WO2021227950A1 - Cancer prognostic method - Google Patents

Abstract

Disclosed is a prognostic method for a cancer patient. The relapse risk and/or survival of a patient are/is predicted on the basis of a methylation difference in genomes of a cancer tissue sample and a paracancerous tissue sample of the patient. Further disclosed are a prognostic system and device for a cancer patient.

Description

Cancer prognosis methods

This application claims the priority of the Chinese patent application filed on May 9, 2020 with the application number 202010385996.8 and the invention title "Methods for Cancer Prognosis", and the content of the above patent application is quoted here in full as a part of this application.

Technical field

The present disclosure generally relates to the field of biological detection and diagnosis. Specifically, the present disclosure relates to a method for prognosing a patient based on the difference in methylation in the genome of the cancer tissue sample and the adjacent tissue sample of the patient. The present disclosure also relates to systems and devices for prognosing cancer patients.

technical background

Globally, cancer is one of the main causes of death and disease burden. Cancer prognosis is the prediction of the possible outcomes of an individual's current medical condition, and it is an important tool for improving patient diagnosis and treatment management. An accurate prognosis is essential for choosing the right cancer treatment and predicting survival rates. At present, clinically, the clinical stage and pathological analysis of tumors are mainly used to assist some related molecular characteristics (such as immunohistochemistry, DNA mutations, mRNA or microRNA expression, etc.) to evaluate and predict the prognosis of patients. However, the above molecular assessment methods still have great limitations. On the one hand, molecular markers of 21 genes in breast cancer can accurately predict the risk of postoperative recurrence of patients and bring significant clinical benefits to breast cancer patients. , And thus entered the guidelines for the diagnosis and treatment of breast cancer; on the other hand, molecular markers with high accuracy and clinical application efficiency are still very scarce, and most cancer types still require frequent imaging follow-up of patients in clinical practice. To monitor the course of the disease, it will cause a certain burden to the patient, and there may be a delay in the discovery of tumor progression. Therefore, independent of pathological stage and other clinical factors, molecular markers that can accurately predict the prognosis of patients have huge clinical needs.

In many studies, it has been proved that DNA methylation mutations are closely related to the occurrence of cancer, and compared with gene mutations, DNA methylation mutations have the characteristics of wider coverage, higher stability, and earlier occurrence. , So it is more suitable for early detection of cancer. However, methods and strategies for predicting the prognosis of cancer patients using DNA methylation mutations are still very lacking.

Summary of the invention

The theory of "regional carcinogenesis" believes that under the action of a certain mechanism, normal tissues will gradually begin the process of carcinogenesis at the molecular level, and this change will first appear in the DNA. Based on the theory of "regional carcinogenesis", the inventors designed a new system that uses DNA methylation detection technology to predict the prognostic recurrence risk of cancer patients-malignancy density ratio (MD ratio) Evaluation system.

The MD ratio evaluation system is based on the following theory, that is, normal tissues in patients are in the process of transforming from normal cells to cancer cells. Through the detection of methylation mutations, the process of canceration can be evaluated, so as to predict the risk of tumor recurrence in patients. Compared with traditional testing, the MD ratio evaluation system predicts the patient’s recurrence risk only by testing tissue samples from the patient, which can better manage the patient’s prognosis, avoid frequent follow-up after surgery, and is simple and efficient. , Personalized features, so it has a better application prospect in prognosis management. Moreover, in the analysis of real samples, the system has higher accuracy than mutation detection in predicting recurrence.

Accordingly, in one aspect, the present disclosure relates to a method for prognosing cancer patients, the method comprising:

a. Detecting the methylation level in one or more regions of the genome of cancer tissues and paracancerous tissues from the patient by high-throughput sequencing;

b. Determine the difference in the methylation level between the cancer tissue and the adjacent tissue; and

c. Use the difference information of the methylation level to predict the patient's prognosis through mathematical modeling,

Wherein, the smaller the difference of the methylation level between the cancer tissue and the adjacent tissues is, the worse the prognosis of the patient is.

In some embodiments, the method includes:

a. Detecting the methylation level in one or more regions of the genome of cancer tissues and paracancerous tissues from the patient by high-throughput sequencing;

b. Determine the differential methylation block (DMB) in the cancer tissue and adjacent tissues, including:

b1. For each CpG site in the detected area, record the number of methylated reads covering this site as M, and the number of unmethylated reads as U;

. b2 adjacent CpG sites are combined and defined as methylated block (MB); M on the i-th MB patients all CpG sites are added, referred to as M _i; the sum of all U , Denoted as U _i ; the total coverage is N _i =M _i +U _i , and the methylation level is β _i =M _i /N _i ;

b3. For the i th MB, the level of methylation in cancer tissues referred to as ^{_{β i (T) = M i}} (T) / N i (T), the level of methylation adjacent tissue cancer referred to as beta] _i ^(A) =M _i ^(A) /N _i ^(A) , where the MB with a significant difference between _{β i} ^(T) and β _i ^{(A) is determined as DMB;}

c. Prognosis of the patient, including:

c1. Introduce a parameter α indicating the degree of similarity between the methylation levels of the paracancerous tissue and the cancer tissue of the patient, and calculate the patient's α value by the following algorithm:

表示先验分布的参数族； Where 0＜α＜1 and a larger value of α indicates a higher degree of similarity, where f(·), g(·), h(·) respectively represent the conditional distribution of the methylation level of adjacent tissues, cancer The prior distribution of the methylation level of the tissue and the prior distribution of the baseline methylation level,

Represents the parameter family of the prior distribution;

并通过fisher信息矩阵计算参数估计的方差

c2. Use maximum likelihood estimation algorithm to calculate parameter estimates in the range of 0＜α＜1

And calculate the variance of parameter estimates through the fisher information matrix

c3. Calculate the p value of the null hypothesis α=0 for each patient’s adjacent tissues,

Among them, the smaller the p value indicates the worse the prognosis of the patient.

In some embodiments, combining adjacent CpG sites in step b2 can be, for example, combining a distance of less than 50 bp, a distance of less than 100 bp, a distance of less than 150 bp, a distance of less than 200 bp, a distance of less than 250 bp, a distance of less than 300 bp, and a distance of less than 350 bp. , Combine CpG sites with a distance of less than 400 bp, a distance of less than 450 bp, or a distance of less than 500 bp. In some embodiments, CpG sites with a distance of less than 50 bp or a distance of less than 100 bp are combined in step b2.

的MB确定为DMB，其中σ为0.05至1之间的值且τ为大于0.1的值。在一些实施方案中，σ＝0.1且τ＝0.4。 In some embodiments, in step b3, _{determining the MB with a significant difference between β i} ^(T) and β _i ^(A) as DMB can be, for example, determining |β _i ^(T) -β _i ^(A) |> σ, and

The MB of is determined as DMB, where σ is a value between 0.05 and 1 and τ is a value greater than 0.1. In some embodiments, σ=0.1 and τ=0.4.

In some embodiments of the above method, in step c1, the patient's alpha value is calculated by the following algorithm:

_{l (α) = Σ i log} L (α; M i (A), N i (A), p i (0), q i (0), β i (T))

Among them, l(α) is the log-likelihood function of the observation data, p _i ⁽⁰⁾ , q _i ⁽⁰⁾ is the Beta-binomial distribution subject to the baseline methylation level of the i-th MB (p _i ^{(0 )} , The shape parameter in _{q i} ^{(0) ).}

In some embodiments, one or more regions of the genome are regions of the genome where there are methylation variants in the population of cancer patients. For example, the region of the genome that is known to have methylation variants in a patient population of a specific cancer type can be obtained through a public database and used in the detection method of the present disclosure. For example, it is possible to download the methylation chip data of cancer tissues and para-cancerous samples in various cancer patient groups from these public databases, and use these data to determine the methylation characteristic regions closely related to the occurrence of each cancer. It can be determined by using, for example, parameter testing methods (such as t-test, linear regression, etc.), non-parametric testing methods (such as Wilcoxon rank sum test, etc.). The aforementioned public databases are, for example, the TCGA (The Cancer Genome Atlas) database and the GEO (Gene Expression Omnibus) database.

In some embodiments, one or more regions of the genome cover a region of at least 0.3M (megabases), such as at least 0.3M, 0.4M, at least 0.5M, at least 0.6M, at least 0.7M, at least 0.8M , At least 0.9M or at least 1.0M area.

In some embodiments, one or more regions of the genome cover about 0.3M-10.0M regions of the genome, such as 0.3M-5.0M, 0.3M-4.0M, 0.3M-3.0M regions, 0.3M-3.0M regions, 2.0M area, 0.3M-1.5M area, 0.3M-1.0M area, 0.4M-5.0M area, 0.4M-4.0M area, 0.4M-3.0M area, 0.4M-2.0M Area, 0.4M-1.5M area, 0.4M-1.0M area, 0.5M-5.0M area, 0.5M-4.0M area, 0.5M-3.0M area, 0.5M-2.0M area , 0.5M-1.5M area, 0.5-1.0M area, or 1.0M-5.0M area, 1.0M-4.0M area, 1.0M-3.0M area, 1.0M-2.0M area or 1.0 M-1.5M area. The above range also includes endpoint values and any subset ranges in between.

In some embodiments, the cancer is a solid tumor. Examples of solid tumors include, but are not limited to, lung cancer (including small cell lung cancer, non-small cell lung cancer, lung adenocarcinoma, and lung squamous cell carcinoma), colorectal cancer, liver cancer, ovarian cancer, pancreatic cancer, gallbladder cancer, gastric cancer, esophageal cancer , Kidney cancer, melanoma, breast cancer, cervical cancer, endometrial cancer, prostate cancer, bladder cancer, testicular cancer, thyroid cancer, salivary gland cancer, skin cancer, squamous cell carcinoma, neuroblastoma, glioblastoma Tumor, retinoblastoma, lymphoma (including Hodgkin’s lymphoma and non-Hodgkin’s lymphoma), bone cancer, myeloma, basal cell carcinoma, peritoneal cancer, choriocarcinoma, eye cancer, head and neck cancer, laryngeal cancer , Oral cancer and rhabdomyosarcoma.

In some embodiments, the cancer may be selected from lung cancer, colorectal cancer, liver cancer, ovarian cancer, pancreatic cancer, gallbladder cancer, gastric cancer, and esophageal cancer.

In some embodiments, the cancer is a primary cancer. In other embodiments, the cancer is a secondary or metastatic cancer. The cancer may be in any stage of cancer development, such as early, middle or late stages of cancer development, or the cancer may be in clinical stages I, II, III, or IV.

In some embodiments, the cancer is lung cancer, such as non-small cell lung cancer (NSCLC). In the case where the cancer is lung cancer, one or more regions of the detected genome may include one or more regions selected from those listed in Table 1.

Table 1.

In some embodiments, the cancer is lung cancer, and one or more regions of the detected genome include at least 100 regions, at least 200 regions, at least 300 regions, and at least 400 regions selected from those listed in Table 1. , At least 500 areas, at least 600 areas, at least 700 areas, at least 800 areas, at least 900 areas, or at least 1000 areas. In some embodiments, the one or more regions of the detected genome include all regions selected from the list in Table 1.

In some embodiments, the cancer patient has undergone previous cancer treatment methods, such as surgical treatment, radiation therapy, chemotherapy, targeted drug therapy, immunotherapy, or a combination thereof.

In some embodiments of the above methods, the cancer tissue and para-cancerous tissue used may be tissues surgically removed from the patient.

In some embodiments of the methods described above, using methylation data from normal tissues (e.g., normal tissues from healthy donors) to establish the baseline methylation level includes:

For the i-th MB, methylated normal tissue data referred to as M _i ⁽⁰⁾ and N _i ^(0), that at a given N _i ⁽⁰⁾ of M _i ⁽⁰⁾ is subject to a shape parameter ( Beta-binomial distribution of p _i ⁽⁰⁾ , q _i ^{(0) ):}

M _i ⁽⁰⁾ |N _i ⁽⁰⁾ , β _i ⁽⁰⁾ ～ Binomial(N _i ⁽⁰⁾ , β _i ⁽⁰⁾ )

β _i ⁽⁰⁾ ～Beta(p _i ⁽⁰⁾ , q _i ⁽⁰⁾ ),

And use the maximum likelihood algorithm to calculate the parameters p _i ⁽⁰⁾ and q _i ⁽⁰⁾ .

In some embodiments of the above method, the p-value is calculated by Wald test in step c3. In other embodiments, the p-value is calculated by, for example, a likelihood ratio test.

In some embodiments of the above methods, the method is used to predict the postoperative recurrence risk and/or survival of the cancer patient. In some embodiments, patients with p<0.05 are identified as having a high risk of recurrence and/or low postoperative survival.

In a second aspect, the present disclosure relates to a system for prognosing cancer patients, the system including:

Methylation detection module; and

Prognostic analysis module,

Wherein, the methylation sequencing module is configured to detect the methylation level in one or more regions of the genome of cancer tissue and paracancerous tissue from the patient through high-throughput sequencing, and the prognostic analysis module is configured For the prognosis of the patient through the following methods:

a. Determine the difference in the methylation level between the cancer tissue and the paracancerous tissue; and

b. Using the difference information of the methylation level, the prognosis of the patient is performed by mathematical modeling,

Wherein, the smaller the difference of the methylation level between the cancer tissue and the adjacent tissues is, the worse the prognosis of the patient is.

In some embodiments, the system is configured to implement the prognostic method according to the first aspect of the present disclosure.

In some embodiments, the prognostic analysis module is configured to prognose the patient through the following methods:

a. Determine the differential methylation block (DMB) in the cancer tissue and the adjacent tissue, including:

a1. For each CpG site in the detected area, record the number of methylated reads covering this site as M, and the number of unmethylated reads as U;

. a2 adjacent CpG sites are combined and defined as methylated block (MB); M on the i-th MB patients all CpG sites are added, referred to as M _i; the sum of all U , Denoted as U _i ; the total coverage is N _i =M _i +U _i , and the methylation level is β _i =M _i /N _i ;

a3. For the i th MB, the level of methylation in cancer tissues referred to as ^{_{β i (T) = M i}} (T) / N i (T), the level of methylation adjacent tissue cancer referred to as beta] _i ^(A) =M _i ^(A) /N _i ^(A) , where the MB with a significant difference between _{β i} ^(T) and β _i ^{(A) is determined as DMB;}

b. Prognosis of the patient, including:

b1. Introduce a parameter α indicating the degree of similarity between the methylation levels of the paracancerous tissue and the cancer tissue of the patient, and calculate the patient's α value through the following algorithm:

表示先验分布的参数族； Where 0＜α＜1 and the larger the value of α indicates the higher the degree of similarity, where f(·), g(·), h(·) respectively represent the conditional distribution of methylation levels of adjacent tissues, cancer tissues The prior distribution of methylation levels and the prior distribution of baseline methylation levels,

Represents the parameter family of the prior distribution;

并通过fisher信息矩阵计算参数估计的方差

b2. Use maximum likelihood estimation algorithm to calculate parameter estimates in the range of 0＜α＜1

And calculate the variance of parameter estimates through the fisher information matrix

b3. Calculate the p value of the null hypothesis α=0 for each patient’s adjacent tissues,

Among them, the smaller the p value indicates the worse the prognosis of the patient.

In some embodiments, the prognostic analysis module is configured to set a distance of less than 50 bp, a distance of less than 100 bp, a distance of less than 150 bp, a distance of less than 200 bp, a distance of less than 250 bp, a distance of less than 300 bp, a distance of less than 350 bp, and a distance of less than 400 bp in step a2. Combine CpG sites with a distance of less than 450 bp or a distance of less than 500 bp. In some embodiments, the prognostic analysis module is configured to combine CpG sites with a distance of less than 50 bp or a distance of less than 100 bp in step b2.

的MB确定为DMB，其中σ为0.05至1之间的值且τ为大于0.1的值。在一些实施方案中，σ＝0.1且τ＝0.4。 In some embodiments, the prognostic analysis module is configured to determine | β _i ^(T) -β _i ^(A) |>σ in step b3, and

The MB of is determined as DMB, where σ is a value between 0.05 and 1 and τ is a value greater than 0.1. In some embodiments, σ=0.1 and τ=0.4.

In some embodiments, the prognostic analysis module is configured to calculate the alpha value of the patient through the following algorithm in step b1:

_{l (α) = Σ i log} L (α; M i (A), N i (A), p i (0), q i (0), β i (T))

Among them, l(α) is the log-likelihood function of the observation data, p _i ⁽⁰⁾ , q _i ⁽⁰⁾ is the Beta-binomial distribution subject to the baseline methylation level of the i-th MB (p _i ^{(0 )} , The shape parameter in _{q i} ^{(0) ).}

In some embodiments, the prognostic analysis module is further configured to use methylation data from normal tissues to establish the baseline methylation level, including:

For the i-th MB, methylated normal tissue data referred to as M _i ⁽⁰⁾ and N _i ^(0), that at a given N _i ⁽⁰⁾ of M _i ⁽⁰⁾ is subject to a shape parameter ( Beta-binomial distribution of p _i ⁽⁰⁾ , q _i ^{(0) ):}

M _i ⁽⁰⁾ |N _i ⁽⁰⁾ , β _i ⁽⁰⁾ ～ Binomial(N _i ⁽⁰⁾ , β _i ⁽⁰⁾ )

β _i ⁽⁰⁾ ～Beta(p _i ⁽⁰⁾ , q _i ⁽⁰⁾ ),

And use the maximum likelihood algorithm to calculate the parameters p _i ⁽⁰⁾ and q _i ⁽⁰⁾ .

In some embodiments of the above system, the system is further configured to calculate the p-value by Wald test in step b3. In other embodiments, the p-value is calculated by, for example, a likelihood ratio test.

In some embodiments of the aforementioned system, the system is further configured to predict the postoperative recurrence risk and/or survival of the cancer patient. In some embodiments, patients with p<0.05 are identified as having a high risk of recurrence and/or low postoperative survival.

In a third aspect, the present disclosure relates to a device for prognosing cancer patients, which includes:

Memory for storing computer program instructions; and

A processor for executing computer program instructions,

Wherein when the computer program instructions are executed by the processor, the device executes the method according to the first aspect of the present disclosure.

In a fourth aspect, the present disclosure relates to a computer-readable medium storing computer program instructions, wherein when the computer program instructions are executed by a processor, the computer program instructions according to the first aspect of the present disclosure are implemented. The method described.

Description of the drawings

Figure 1 shows the line chart of the disease-free survival (DFS) results of patients with high risk of recurrence and low risk of recurrence predicted by detecting the cancer tissue and the genomic regions listed in Table 1 in the patient's cancer tissue and paracancerous group using the MD ratio method .

Figure 2 shows the line graph of the disease-free survival (DFS) results of patients with high risk of recurrence and low risk of recurrence predicted by detecting the cancer tissue and the genomic regions listed in Table 3 in the patient's cancer tissue and paracancerous group using the MD ratio method .

Figure 3 shows a line chart of the disease-free survival (DFS) results of patients with high risk of recurrence and low risk of recurrence predicted by detecting the cancer tissue and the genomic regions listed in Table 4 in the patient's cancer tissue and paracancerous group using the MD ratio method .

Detailed ways

The present invention will be further described below in conjunction with specific embodiments, and the advantages and features of the present invention will become clearer with the description. However, these embodiments are only exemplary, and do not constitute any limitation to the scope of the present invention. Those skilled in the art should understand that the details and forms of the technical solution of the present invention can be modified or replaced without departing from the spirit and scope of the present invention, but these modifications and replacements fall within the protection scope of the present invention.

data preparation

The main strategy of the MD ratio assessment system is to sequence the methylation levels of tissue samples to find areas with differences in methylation status in the genome of cancer tissues and normal tissues (such as paracancerous tissues) in patients. Based on the paracancerous tissues in the process of transforming from normal cells to cancer cells, calculate the degree of similarity between the methylation level of the paracancerous tissues and the methylation level of cancer tissues in the different regions, and then infer the cancerous transformation of normal cells in the patient Risk level.

The MD ratio assessment system for cancer recurrence risk detects the patient’s cancer tissue and para-cancerous tissue samples. Take lung cancer as an example. In lobectomy and segmentectomy, the adjacent tissue is defined as the tissue outside the resection margin 5cm; in the wedge resection, the adjacent tissue is defined as the tissue outside the resection margin 3cm, and pass The pathological evaluation of the tissue cells verifies that they do not contain tumor cells; at the same time, the cell types of the cancer tissue and the adjacent tissues are the same. In addition, 68 normal tissue samples from healthy donors (age distribution 30-86 years; median age 57.5 years; 35 men and 33 women) were used for the baseline construction process, and the above samples were also pathologically evaluated Verify that it does not contain tumor cells and is consistent with the cell type of cancer tissue.

The sample library is prepared using the brELSATM method (Burning Rock Biotech, Guangzhou, China), which includes the following steps: 1) DNA extraction and purification; 2) sodium bisulfite treatment; 3) single-stranded DNA amplification by DNA polymerase; 4) Use a customized cancer methylation profile RNA bait to enrich the target region as shown in Table 1 (covering a region of about 0.9M of the human genome); 5) Quantify the target library by real-time PCR. Finally, the sequencer NovaSeq 6000 released by Illumina was used for sequencing, and the average sequencing depth was 1,000 layers.

)。 The original output files of sequencing were analyzed using sequence comparison software BWA-meth and methylation data statistics software MethylDackel to obtain the methylation detection output files of each sample. It contains the location information of each CpG site in the specific capture area, and the methylation information in the reads covering this site. The number of methylated reads covering this site is recorded as M, and the number of unmethylated reads is recorded as U. Combining adjacent CpG sites (with a distance of less than 50 bp), this set of multiple CpG sites is called a methylation block (MB). M on the i-th MB patients all sites are added, referred to as M _i, referred to as the sum of all U _i U, the total number of coverage referred to as _{N i (N i = M i} + U i), The methylation level is denoted as β _i (the moment estimation of _{β i}

).

Differential methylation region screening, observation data modeling and parameter solving

In order to detect the degree to which the patient’s para-cancerous tissue is similar to cancer tissue but different from normal tissues, the research target is targeted at the regions of the cancer and para-cancerous tissues where there is a difference in methylation in the genome, which is defined as a differentially methylated block ( differential methylated block, referred to as DMB).

的MB确定为该患者的个性化DMB。 In order to screen the individual DMB of cancer patients, the methylation level β ^(T) of the cancer tissue sample of the patient and the methylation level β ^{(A) of the} adjacent tissue sample are used for testing. For the i-th MB, the level of methylation in cancer tissues referred to as _{^{β i (T) = M i}} (T) / N i (T), the level of methylation adjacent tissue cancer referred to as β _i ^{(A )} = M _i ^(A) /N _i ^(A) , which will conform to |β _i ^(T) -β _i ^(A) |>0.1, and

The MB is determined to be the patient’s personalized DMB.

In order to establish a corresponding statistical model, first use normal tissue samples (as described in Example 1) to establish a baseline of methylation level. For the i-th MB, methylated normal tissue data referred to as M _i ⁽⁰⁾ and N _i ^(0), that at a given N _i ⁽⁰⁾ of M _i ⁽⁰⁾ is subject to a shape parameter ( Beta-binomial distribution of p _i ⁽⁰⁾ , q _i ^{(0) ):}

M _i ⁽⁰⁾ |N _i ⁽⁰⁾ , β _i ⁽⁰⁾ ～ Binomial(N _i ⁽⁰⁾ , β _i ⁽⁰⁾ )

β _i ⁽⁰⁾ ～Beta(p _i ⁽⁰⁾ , q _i ⁽⁰⁾ ),

Wherein β _i ⁽⁰⁾ represents the baseline of the actual degree of methylation using the maximum likelihood algorithm (maximum likelihood estimation, referred MLE) solving the shape parameter {p _i, q _i}, the maximum likelihood estimates of the parameters in mind It is {p _i ⁽⁰⁾ , q _i ⁽⁰⁾ }.

For cancer patients, that methylation sequencing data ^{_{(M i (A), N}} i (A), β i (A)) which is adjacent to obey a cancerous tissue sample hybrid Beta-Binomial distribution may be specifically expressed as:

M _i ^(A) |N _i ^(A) , β _i ^(A) ～ Binomial(N _i ^(A) , β _i ^(A) )

β _i ^(A) _i = αβ _i ^(T) + (1-α)β _i ⁽⁰⁾

β _i ⁽⁰⁾ ～Beta(p _i ⁽⁰⁾ , q _i ⁽⁰⁾ )

代替；β _i ⁽⁰⁾表示基线的甲基化水平，如上所述其服从形状参数为(p _i ⁽⁰⁾，q _i ⁽⁰⁾)的Beta分布。α为[0，1]上的比例参数，表示癌旁组织和癌症组织的相似性。α越接近于1则说明癌旁组织的甲基化程度越接近癌症组织，可以推断患者的复发风险越高。 Where β _i ^(T) represents the methylation level of the cancer tissue of the patient, which can be estimated by the moment of the sequencing data of the cancer tissue sample

Place; β _i ⁽⁰⁾ represents the baseline level of methylation, as described above, which is subject to the shape parameter ^{_{(p i (0), q}} i (0)) of the Beta distribution. α is a ratio parameter on [0,1], which indicates the similarity between adjacent tissues and cancerous tissues. The closer α is to 1, the closer the degree of methylation of the adjacent tissues is to the cancer tissue, and it can be inferred that the patient's risk of recurrence is higher.

According to the above distribution, the log likelihood function of the observation data can be written:

_{l (α) = Σ i log} L (α; M i (A), N i (A), p i (0), q i (0), β i (T))

The parameter α satisfies:

0＜α＜1

并通过fisher信息矩阵计算参数估计的方差

Calculate the parameter α through the maximum likelihood algorithm, and use the quasi-Newton algorithm to calculate the parameter estimation in the range of 0＜α＜1

And calculate the variance of parameter estimates through the fisher information matrix

Infer the risk of cancer recurrence through the similarity between adjacent tissues and cancer tissues: make statistical inferences on the null hypothesis of parameter α: α=0. Use Wald test to infer the null hypothesis:

Under the chi-square distribution with 1 degree of freedom, the p value of α=0 for each patient’s paracancerous tissue can be calculated. The smaller the p value, the more similar the patient’s paracancerous tissue and the cancer tissue, and the greater the risk of future cancer recurrence. Big.

Numerical simulation test results of the MD ratio of cancer recurrence risk

Application of beta-binomial mixture model to generate analog data, the parameter is set to _{N i = 1000, m i =} 1000, p i (0) = 11, q i (0) = 383, where p _i ⁽⁰⁾ and q _i ^{( The value of 0)} is derived from the maximum likelihood value of normal tissue samples. The values of α are 0, 0.001, 0.003, 0.01, 0.003, 0.1, and each group of parameters is repeated 50 times. The numerical simulation results are shown in Table 2 below.

Table 2.

的均值，MSE表示

的均值，Std表示

的标准差，Wald表示Wald检验的中位数，Power表示p-value＜0.05的概率。从表中可以看出极大似然的算法误差非常小；假阳率＝0.06，检验功效＝1，符合预期。 Among them, Bias said

Mean of, MSE means

Mean of, Std means

The standard deviation of, Wald represents the median of Wald test, and Power represents the probability of p-value<0.05. It can be seen from the table that the error of the maximum likelihood algorithm is very small; the false positive rate=0.06, the test power=1, which is in line with expectations.

Example 2. Real sample detection results of the MD ratio of cancer recurrence risk

Using the MD ratio evaluation system as described above, 39 patients with stage IA non-small cell lung cancer (age distribution 40-82 years old; median age 61 years old; 24 men and 15 women) cancer tissue samples and cancer Next to tissue samples for testing. Detect the methylation level of the genomic regions listed in Table 1 in cancer tissues and adjacent tissues. The p-value of the patient is calculated by the above algorithm. According to the test p-value<0.05, the test results are divided into two groups: high risk of recurrence and low risk of recurrence. The results of disease-free survival (DFS) are shown in Figure 1.

It can be seen from the results in Figure 1 that compared with low-risk patients, high-risk patients have a faster relapse, and the difference in DFS between the two groups of patients is statistically significant (p-value=0.039). The above results prove that the MD ratio evaluation system can be used to accurately predict the recurrence risk and subsequent survival of cancer patients.

In order to further verify the prognostic effect of the MD ratio evaluation system on cancer patients, the genomic regions listed in Table 4 and Table 5 in cancer tissues and adjacent tissues (which include 522 and 532 randomly selected from Table 1 respectively) These regions cover the methylation levels of 0.47M and 0.48M in the genome, respectively. The patient's p-value is calculated by the same algorithm, and the test results are divided into high-risk recurrence and low-risk recurrence according to the test p-value<0.05. The disease-free survival (DFS) results are shown in Figure 2 and Figure 3, respectively.

table 3.

Table 4.

In order to compare the effectiveness of the MD ratio evaluation system and traditional tumor driver gene detection in prognosticating cancer patients, the method of somatic mutation detection is used to compare the correlation between the main mutation types EGFR19del and EGFR L858R and DFS with the MD ratio evaluation system The correlation between the test results and DFS is compared, and the results are shown in Table 5. Among them, MD ratio-1 represents the result obtained by detecting the area shown in Table 1 (corresponding to Figure 1); MD ratio-2 and MD-ratio-3 represent the result obtained by detecting the area shown in Table 3 and Table 4, respectively (Corresponding to Figure 2 and Figure 3 respectively).

table 5.

To p-value harzard ratio(HR) EGFR L858R 0.230 0.464 EGFR 19del 0.130 2.442 MD ratio-1 0.039 4.692 MD ratio-2 0.030 3.401 MD ratio-3 0.018 4.453

The above results indicate that the MD ratio assessment system can more effectively assess the patient's cancer recurrence risk and subsequent survival than the somatic test, and play a better role in the patient's prognosis management and clinical treatment.

Claims (10)

A method for prognosing cancer patients, the method comprising: a. Detecting the methylation level in one or more regions of the genome of cancer tissues and paracancerous tissues from the patient by high-throughput sequencing; b. Determine the difference in the methylation level between the cancer tissue and the adjacent tissue; and c. Use the difference information of the methylation level to predict the patient's prognosis through mathematical modeling, Wherein, the smaller the difference of the methylation level between the cancer tissue and the adjacent tissues is, the worse the prognosis of the patient is. The method of claim 1, comprising: a. Detecting the methylation level in one or more regions of the genome of cancer tissues and paracancerous tissues from the patient by high-throughput sequencing; b. Determine the differential methylation block (DMB) in the cancer tissue and adjacent tissues, including: b1. For each CpG site in the detected area, the number of methylated reads covering the site is recorded as M, and the number of unmethylated reads is recorded as U; . b2 adjacent CpG sites are combined and defined as methylated block (MB); M on the i-th MB patients all CpG sites are added, referred to as M _i; the sum of all U , Denoted as U _i ; the total coverage is N _i =M _i +U _i , and the methylation level is β _i =M _i /N _i ; b3. For the i th MB, the level of methylation in cancer tissues referred to as ^{_{β i (T) = M i}} (T) / N i (T), the level of methylation adjacent tissue cancer referred to as beta] _i ^(A) =M _i ^(A) /N _i ^(A) , where the MB with a significant difference between _{β i} ^(T) and β _i ^{(A) is determined as DMB;} c. Prognosis of the patient, including: c1. Introduce a parameter α indicating the degree of similarity between the methylation levels of the paracancerous tissue and the cancer tissue of the patient, and calculate the patient's α value by the following algorithm:

Where 0＜α＜1 and a larger value of α indicates a higher degree of similarity, where f(·), g(·), h(·) respectively represent the conditional distribution of the methylation level of adjacent tissues, cancer The prior distribution of the methylation level of the tissue and the prior distribution of the baseline methylation level,

Represents the parameter family of the prior distribution; c2. Use maximum likelihood estimation algorithm to calculate parameter estimates in the range of 0＜α＜1

And calculate the variance of parameter estimates through the fisher information matrix

c3. Calculate the p value of the null hypothesis α=0 for each patient’s adjacent tissues, Among them, the smaller the p value indicates the worse the prognosis of the patient. The method according to claim 2, which optionally includes any one or more of the following features: (1) Combine CpG sites with a distance of less than 50 bp or a distance of less than 100 bp in step b2; (2) In step b3, set |β _i ^(T) -β _i ^(A) |>σ, and

The MB of is determined as DMB, where σ is a value between 0.05 and 1 and τ is a value greater than 0.1; Preferably, σ=0.1 and τ=0.4; (3) In step c1, the patient's α value is calculated by the following algorithm:

Among them, l(α) is the log-likelihood function of the observation data, p _i ⁽⁰⁾ , q _i ⁽⁰⁾ is the Beta-binomial distribution subject to the baseline methylation level of the i-th MB (p _i ^{(0 )} , The shape parameter in _{q i} ^{(0) );} (4) Calculate the p value by Wald test in step c3; (5) Patients with p<0.05 are identified as having a high risk of recurrence and/or low postoperative survival. The method according to any one of claims 1-3, which optionally includes any one or more of the following features: (1) One or more regions of the genome are regions of the genome with methylation variants in the population of cancer patients; (2) One or more regions of the genome cover at least a 0.3M (megabase) region of the genome, such as a 0.3M-5M region; (3) The method is used to predict the postoperative recurrence risk and/or survival of the cancer patient. The method of any one of claims 1 to 4, wherein the cancer is selected from lung cancer, colorectal cancer, liver cancer, ovarian cancer, pancreatic cancer, gallbladder cancer, gastric cancer, and esophageal cancer; preferably, wherein the cancer Is lung cancer, such as non-small cell lung cancer (NSCLC); preferably, wherein one or more regions of the genome include one or more regions selected from those listed in Table 1; preferably, wherein the cancer tissue and cancer Parasite tissue is the tissue surgically removed from the patient. The method of any one of claims 2-5, wherein using methylation data from normal tissues to establish the baseline methylation level comprises: For the i-th MB, methylated normal tissue data referred to as M _i ⁽⁰⁾ and N _i ^(0), that at a given N _i ⁽⁰⁾ of M _i ⁽⁰⁾ is subject to a shape parameter ( Beta-binomial distribution of p _i ⁽⁰⁾ , q _i ^{(0) ):} M _i ⁽⁰⁾ |N _i ⁽⁰⁾ , β _i ⁽⁰⁾ ～ Binomial(N _i ⁽⁰⁾ , β _i ⁽⁰⁾ ) β _i ⁽⁰⁾ ～Beta(p _i ⁽⁰⁾ , q _i ⁽⁰⁾ ), And use the maximum likelihood algorithm to calculate the parameters p _i ⁽⁰⁾ and q _i ⁽⁰⁾ . A system for prognosing cancer patients, the system comprising: Methylation detection module; and Prognostic analysis module, Wherein, the methylation sequencing module is configured to detect the methylation level in one or more regions of the genome of cancer tissues and paracancerous tissues from the patient through high-throughput sequencing, and the prognostic analysis module is configured For the prognosis of the patient through the following methods: a. Determine the difference in the methylation level between the cancer tissue and the paracancerous tissue; and b. Using the difference information of the methylation level, the prognosis of the patient is performed by mathematical modeling, Wherein, the smaller the difference of the methylation level between the cancer tissue and the adjacent tissues is, the worse the prognosis of the patient is. The system of claim 7, wherein the prognostic analysis module is configured to prognose the patient by the following method: a. Determine the differential methylation block (DMB) in the cancer tissue and the adjacent tissue, including: a1. For each CpG site in the detected area, record the number of methylated reads covering this site as M, and the number of unmethylated reads as U; . a2 adjacent CpG sites are combined and defined as methylated block (MB); M on the i-th MB patients all CpG sites are added, referred to as M _i; the sum of all U , Denoted as U _i ; the total coverage is N _i =M _i +U _i , and the methylation level is β _i =M _i /N _i ; a3. For the i th MB, the level of methylation in cancer tissues referred to as ^{_{β i (T) = M i}} (T) / N i (T), the level of methylation adjacent tissue cancer referred to as beta] _i ^(A) =M _i ^(A) /N _i ^(A) , where the MB with a significant difference between _{β i} ^(T) and β _i ^{(A) is determined as DMB;} b. Prognosis of the patient, including: b1. Introduce a parameter α indicating the degree of similarity between the methylation levels of the paracancerous tissue and the cancer tissue of the patient, and calculate the patient's α value through the following algorithm:

Where 0＜α＜1 and a larger value of α indicates a higher degree of similarity, where f(·), g(·), h(·) respectively represent the conditional distribution of the methylation level of adjacent tissues, cancer The prior distribution of the methylation level of the tissue and the prior distribution of the baseline methylation level,

Represents the parameter family of the prior distribution; b2. Use maximum likelihood estimation algorithm to calculate parameter estimates in the range of 0＜α＜1

And calculate the variance of parameter estimates through the fisher information matrix

b3. Calculate the p value of the null hypothesis α=0 for each patient’s adjacent tissues, Among them, the smaller the p value indicates the worse the prognosis of the patient; Preferably, the prognostic analysis module is configured to combine CpG sites with a distance of less than 50 bp or a distance of less than 100 bp in step a2; Preferably, the prognostic analysis module is configured to set |β _i ^(T) -β _i ^(A) |>σ in step a3, and

The MB of is determined to be DMB, where σ is a value between 0.05 and 1 and τ is a value greater than 0.1; preferably, where σ=0.1 and τ=0.4; preferably, the prognostic analysis module is configured to be in step b1 The α value of the patient is calculated by the following algorithm:

Among them, l(α) is the log-likelihood function of the observation data, p _i ⁽⁰⁾ , q _i ⁽⁰⁾ is the Beta-binomial distribution subject to the baseline methylation level of the i-th MB (p _i ^{(0 )} , _{the shape parameter in q i} ⁽⁰⁾ ); preferably, the prognostic analysis module is further configured to use methylation data from normal tissues to establish the baseline methylation level, including: For the i-th MB, methylated normal tissue data referred to as M _i ⁽⁰⁾ and N _i ^(0), that at a given N _i ⁽⁰⁾ of M _i ⁽⁰⁾ is subject to a shape parameter ( Beta-binomial distribution of p _i ⁽⁰⁾ , q _i ^{(0) ):} M _i ⁽⁰⁾ |N _i ⁽⁰⁾ , β _i ⁽⁰⁾ ～ Binomial(N _i ⁽⁰⁾ , β _i ⁽⁰⁾ ) β _i ⁽⁰⁾ ～Beta(p _i ⁽⁰⁾ , q _i ⁽⁰⁾ ), The maximum likelihood algorithm is used to calculate the parameters p _i ⁽⁰⁾ and q _i ⁽⁰⁾ ; preferably, the prognostic analysis module is configured to calculate the p value through Wald test in step b3; preferably, the The system is used to predict the postoperative recurrence risk and/or survival of the cancer patient; preferably, patients with p<0.05 are identified as having a high recurrence risk and/or low postoperative survival. A device for prognosing cancer patients, which includes: Memory for storing computer program instructions; and A processor for executing computer program instructions, Wherein when the computer program instructions are executed by the processor, the device executes the method according to any one of claims 1-6. A computer readable medium storing computer program instructions, wherein when the computer program instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images