learning model described in this application that has been trained by machine learning has good predictive ability for the mixed 1008 multi-cancer test sets, and the AUC has reached 0.9947 (see Figure 6), where cfSvG represents the applicant's developed The name of the algorithm.

Additionally, the model was tested for its ability to classify each cancer type. It was found that the AUC of the model was stable above 0.99 in almost all 11 cancers. But in the bladder cancer data, the performance dropped slightly, but its AUC also reached 0.9886 (see Figure 7 for AUC results).

The method and/or model described in this application not only perform well in lung cancer, but also have excellent performance in the classification ability of pan-cancer.

The embodiments of the present application have been described in detail above, but the present application is not limited to the specific details in the above-mentioned embodiments. Within the scope of the technical concept of the present application, various simple modifications can be made to the technical solutions of the present application. These simple modifications are all Belong to the protection scope of this application. In addition, it should be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable way if there is no contradiction. The combination method will not be described separately. In addition, any combination of various implementations of the present application can also be made, as long as they do not violate the idea of the present application, they should also be regarded as the content disclosed in the present application.

Claims

A method for distinguishing somatic mutations from germline mutations, comprising the steps of:

(1) Obtaining at least one mutation site derived from a subject sample; wherein, the mutation site is obtained by gene sequencing,

(2) Obtain a wild-type support fragment and a mutant support fragment for each of the mutation sites;

Wherein, the wild-type supporting fragment is a cfDNA fragment comprising a wild-type base sequence,

The mutant supporting fragment is a cfDNA fragment comprising a mutant base sequence,

Wherein, the wild-type base sequence is, compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome, the same sequence,

Wherein, the mutant base sequence is a different sequence compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome,

Wherein said reference genome is the human reference genome in said gene sequencing;

(3) For each mutation site, obtain the number of wild-type supporting fragments of at least one length, and obtain the corresponding number of mutant-type supporting fragments of the same length,

calculating the ratio WC of the number of wild-type support fragments of the length to the total number of wild-type support fragments;

calculating the ratio MC of the number of mutant supporting fragments of the same length to the total number of mutant supporting fragments;

calculating the difference between said ratio WC and said ratio MC under the same length;

(4) Using the difference or the set of differences as an index for distinguishing whether the mutation site is a somatic mutation or a germline mutation.
A method for identifying ctDNA in cfDNA comprising the steps of:

(1) Obtaining at least one mutation site derived from a subject sample; wherein, the mutation site is obtained by gene sequencing,

(2) Obtain a wild-type support fragment and a mutant support fragment for each of the mutation sites;

Wherein, the wild-type supporting fragment is a cfDNA fragment comprising a wild-type base sequence,

The mutant supporting fragment is a cfDNA fragment comprising a mutant base sequence,

Wherein, the wild-type base sequence is, compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome, the same sequence,

Wherein, the mutant base sequence is a different sequence compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome,

Wherein said reference genome is the human reference genome in said gene sequencing;

(3) For each mutation site, obtain the number of wild-type supporting fragments of at least one length, and obtain the corresponding number of mutant-type supporting fragments of the same length,

calculating the ratio WC of the number of wild-type support fragments of the length to the total number of wild-type support fragments;

calculating the ratio MC of the number of mutant supporting fragments of the same length to the total number of mutant supporting fragments;

calculating the difference between said ratio WC and said ratio MC under the same length;

(4) Using the difference value or the set of difference values as an indicator for identifying whether the mutation site is ctDNA.
A training method for a machine learning model, comprising the following steps:

(1) Obtaining at least one mutation site derived from a subject sample; wherein, the mutation site is obtained by gene sequencing,

(2) Obtain a wild-type support fragment and a mutant support fragment for each of the mutation sites;

Wherein, the wild-type supporting fragment is a cfDNA fragment comprising a wild-type base sequence,

The mutant supporting fragment is a cfDNA fragment comprising a mutant base sequence,

Wherein, the wild-type base sequence is, compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome, the same sequence,

Wherein, the mutant base sequence is a different sequence compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome,

Wherein said reference genome is the human reference genome in said gene sequencing;

(3) For each mutation site, obtain the number of wild-type supporting fragments of at least one length, and obtain the corresponding number of mutant-type supporting fragments of the same length,

calculating the ratio WC of the number of wild-type support fragments of the length to the total number of wild-type support fragments;

calculating the ratio MC of the number of mutant supporting fragments of the same length to the total number of mutant supporting fragments;

calculating the difference between said ratio WC and said ratio MC under the same length;

(4) Inputting the difference or a set of difference values into the machine learning model as a training index to perform machine learning training.
A database establishment method, it comprises the following steps:

(1) Obtaining at least one mutation site derived from a subject sample; wherein, the mutation site is obtained by gene sequencing,

(2) Obtain a wild-type support fragment and a mutant support fragment for each of the mutation sites;

Wherein, the wild-type supporting fragment is a cfDNA fragment comprising a wild-type base sequence,

The mutant supporting fragment is a cfDNA fragment comprising a mutant base sequence,

Wherein, the wild-type base sequence is, compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome, the same sequence,

Wherein, the mutant base sequence is a different sequence compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome,

Wherein said reference genome is the human reference genome in said gene sequencing;

(3) For each mutation site, obtain the number of wild-type supporting fragments of at least one length, and obtain the corresponding number of mutant-type supporting fragments of the same length,

calculating the ratio WC of the number of wild-type support fragments of the length to the total number of wild-type support fragments;

calculating the ratio MC of the number of mutant supporting fragments of the same length to the total number of mutant supporting fragments;

calculating the difference between said ratio WC and said ratio MC under the same length;

(4) storing the difference or the set of differences in a database for distinguishing somatic mutations from germline mutations, and/or identifying ctDNA from cfDNA.
The method according to any one of claims 1-4, wherein the wild-type support fragment and/or the mutant support fragment have a length ranging from about 1 nucleotide to about 550 nucleotides, Or about 1 nucleotide to about 400 nucleotides, or about 1 nucleotide to about 200 nucleotides.
The method according to any one of claims 1-4, comprising the steps of:

(4') Obtain the distribution of the difference in step (3), select the maximum value in the distribution as Dev(Max), use the Dev(Max) as the distinguishing index and/or as the training sample.
The method according to any one of claims 1-4, comprising the steps of:

(4') Obtain the distribution of the difference described in step (3), which is called the first distribution.
The method according to claim 7, comprising the steps of:

(5) within the length range of the effective fragment interval, each difference in the first distribution is sequentially accumulated to obtain an added value, wherein the length of the effective fragment interval covers the nucleic acid sequence wound around the nucleosome length.
The method according to claim 8, wherein the length of the effective fragment interval is about 1 to about 167 nucleotides, and/or about 200 or more nucleotides, for example, about 250 to about 400 nucleosides acid.
The method according to claim 8, comprising the steps of:

(6) Obtaining the second distribution of the added value in step (5), and calculating the maximum value of the added value in the second distribution.
The method according to claim 10, wherein the maximum value of the added value is used as Dev(Max), and the Dev(Max) is used as the distinguishing index and/or as the training sample.
The method according to claim 1, wherein said difference is smoothed, wherein said smoothing comprises the steps of:

(a) determining a smoothing window value, wherein the smoothing window value is an integer from about 1-10;

(b) determine several smoothing sampling length ranges whose length values are equal to the smoothing window value, wherein the minimum value of each smoothing sampling length range is the initial length,

Wherein the range of the initial length is the range of the length of the wild type support fragment and/or the length of the mutant support fragment;

(c) Obtain the number of wild-type support fragments of at least one smoothed sampling length in any smoothed sampling length range, obtain the corresponding number of mutant support fragments of the same length, and calculate the length of the The ratio WC of the number of wild-type support fragments to the total number of wild-type support fragments; calculate the ratio MC of the number of mutant support fragments of the same length to the total number of mutant support fragments;

calculating the difference between said ratio WC and said ratio MC under the same length;

(d) calculating an average difference for a range of smoothed sample lengths based on said difference for said at least one smoothed sample length;

(e) Taking the obtained average difference as a representative value of the smoothed sampling length range.
The method according to claim 12, wherein the smoothing window value is an integer between about 2-6, for example, the smoothing window value is 3.
The method according to claim 12, wherein said smoothing process comprises the steps of:

(f) Obtaining a first distribution of said mean differences of step (e).
The method according to claim 14, wherein said smoothing process comprises the steps of:

(g) within the length range of the effective segment interval, each average difference in the first distribution is sequentially accumulated to obtain the bonus value,

Wherein, the length of the effective fragment interval is the length of the nucleic acid sequence wound around the nucleosome.
The method according to claim 15, wherein said smoothing process comprises the steps of:

(h) Obtaining the second distribution of the added value in step (g), and calculating the maximum value of the added value in the second distribution.
The method according to claim 16, wherein the maximum value is used as Dev(Max), and the Dev(Max) is used as the distinguishing index and/or as the training sample.
A means for distinguishing between somatic and germline mutations comprising:

Calculation module, for calculating the difference between the ratio WC and the ratio MC of the same length;

Wherein, for each mutation site, according to the number of wild-type supporting fragments of at least one length, and the corresponding number of mutant-type supporting fragments of the same length;

The ratio WC is the ratio of the number of wild-type support fragments of one length to the total number of wild-type support fragments;

Wherein the ratio MC is the ratio of the number of corresponding mutant-type supporting fragments of the same length to the total number of the mutant-type supporting fragments;

Wherein, the wild-type support fragment is a cfDNA fragment comprising a wild-type base sequence, and the mutant support fragment is a cfDNA fragment comprising a mutant base sequence,

Wherein, the wild-type base sequence is, compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome, the same sequence,

Wherein, the mutant base sequence is a different sequence compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome,

Wherein, the reference genome is the human reference genome in the gene sequencing;

The mutation site is derived from a subject sample, wherein the mutation site is obtained by gene sequencing,

a judging module, configured to obtain a recognition result for identifying the somatic mutation according to a machine learning model that has been trained by machine learning,

The machine learning training includes inputting the difference as a training sample into the machine learning model for machine learning training.
Means for identifying ctDNA in cfDNA, comprising:

Calculation module, for calculating the difference between the ratio WC and the ratio MC of the same length;

Wherein, for each mutation site, according to the number of wild-type supporting fragments of at least one length, and the corresponding number of mutant-type supporting fragments of the same length;

The ratio WC is the ratio of the number of wild-type support fragments of one length to the total number of wild-type support fragments;

Wherein the ratio MC is the ratio of the number of corresponding mutant-type supporting fragments of the same length to the total number of the mutant-type supporting fragments;

Wherein, the wild-type support fragment is a cfDNA fragment comprising a wild-type base sequence, and the mutant support fragment is a cfDNA fragment comprising a mutant base sequence,

Wherein, the wild-type base sequence is, compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome, the same sequence,

Wherein, the mutant base sequence is a different sequence compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome,

Wherein, the reference genome is the human reference genome in the gene sequencing;

The mutation site is derived from a subject sample, wherein the mutation site is obtained by gene sequencing,

a judging module, configured to obtain a judging result of identifying ctDNA from the cfDNA according to a machine learning model that has been trained by machine learning,

The machine learning training includes inputting the difference as a training sample into the machine learning model for machine learning training.
A training device for a machine learning model, comprising:

Calculation module, for calculating the difference between the quantity of the wild-type support fragment and the quantity of the mutant support fragment of the same length;

Wherein, the number of wild-type support fragments includes the number of wild-type support fragments of at least one length for each mutation site, and the number of mutant support fragments includes the corresponding mutant support fragments of the same length. number of fragments,

Wherein, the wild-type support fragment is a cfDNA fragment comprising a wild-type base sequence, and the mutant support fragment is a cfDNA fragment comprising a mutant base sequence,

Wherein, the wild-type base sequence is, compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome, the same sequence,

Wherein, the mutant base sequence is a different sequence compared with the nucleotide sequence at the corresponding position of the mutation site in the reference genome,

Wherein, the reference genome is the human reference genome in the gene sequencing;

The mutation site is derived from a subject sample, wherein the mutation site is obtained by gene sequencing,

A training module, configured to input the difference as a training sample into the machine learning model for machine learning training.