CN115798582A - Model for identifying benign and malignant lung nodules or predicting risk of postoperative recurrence of lung cancer, kit and application - Google Patents

Abstract

The invention discloses a model for identifying benign and malignant lung nodules or predicting risk of postoperative recurrence of lung cancer, a kit and application, and relates to the technical field of biological medical treatment. The invention discovers that the methylation levels of the PTGER4 gene, the RASSF1 gene, the SHOX2 gene and the PCDHGB6 gene, the genomic instability of the chr6_46000000, the chr14_78000000, the chr 80000000 and the chr5_146000000 windows 148000000 and the fragment size distribution of the chr2q, the chr11p and the chr14p can be used as markers, so that the effective identification of the benign and malignant pulmonary nodules and the postoperative monitoring of the recurrence of the lung cancer are realized, and the invention has the advantages of high sensitivity, good specificity, low detection cost and the like.

Description

A model for differentiating benign from malignant pulmonary nodules or predicting the risk of lung cancer recurrence after surgery and kits and applications

technical field

The invention relates to the field of biomedical technology, in particular to a model, kit and application for differentiating benign and malignant pulmonary nodules or predicting the risk of postoperative recurrence of lung cancer.

Background technique

Since the popularization of low-dose CT scans and the higher resolution, more and more pulmonary nodules have been found clinically. In clinical practice, pulmonary nodule resection is not difficult. The real test of the technical level is how to correctly identify properties of pulmonary nodules. A pulmonary nodule is defined radiographically as a well-circumscribed lesion ≤30 mm in diameter completely surrounded by lung parenchyma. Lesions >30 mm in diameter are masses rather than nodules and have a significantly higher probability of malignancy. Incidental pulmonary nodules refer to pulmonary nodules that are discovered incidentally without corresponding signs and symptoms. Occasionally, multiple pulmonary nodules are found, and diagnostic evaluation is directed at the dominant type or the most suspicious nodules (eg, largest or growing nodules). Nodules can be classified into solid or subsolid nodules, and subsolid nodules can be further divided into pure ground-glass nodules (without solid components) and part-solid nodules (with ground-glass and solid components). Common benign pulmonary nodules include hamartomas, granulomas, rheumatoid nodules, arteriovenous malformations, infections (including tuberculosis and fungi), intrapulmonary lymph nodes, and amyloidosis. In summary, the nature of pulmonary nodules is the basis for surgery, which can avoid mistaking benign nodules for malignant resection, and mistaking malignant nodules for benign and delaying. Due to the high mortality rate of lung cancer among malignant tumors, many patients are already diagnosed at an advanced stage. Therefore, patients with pulmonary nodules need to be assessed for their malignant risk, and whether to take further diagnostic and treatment measures based on the level of malignant risk.

To date, low-dose computed tomography (LDCT) has been the main strategy for substantially reducing lung cancer-related mortality over the long term in high-risk asymptomatic populations. Two large randomized controlled trials, the National Lung Screening Trial (NLST) and the Netherlands-Belgium Lung Screening Trial for Lung Cancer (NELSON) have demonstrated that LDCT-based screening can reduce lung cancer-related mortality statistically significantly in high-risk populations20 %above. Low-dose spiral CT can be used in lung cancer screening to find some patients with asymptomatic pulmonary nodules, but it cannot accurately distinguish between benign and malignant pulmonary nodules. Suspected nodules detected by LDCT can be further diagnosed with lung biopsy, including bronchoscopy and percutaneous aspiration. However, peripheral and bronchoscopically invisible lesions remain a diagnostic challenge on lung biopsy. PET-CT is recommended for the identification of solid nodules > 8mm. The sensitivity of PET-CT in diagnosing malignant pulmonary nodules is 72%-94%, but the detection cost is high, and it is not suitable for patients with benign and malignant nodules identification. Histopathological biopsy is the "gold standard" for the diagnosis of malignant tumors. However, tissue biopsy is invasive and relatively complicated to operate, and small tumors may require multiple operations to obtain enough biopsy tissue for large-scale lung cancer Screening is not realistic.

In view of this, the present invention is proposed.

Contents of the invention

The purpose of the present invention is to provide a model, kit and application for differentiating benign from malignant pulmonary nodules or predicting the risk of postoperative recurrence of lung cancer.

The present invention is achieved like this:

In the first aspect, the embodiment of the present invention provides the application of a reagent for detecting a combination of markers in the preparation of a product for distinguishing benign from malignant pulmonary nodules and/or predicting the risk of recurrence of lung cancer after surgery. The combination of markers includes the following Three types of markers: the methylation level of the target gene, the genome instability within the target chromosome window, and the fragment size distribution within the target chromosome arm; wherein, the target genes include: PTGER4 gene, RASSF1 gene, SHOX2 gene and PCDHGB6 At least one of the genes; the target chromosome window includes: at least one of chr6_46000000_48000000, chr14_78000000_80000000 and chr5_146000000_148000000; the target chromosome arm includes: at least one of chr2q, chr11p and chr14p.

In the second aspect, an embodiment of the present invention provides a reagent or kit for distinguishing benign from malignant pulmonary nodules and/or predicting the risk of lung cancer recurrence after surgery, which includes: the combination of detection markers described in the foregoing embodiments reagents.

In a third aspect, an embodiment of the present invention provides a method for training a predictive model of benign and malignant pulmonary nodules and/or predicting the risk of postoperative recurrence of lung cancer, which includes: obtaining each marker in the marker combination in the training sample The detection result of the substance and the corresponding labeling result; wherein, the labeling result is a label representing the benign and malignant of the sample pulmonary nodules and/or predicting the risk of recurrence of lung cancer after surgery, and the combination of markers is the one described in the preceding examples Marker combination; input the detection results of all markers in the marker combination or the scores of the three types of markers into the pre-built prediction model to obtain the prediction results; wherein, the scoring method of each type of markers is obtained as follows : Obtain the detection results of each marker in each class of markers in the training samples and the corresponding labeling results; input the detection results of all markers in each class of markers into the pre-built prediction model, and use the predicted results as each The score of a class of markers; the pre-built prediction model can predict the sample pulmonary nodules according to the detection results of the marker combination, the scores of the three types of markers or the detection results of each marker in each type of markers. A machine learning model for benign and malignant nodes and/or predicting the risk of postoperative recurrence of lung cancer; based on the labeling results and the prediction results, the parameters of the prediction model are updated.

In a fourth aspect, an embodiment of the present invention provides a predictive device for benign and malignant pulmonary nodules and/or for predicting the risk of postoperative recurrence of lung cancer, which includes: an acquisition module and a prediction module. The acquisition module is used to obtain the detection result of each marker in the marker combination described in the foregoing embodiment of the sample to be tested; the prediction module is used to input the detection results of all markers or the scores of the three types of markers into the aforementioned implementation In the prediction model trained by the training method described in the example, the prediction result of the sample to be tested is obtained; the method of obtaining the score of each type of marker is as described in the foregoing embodiment.

In a fifth aspect, an embodiment of the present invention provides an electronic device, which includes: a processor and a memory; the memory is used to store a program, and when the program is executed by the processor, the processor realizes the foregoing The training method described in the embodiment or the benign and malignant pulmonary nodules and/or the prediction method for predicting the risk of postoperative recurrence of lung cancer; the prediction method includes: obtaining the detection result of each marker in the marker combination of the sample to be tested The marker combination is the marker combination described in the preceding examples; the detection results of all markers or the scores of the three types of markers are input into the prediction model trained by the training method described in the preceding examples, and the test results are obtained. The prediction result of the sample; the method of obtaining the score of each type of marker is as described in the foregoing examples.

In a sixth aspect, an embodiment of the present invention provides a computer-readable medium, on which a computer program is stored, and when the computer program is executed by a processor, the training method described in the foregoing embodiment or the foregoing implementation is implemented. The forecasting method described in the example.

The present invention has the following beneficial effects:

The present invention uses the methylation detection data to find the unique genome and epigenetic signals of lung cancer carcinogenesis through integrated learning of the fragment size difference, copy number difference and methylation difference between malignant pulmonary nodules and benign pulmonary nodules. It can be used as a new tumor marker to realize effective differentiation of benign and malignant pulmonary nodules, as well as postoperative recurrence monitoring of lung cancer. It has the advantages of high sensitivity, good specificity, and low detection cost.

Description of drawings

In order to illustrate the technical solutions of the embodiments of the present invention more clearly, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and thus It should be regarded as a limitation on the scope, and those skilled in the art can also obtain other related drawings based on these drawings without creative work.

Figure 1 shows the distribution of LC Score in benign nodules and malignant nodules;

Figure 2 is the ROC used by LC Score to identify benign and malignant pulmonary nodules;

Figure 3 shows the LC Score used to monitor the recurrence risk of lung cancer after surgery;

Fig. 4 is the ROC curve figure of the experimental group 2 of embodiment 4;

Fig. 5 is the ROC curve diagram of the experimental group 3 of the embodiment 4.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Those who do not indicate the specific conditions in the examples are carried out according to the conventional conditions or the conditions suggested by the manufacturer. The reagents or instruments used were not indicated by the manufacturer, and they were all conventional products that could be purchased from the market.

First, the embodiment of the present invention provides the application of reagents for detecting marker combinations in the preparation of products for distinguishing benign and malignant pulmonary nodules and/or predicting the risk of recurrence of lung cancer after surgery. The marker combinations include the following three types Markers: methylation levels of target genes, genomic instability within target chromosome windows, and fragment size distribution within target chromosome arms;

Wherein, the target gene includes: at least one of PTGER4 gene, RASSF1 gene, SHOX2 gene and PCDHGB6 gene;

The target chromosome window includes: at least one of chr6_46000000_48000000, chr14_78000000_80000000 and chr5_146000000_148000000;

The target chromosome arm includes: at least one of chr2q, chr11p and chr14p.

Using methylation detection data, the inventors of the present application found the above-mentioned marker combination through integrated learning of the fragment size difference, copy number difference and methylation difference between malignant pulmonary nodules and benign nodules, and then realized the detection of pulmonary nodules. Compared with other marker combinations, it has better sensitivity, specificity and accuracy in the identification of benign and malignant tumors, as well as in the monitoring of postoperative recurrence of lung cancer.

In the case that the target gene, the target chromosome window and the target chromosome arm have been disclosed, the detection method and calculation method of the methylation level of the target gene, the genomic instability within the target chromosome window and the fragment size distribution within the target chromosome arm can be Acquisition based on existing knowledge.

The "methylation level of the target gene" herein refers to the methylation level of the promoter of the target gene.

The "window" in this article refers to the window (Windows) in biological information analysis, chr6_46000000_48000000 refers to the genome segment between the 46000000th base and the 48000000th base of chromosome 6, the reference genome is hg19, other targets Chromosome window and so on.

The "q" in "chr2q" herein refers to the long arm of the corresponding chromosome, and the "p" in "chr11p and chr14p" refers to the short arm of the corresponding chromosome.

In some embodiments, the method for detecting the methylation level of the target gene comprises: methylation-specific PCR (MS-PCR), bisulfite treatment+sequencing, restriction endonuclease combined with sodium bisulfite Any one of enzyme (COBRA), fluorescence quantification, methylation-sensitive high-resolution melting curve analysis, pyrosequencing, chip-based methylation profiling, and high-throughput sequencing.

In some embodiments, the genomic instability within the target chromosomal window is represented by a genomic instability score within the target chromosomal window.

Optionally, the calculation method of the genome instability score is selected from: any one of the Z value algorithm based on sequencing depth, the log2ratio algorithm based on control samples, the algorithm based on soft-clipped break reads and the BinCount _i algorithm .

Optionally, the calculation formula of BinCount _i is as follows:

其中，BinCount_i为基因组不稳定性评分；Fragment_i为第i个窗口内的读段数目；TotalMappedFragments为该样本总体的读段数目；WindowLength_i为第i个窗口的长度。当窗口长度为2000000bp时，WindowLength_i为2E6。

Among them, BinCount _i is the genome instability score; Fragment _i is the number of reads in the i-th window; TotalMappedFragments is the number of reads in the sample overall; WindowLength _i is the length of the i-th window. When the window length is 2000000bp, WindowLength _i is 2E6.

In some embodiments, the fragment size distribution refers to the ratio or difference of the number of long fragments and short fragments in the target chromosome window, the length of the long fragments is 101-220bp, and the length of the short fragments is 20-100bp .

The "quantity ratio or difference between long fragments and short fragments" herein can be understood as the quantitative ratio of long fragments to short fragments, or the quantitative ratio of short fragments to long fragments, or a numerical value representing the difference between long fragments and short fragments.

In some embodiments, the surgery for lung cancer includes conventional lung cancer surgery, specifically including: any one of lobectomy plus lymph node dissection, wedge resection, segmentectomy, lobectomy, pneumonectomy, and thoracoscopic minimally invasive surgery. One or any combination of several.

In some embodiments, the product is selected from any one of reagents, kits and predictive models.

On the other hand, an embodiment of the present invention provides a reagent or kit for identifying benign and malignant pulmonary nodules and/or predicting the risk of lung cancer recurrence after surgery, which includes: the detection markers described in any of the foregoing embodiments Combined reagents.

On the other hand, an embodiment of the present invention provides a method for training a predictive model of benign and malignant pulmonary nodules and/or predicting the risk of postoperative recurrence of lung cancer, which includes:

Obtain the detection result of each marker in the marker combination in the training sample and the corresponding labeling result; wherein, the labeling result is a label representing the benign and malignant of the sample pulmonary nodules and/or predicting the risk of postoperative recurrence of lung cancer, The marker combination is the marker combination described in any of the foregoing embodiments;

Input the detection results of all markers in the marker combination or the scores of the three types of markers into the pre-built prediction model to obtain the prediction results; wherein, the way to obtain the scores of each type of markers is as follows: Obtain training samples The detection results of each marker in each type of markers and the corresponding labeling results; input the detection results of all markers in each type of markers into the pre-built prediction model, and use the predicted results as each type of markers score; the pre-built prediction model can predict the benign and malignant of the sample pulmonary nodules according to the detection results of the marker combination, the scores of the three types of markers, or the detection results of each marker in each type of markers and/or a machine learning model that predicts the risk of lung cancer recurrence after surgery;

The parameters of the prediction model are updated based on the labeling result and the prediction result.

Optionally, the label is a character or a character string.

The "three types of markers" in this article refer to: the methylation level of the target gene, the genomic instability within the target chromosome window, or the fragment size distribution within the target chromosome arm, and "each type of marker" refers to any of them. A sort of.

The prediction model can be constructed using the detection results of all the markers in the marker combination as input data, or the scores of the three types of markers can be constructed as input data, and has similar predictive effects. In contrast, when the scores of the three types of markers are used as input data, the prediction model obtained by construction is more effective.

In some embodiments, the machine learning model includes: a logistic regression model.

Optionally, the number of training samples may be greater than or equal to any number among 10, 50, 100, 200, 300, 400 and 500.

In some embodiments, the sample to be tested or the training sample is independently selected from: plasma samples, serum samples, whole blood samples, negative standards or positive standards. Optionally, the samples to be tested or the training samples may also be selected from: environmental samples containing at least one of plasma samples or serum samples.

On the other hand, an embodiment of the present invention provides a device for predicting benign and malignant pulmonary nodules and/or predicting the risk of postoperative recurrence of lung cancer, which includes:

An acquisition module, configured to acquire the detection result of each marker in the marker combination of the sample to be tested; wherein, the marker combination is the marker combination described in any of the foregoing embodiments;

The prediction module is used to input the detection results of all markers of the samples to be tested or the detection results of the scores of the three types of markers into the prediction model trained by the training method described in any of the foregoing embodiments, and obtain the prediction results of the samples to be tested The acquisition method of the score of each type of marker is as described in any of the foregoing embodiments.

Optionally, the modules described in the above embodiments can be stored in the memory in the form of software or firmware (Firmware) or solidified in the operating system (Operating System, OS) of the electronic device provided by this application, and can be controlled by the Processor executes. Meanwhile, data necessary for executing the above-mentioned modules, codes of programs, etc. may be stored in the memory.

On the other hand, an embodiment of the present invention provides an electronic device, which includes: a processor and a memory; the memory is used to store a program, and when the program is executed by the processor, the processor realizes the foregoing The training method described in any embodiment or the benign and malignant pulmonary nodules and/or the prediction method for predicting the risk of postoperative recurrence of lung cancer;

The prediction method includes: obtaining the detection result of each marker in the marker combination of the sample to be tested; the marker combination is the marker combination described in any of the foregoing embodiments; The results or the scores of the three types of markers are input into the prediction model trained by the training method described in any of the above-mentioned embodiments to obtain the prediction results of the samples to be tested; the way to obtain the scores of each type of markers is as described in any of the above-mentioned embodiments. .

An electronic device may include a memory, a processor, a bus, and a communication interface, and the memory, the processor, and the communication interface are electrically connected to each other directly or indirectly, so as to realize data transmission or interaction. For example, these components may be electrically connected to each other through one or more buses or signal lines.

Memory can be, but not limited to, random access memory (Random Access Memory, RAM), read-only memory (ReadOnly Memory, ROM), programmable read-only memory (Programmable Read-Only Memory, PROM), erasable read-only memory (Erasable Programmable Read-Only Memory, EPROM), Electric Erasable Programmable Read-Only Memory (EEPROM), etc.

A processor may be an integrated circuit chip with signal processing capabilities. The processor 120 can be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (NetworkProcessor, NP), etc.; it can also be a digital signal processor (Digital Signal Processing, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), field programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

The electronic device may be a server, a cloud platform, a mobile phone, a tablet computer, a notebook computer, an ultra-mobile personal computer (UMPC), a handheld computer, a netbook, a personal digital assistant (personal digital assistant, PDA), a wearable electronic device, etc. Devices, virtual reality devices and other devices, therefore, the embodiments of the present application do not limit the types of electronic devices.

In addition, an embodiment of the present invention provides a computer-readable medium, on which a computer program is stored. When the computer program is executed by a processor, the training method described in any of the foregoing embodiments or any of the foregoing implementations can be implemented. The forecasting method described in the example.

The computer-readable medium may be a general storage medium, such as a removable disk, a hard disk, and the like.

The characteristics and performance of the present invention will be described in further detail below in conjunction with the examples.

Example 1

Lung cancer-specific differentially methylated region (DMR) gene panel (panel) design:

Download the pan-cancer methylation data TCGA Pan-Cancer (PANCAN) from the XENA database, including adrenocortical carcinoma, bladder urothelial carcinoma, breast cancer invasive carcinoma, cervical squamous cell carcinoma and endocervical adenocarcinoma, cholangiocarcinoma, Colon adenocarcinoma, cellular lymphoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, chromophobe renal cell carcinoma, clear cell renal cell carcinoma, papillary renal cell carcinoma, acute myeloid leukemia, Brain low-grade glioma, hepatocellular carcinoma, lung adenocarcinoma, lung squamous cell carcinoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic cancer, pheochromocytoma and paraganglioma, prostate cancer, sarcoma, skin Melanoma, gastric adenocarcinoma, testicular germ cell tumor, thyroid cancer, endometrial cancer, uterine carcinosarcoma, uveal melanoma and other malignant tumors (see Table 1 for the number of samples) and the methylation sites and Its methylation level (β value) data.

Table 1: TCGA sample number statistics

Differential methylation site combination panel of each cancer type and healthy person is each cancer type (adrenocortical carcinoma, bladder urothelial carcinoma, breast cancer invasive carcinoma, cervical squamous cell carcinoma and endocervical adenocarcinoma, cholangiocarcinoma, colon carcinoma Adenocarcinoma, cellular lymphoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, chromophobe renal cell carcinoma, clear cell renal cell carcinoma, papillary renal cell carcinoma, acute myeloid leukemia, brain Low-grade glioma, hepatocellular carcinoma, lung adenocarcinoma, lung squamous cell carcinoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic cancer, pheochromocytoma and paraganglioma, prostate cancer, sarcoma, skin melanoma Tumor, gastric adenocarcinoma, testicular germ cell tumor, thyroid cancer, endometrial cancer, uterine carcinosarcoma, uveal melanoma) and paracancerous healthy control rank sum test (Wilcoxon Rank SumTest) P<0.05 and the multiple of difference>1.2 The union of CpG sites, where the multiple of difference is defined as: the mean value of the methylation level of the CpG site in the positive population/the mean value of the methylation level of the CpG site in the negative population.

Tissue traceability characteristics CpG site combination panel is defined as: CpG sites of adrenocortical carcinoma τ>0.85, CpG sites of bladder urothelial carcinoma τ>0.85, CpG sites of breast cancer invasive cancer τ>0.85, cervical squamous CpG sites of cell carcinoma and endocervical adenocarcinoma τ>0.85, CpG sites of cholangiocarcinoma τ>0.85, CpG sites of colon adenocarcinoma τ>0.85, CpG sites of cell lymphoma τ>0.85, esophageal cancer τ>0.85 CpG sites >0.85, glioblastoma multiforme τ>0.85 CpG sites, head and neck squamous cell carcinoma τ>0.85 CpG sites, renal chromophobe cell carcinoma τ>0.85 CpG sites , CpG sites of renal clear cell carcinoma τ>0.85, renal papillary cell carcinoma τ>0.85 CpG sites, acute myeloid leukemia τ>0.85 CpG sites, brain low-grade glioma τ>0.85 CpG sites , CpG sites of hepatocellular carcinoma τ>0.85, CpG sites of lung adenocarcinoma τ>0.85, CpG sites of lung squamous cell carcinoma τ>0.85, CpG sites of mesothelioma τ>0.85, ovarian serous CpG sites for cystadenocarcinoma τ>0.85, CpG sites for pancreatic cancer τ>0.85, CpG sites for pheochromocytoma and paraganglioma τ>0.85, CpG sites for prostate cancer τ>0.85, sarcoma τ CpG sites of >0.85, CpG sites of skin melanoma τ>0.85, CpG sites of gastric adenocarcinoma τ>0.85, CpG sites of testicular germ cell tumors τ>0.85, CpG sites of thyroid cancer τ>0.85, The union of CpG sites with τ>0.85 in endometrial cancer, CpG sites with τ>0.85 in uterine carcinosarcoma, and CpG sites in uveal melanoma with τ>0.85, this panel can represent the methylation sites that drive cancer point. where τ is defined as:

表示为任意癌种的同类型样本中甲基化水平的极大值。In formula 1, n represents the number of samples of any cancer type. In formula 2, x _i represents the methylation level of any sample of any cancer type,

Expressed as the maximum value of the methylation level in the same type of samples of any cancer type.

Combine the differential methylation sites of cancer types and healthy people and the combination of CpG sites with tissue traceability characteristics of each cancer type, so as to obtain the methylation panel combination that characterizes the general characteristics of pan-cancer types (as shown in Table 2) .

Table 2: Methylated Gene Panel

Methylation library construction and library sequencing:

1. Use 5-30ng cfDNA for methylation library construction, add 50pg internal reference DNA mixture (self-made, 166bp) and use Zymo Research’s Lightning conversion reagent kit to treat the sample with bisulfite and purify and recover.

2. Use the methylation library construction kit of Yisheng Biotechnology for library construction. During the library construction process, DNA purification and recovery use AMPureXP beads (Beckman), and use EB buffer (Qiagen) to elute and collect the library.

3. Take 500ng of the pre-library DNA, and use Twist Bioscience's hybrid capture and washing related reagents to capture the target region DNA (DNA fragments carrying the genes shown in Table 2).

4. Use KAPA HiFi Hotstart ready Mix to amplify the washed DNA, and use AMPure XP beads (Beckman) to purify the amplified product, which is the final library.

5. The final library was quantified using qPCR (KAPA SYBR Fast Kit, Roche), and then performed paired-end 150bp sequencing on the IlluminaNovaSeq 6000 sequencing platform.

Obtain methylation level observations:

The off-machine data of the sequencer was recognized by the bcl2fastq software and converted into a fastq file, and the cutadapt software was used to remove low-quality reads, reads containing more than 5% of N bases, and reads shorter than 50 bp in the sequencing data. The above reads were aligned to the hg19 reference genome using BSMAP software, and then redundant PCR sequences were removed. Use Bismark software to detect the methylation level observations of all CpG sites in the target region (DNA fragments carrying the genes described in Table 2).

Genome Instability Assessment:

Divide the whole genome into windows of equal length with a length of 2M (two million) bases, and exclude X, Y sex chromosomes, mitochondria and other contigs. Use featureCounts software to count the number of reads in any window. For different samples, due to the inconsistent size of the library, it is necessary to perform a homogenization operation within the sample, and introduce the "every million reads" normalization method, specifically:

Among them, Fragment _i represents the number of reads in the i-th window, TotalMappedFragments represents the total number of reads in the sample, and WindowLength _i represents the length of the i-th window, in particular, 2E6.

Fragment size:

Restore the reads of the BAM file of the sample to be tested to the original DNA template (fragment), and divide the whole genome according to the chromosome arm (arm), in particular, including chr1p, chr1q, chr2p, chr2q, chr3p, chr3q, chr4p, chr4q, chr5p, chr5q, chr6p, chr6q, chr7p, chr7q, chr8p, chr8q, chr9p, chr9q, chr10p, chr10q, chr11p, chr11q, chr12p, chr12q, chr13p, chr13q, chr14p, chr14q, chr15p, chr15q, chr16p, chr17p, chr17 chr18p, chr18q, chr19p, chr19q, chr20p, chr20q, chr21p, chr21q, chr22p, chr22q. The number of fragments in the intervals of 20-100bp and 101-220bp in each chromosome arm were counted respectively. For different samples, due to the inconsistent size of the library, it is necessary to perform a homogenization operation within the sample. Specifically, divide the number of fragments in a specific length interval by the overall number of fragments in the chromosome, specifically divide the number of fragments in the range of 20-100bp by The overall number of fragments in the chromosome arm to obtain the proportion of "short fragments (that is, the length is 20-100bp)"; divide the number of fragments in the range of 101 to 220 by the overall number of fragments in the chromosome arm to obtain the "long Fragments (that is, the length is between 101 and 220)", and at the same time calculate the number of fragments with a length of 20-100 bp (ie "short fragments") in each chromosome arm and the number of fragments with a length of 101 to 220 bp in each chromosome arm (that is, The ratio of the number of fragments of the "long fragment") as the fragment size distribution within the chromosome arm.

Multimodal ensemble learning model construction:

1. Feature data extraction: Extract the methylation level, genome instability score, and fragment size generated by the above methods in the sequencing data of each sample as input data, and use the machine learning method LASSO (least absoluteshrinkage and selection operator) regression algorithm to analyze the above The three markers were further dimensionally reduced, and the markers with weights not equal to 0 were selected as the regional combination for model construction;

2. Determination of the optimal parameters of the model: use logistic regression for model construction and iterative training, and determine the optimal threshold through the Youden index method;

3: Model performance verification: Use the determined optimal parameters and optimal thresholds of the model to verify in an independent test set, draw the receiver operating characteristic ROC curve, and calculate the AUC value under the curve. Define LC Score as the final result of the model:

Z＝coef _i ×P _meth +coef _j ×P _bincount +coef _k ×P _fragment -intercept;

Among them, P _meth represents the methylation model score, P _bincount represents the genome instability model score, and P _fragment represents the fragment size model score; coef _i , coef _j , coef _k represent methylation, genome instability in logistic regression, respectively. Stability, the weight of the fragment size distribution; intercept represents the intercept term.

In this embodiment, coef _i , coef _j , and coef _k are respectively: 2.98, 1.59, and 3.27, and intercept is: -5.75. In other embodiments, these parameters may vary based on specific samples of the training set and test set, but will not have a significant impact on the final result.

It should be noted that in this example, the methylation model, genome instability model and fragment size distribution model (models for each type of marker) are constructed in the same way as the final prediction model, all based on the same training The samples are obtained after training the logistic regression model respectively, and the difference is only in the input data.

Example 2

In 52 cases of benign nodules (including 10 cases of hamartoma, 5 cases of bronchial epithelial hyperplasia, 8 cases of fibrous tissue hyperplasia with hyalinization, 12 cases of interstitial lung disease, 7 cases of cryptococcal infection, and 10 cases of fungal infection) plasma Candidate biomarker combinations were screened in samples and plasma samples of 76 lung cancer patients (including 30 cases of stage I lung cancer, 12 cases of stage II lung cancer, 14 cases of stage III lung cancer, and 20 cases of stage IV lung cancer). See Table 7 at the end of the specification for specific information on the samples.

(1) The screening steps of the methylation panel are as follows: extract the methylation levels of all gene promoters in the methylation gene panel shown in Table 2 in Example 1 of each sample as input data, and retain benign nodules and The union of CpG sites with Wilcoxon Rank Sum Test (Wilcoxon Rank Sum Test) P<0.05 and the multiple of difference>1.2, where the multiple of difference is defined as: the mean value of the methylation level of the CpG site in the lung cancer population/benign nodule The mean value of the methylation level of this CpG site in the population. Further, the remaining CpG sites are further dimensionally reduced using a machine learning method called recursive feature elimination (RFE, recursive feature elimination). Based on the current sample, the candidate CpG sites were used to fit the logistic regression model, and the area under the curve (AUC) was evaluated by 5-fold cross-validation. The AUC was sorted from large to small and the top100 genes were retained. The results are shown in Table 3.

Table 3: Diagnostic performance of different methylated genes for benign and malignant pulmonary nodules

After evaluation, the combined top4 gene combination (PTGER4, RASSF1, SHOX2, PCDHGB6) can obtain excellent diagnostic performance, therefore, the methylation panel selects PTGER4, RASSF1, SHOX2, PCDHGB6 genes.

(2) The screening steps of the genome instability panel are as follows: extract the Bincount value of each window (the whole genome is divided into equal-length windows with a length of 2M bases) within the scope of the whole genome of each sample as input data, and keep The union of benign nodules and lung cancer (Wilcoxon Rank Sum Test) windows with P<0.05 and multiples of difference>1.2. Further, the remaining windows are further dimensionally reduced using the machine learning method recursive feature elimination (RFE, recursive feature elimination). Based on the current sample, use the candidate window to fit the logistic regression model, and evaluate the area under the curve (AUC) through 5-fold cross-validation. After sorting the AUC from large to small, keep the top30 windows. The results are shown in Table 4.

Table 4: Diagnostic performance of different windows of genomic instability on benign and malignant pulmonary nodules

window AUC chr6_46000000_48000000 0.714252006 chr14_78000000_80000000 0.712321421 chr5_146000000_148000000 0.710674157 chr19_2000000_4000000 0.610669698 chr12_132000000_133851895 0.609940366 chr7_38000000_40000000 0.60479214 chr19_1_2000000 0.603827247 chr19_48000000_50000000 0.603462182 chr19_16000000_18000000 0.603125 chr3_130000000_132000000 0.603097516 chr19_58000000_59128983 0.60273285 chr7_68000000_70000000 0.602444698 chr16_1_2000000 0.602203301 chr19_10000000_12000000 0.600447683 chr7_24000000_26000000 0.597572858 chr16_2000000_4000000 0.597550913 chr19_44000000_46000000 0.59498906 chr19_12000000_14000000 0.591707066 chr4_1_2000000 0.587697507 chr10_134000000_135534747 0.586424684 chr21_8000000_10000000 0.581465063 chr19_52000000_54000000 0.580675035 chr16_88000000_90000000 0.580552576 chr12_24000000_26000000 0.567310393 chr1_232000000_234000000 0.563031074 chr11_1_2000000 0.56020014 chr19_46000000_48000000 0.554406601 chr9_18000000_20000000 0.551078982 chr3_152000000_154000000 0.544772647 chr8_120000000_122000000 0.539813027

After evaluation, the combination of the TOP3 window (CHR6_46000000_48000000, CHR14_78000000_80000000, CHR5_146000000_148000000) can obtain excellent diagnosis performance. Therefore, the segment size model selects CHR6_46000000000000, CHR14_8000000000, and CHR5146666.

(3) The screening steps of the fragment size distribution panel are as follows: extract the fragment value of any chromosome arm of each sample as input data, and retain benign nodules and lung cancer rank sum test (Wilcoxon Rank Sum Test) P<0.05 and the multiple of difference>1.2 Union of chromosome arms. Further, the remaining windows are further dimensionally reduced using the machine learning method recursive feature elimination (RFE, recursive feature elimination). Based on the current sample, the candidate chromosome arm was used to fit the logistic regression model, and the area under the curve (AUC) was evaluated by 5-fold cross-validation, and the AUC was sorted from large to small. The results are shown in Table 5.

Table 5: Diagnostic performance of fragment sizes of different chromosome arms on benign and malignant pulmonary nodules

After evaluation, the combination of top3 chromosome arms (chr2q, chr11p, chr14p) can obtain excellent diagnostic performance, so the fragment size model selects the chromosome arms of chr2q, chr11p, chr14p.

In summary, the tumor marker panel for the identification of benign and malignant pulmonary nodules combined with fragment size pattern, genomic instability and methylation level is defined as: including PTGER4, RASSF1, SHOX2, PCDHGB6 gene methylation, chr6_46000000_48000000, chr14_78000000_80000000, chr5_146000000_148000000 Logistic regression models for window instability values and size ratios of chr2q, chr11p, chr14p chromosome arm fragments.

Based on the current sample (randomly shuffling the order of 128 samples, using 80% of the 128 samples for training each time, and the remaining 20% as independent verification), for each type of marker, respectively fit a logistic regression model to obtain each type Scoring of markers.

Based on the current sample, the scores of the three types of markers are used as input data to fit the logistic regression model, and the malignancy probability value (hereinafter referred to as LC Score) can be marked for the sample to be tested (see Table 7 for specific information of the sample). In order to reflect the stable performance of the model, parameters such as sensitivity and specificity are the average value of 100 iterative tests. The distribution of malignant probability in benign nodules and lung cancer is shown in Figure 1, and the performance of LC Score for differentiating benign and malignant pulmonary nodules is shown in Figure 2 and Table 6.

Table 6 Confusion matrix of LC Score for differentiating benign from malignant pulmonary nodules

/ malignant nodules (lung cancer) benign nodules Positive Predictive Value/Negative Predictive Value Positive 68 7 90.67% Negative 8 45 84.91% Sensitivity/Specificity 89.47% 86.54% 88.28%

Remarks: The sensitivity is 89.47%, the specificity is 86.54%, and the accuracy is 88.28%.

Example 3

The LC Score model was used to compare 20 pairs of patients with preoperative positive and postoperative recurrence (a total of 40 samples), and 11 pairs of patients with preoperative positive and postoperative recurrence of lung cancer (a total of 22 samples) to evaluate the effect of the LC Score model on lung cancer surgery. Sensitivity of post-relapse risk monitoring. It can be seen from the results that LC Score can accurately distinguish postoperative recurrence and non-recurrence samples, and the evaluation results are shown in Figure 3.

Example 4

Combining the 10 markers of Example 2 (methylation levels of PTGER4 gene, RASSF1 gene, SHOX2 gene and PCDHGB6 gene, genomic instability of chr6_46000000_48000000, chr14_78000000_80000000 and chr5_146000000_148000000 windows, and chr2q and size distribution of chr1p fragments ) as the experimental group 1, and at the same time the experimental groups 2-3:

Experimental group 2 is 3 markers (methylation level of TGER4 gene, RASSF1 gene, SHOX2 gene);

Experimental group 3 is a combination of 15 markers (referring to 10 markers plus chr19_2000000_4000000, chr12_132000000_133851895 window genomic instability and chr1p, chr6p, chr9p fragment size distribution);

The three groups of markers were combined to fit the logistic regression model according to the method provided in Example 2, and the receiver curve analysis was performed on the 128 samples described in Example 2.

The ROC curve of the experimental group 2 is shown in Figure 4, and the AUC is 0.772; the ROC curve of the experimental group is shown in Figure 2, and the AUC is 0.928; the ROC curve of the experimental group 3 is shown in Figure 5, and the AUC is 0.92.

Table 7 Sample information

Table 8 Sample Information

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (10)

1. The application of the reagent for detecting the combination of markers in the preparation of products for distinguishing benign from malignant pulmonary nodules and/or predicting the risk of recurrence of lung cancer after surgery, characterized in that the combination of markers includes the following three types of markers : methylation level of the target gene, genomic instability within the target chromosome window and fragment size distribution within the target chromosome arm; Wherein, the target gene includes: at least one of PTGER4 gene, RASSF1 gene, SHOX2 gene and PCDHGB6 gene; The target chromosome window includes: at least one of chr6_46000000_48000000, chr14_78000000_80000000 and chr5_146000000_148000000; The target chromosome arm includes: at least one of chr2q, chr11p and chr14p. 2. The application according to claim 1, wherein the genome instability in the target chromosome window is reflected by the genome instability score in the target chromosome window; Preferably, the calculation method of the genome instability score is selected from: any one of the Z value algorithm based on sequencing depth, the log2ratio algorithm based on control samples, the algorithm based on soft-clipped break reads and the BinCount _i calculation formula ; Preferably, the calculation formula of BinCount _i is as follows:

Among them, BinCount _i is the genome instability score; Fragment _i is the number of reads in the i-th window; TotalMappedFragments is the number of reads in the sample overall; WindowLength _i is the length of the i-th window. 3. The application according to claim 1, wherein the fragment size distribution refers to the ratio or difference between the number of long fragments and short fragments in the target chromosome window, and the length of the long fragments is 101-220bp, so The length of the short fragment is 20-100bp. 4. The application according to any one of claims 1-3, wherein the product is selected from any one of reagents, kits and predictive models. 5. A reagent or kit for differentiating benign from malignant pulmonary nodules and/or predicting the risk of postoperative recurrence of lung cancer, characterized in that it comprises: the detection marker according to any one of claims 1 to 4 Combined reagents. 6. A method for training a predictive model of benign and malignant pulmonary nodules and/or the risk of postoperative recurrence of lung cancer, characterized in that it comprises: Obtain the detection result of each marker in the marker combination in the training sample and the corresponding labeling result; wherein, the labeling result is a label representing the benign and malignant of the sample pulmonary nodules and/or predicting the risk of postoperative recurrence of lung cancer, The marker combination is the marker combination according to any one of claims 1-4; Input the detection results of all markers in the marker combination or the scores of the three types of markers into the pre-built prediction model to obtain the prediction results; wherein, the way to obtain the scores of each type of markers is as follows: Obtain training samples The detection results of each marker in each type of markers and the corresponding labeling results; input the detection results of all markers in each type of markers into the pre-built prediction model, and use the predicted results as each type of markers score; the pre-built prediction model can predict the benign and malignant of the sample pulmonary nodules according to the detection results of the marker combination, the scores of the three types of markers, or the detection results of each marker in each type of markers and/or a machine learning model that predicts the risk of lung cancer recurrence after surgery; The parameters of the prediction model are updated based on the labeling result and the prediction result. 7. The training method according to claim 6, wherein the machine learning model comprises: a logistic regression model. 8. A prediction device for benign and malignant pulmonary nodules and/or the risk of postoperative recurrence of lung cancer, characterized in that it comprises: An acquisition module, configured to acquire the detection result of each marker in the marker combination of the sample to be tested; wherein the marker combination is the marker combination according to any one of claims 1-4; The prediction module is used to input the detection results of all markers or the scores of the three types of markers into the prediction model trained by the training method described in claim 6 or 7, to obtain the prediction results of the samples to be tested; each type of markers The way to obtain the score is as described in claim 6. 9. An electronic device, characterized in that it comprises: a processor and a memory; the memory is used to store a program, and when the program is executed by the processor, the processor is made to implement claim 6 or 7 The training method or the benign and malignant pulmonary nodules and/or the prediction method for predicting the risk of postoperative recurrence of lung cancer; The prediction method includes: obtaining the detection result of each marker in the marker combination of the sample to be tested; the marker combination is the marker combination described in any one of claims 1 to 4; the detection of all markers The results or the scores of the three types of markers are input into the prediction model trained by the training method described in claim 6 or 7, and the prediction results of the samples to be tested are obtained. The scoring method of each type of markers is as described in claim 6. . 10. A computer-readable medium, characterized in that, a computer program is stored on the computer-readable medium, and when the computer program is executed by a processor, it realizes the training method described in claim 6 or 7 or the training method described in claim 9. the forecasting method described above.

Images