Method for detecting copy number variation of sample in alveolar lavage fluid, system for diagnosing pneumonia-type lung cancer and computer readable medium
By performing metagenomic filtration and CNV detection of human sequences in alveolar lavage fluid, an auxiliary diagnosis system for pneumonia lung cancer is constructed, which solves the problem of difficult to identify pneumonia lung cancer patients in the prior art, and achieves high sensitivity and specificity identification of sexual lung cancer.
Patent Information
- Application Number
- CN202411624705.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-14
- Publication Date
- 2025-02-07
AI Technical Summary
The prior art is difficult to detect alveolar lavage fluid samples through mNGS and effectively identify patients with pneumonia lung cancer.
By performing metagenomic filtration of human sequences in alveolar lavage fluid and combining CNV detection, an auxiliary diagnosis system for pneumonia lung cancer is constructed, and an mNGS sequencing, comparison module, copy number variation calculation module and classification model construction module are used for identification.
The auxiliary identification of patients with pneumonia lung cancer was achieved. The sensitivity of the classifier model was 71.8%, the specificity was 84.1%, and the AUC was 0.824, which had good clinical application value.
Smart Images
Figure CN119391856A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of molecular biomedical technology, and in particular to a method for detecting copy number variation of samples in alveolar lavage fluid, a diagnosis system for pneumonia-type lung cancer, and a computer-readable medium. Background Art
[0002] Lung cancer is the leading cause of cancer morbidity and mortality worldwide, and lung adenocarcinoma is the main subtype of lung cancer, accounting for more than 50%. The staging of lung adenocarcinoma is crucial to prognosis, and the earlier the staging, the higher the possibility of cure. Lung adenocarcinoma has its own unique characteristics in terms of clinical, imaging and pathological features, diagnostic methods, treatment options and prognosis. Among the various subtypes of lung adenocarcinoma, invasive mucinous adenocarcinoma (IMA) is a rare and unique type with a lower prevalence than invasive non-mucinous adenocarcinoma (INMA), accounting for approximately 4-20% of non-small cell lung cancer. This subtype has a significantly worse prognosis and is more likely to metastasize to the lungs.
[0003] Invasive mucinous adenocarcinoma, also known as pneumonia-type lung cancer, often has radiological manifestations similar to pneumonia. Patients mainly present with cough and sputum, occasionally accompanied by fever. After anti-infection treatment, the lesions in the lungs fail to be significantly absorbed, and in severe cases can lead to respiratory failure, and ultimately rely on pathological diagnosis to confirm it as lung adenocarcinoma. This type of lung adenocarcinoma is often delayed in diagnosis in the early stages of the disease because it is difficult to distinguish it from lung infection, and its pathological subtype is mostly invasive mucinous adenocarcinoma. Usually, such patients are already in the advanced stage when diagnosed, and common lung adenocarcinoma driver genes such as EGFR or ALK are often undetectable, constituting a special type of lung cancer. Due to the overlap of clinical symptoms and radiological features, IMA in the lungs is often misdiagnosed as infectious pneumonia, which leads to delayed treatment and increases the risk of death for patients.
[0004] In recent years, the application of mNGS in multiple clinical fields has gradually increased, including bloodstream infection, lower respiratory tract infection, central nervous system infection and orthopedic infection. In the detection of pathogens of lower respiratory tract infection, the sensitivity of mNGS is as high as 93.5%, which can help 80.7% of clinical patients adjust their treatment plans. In addition, mNGS can also provide targeted guidance for the use of drugs for cancer patients. 77.4% of patients with tumor infection have shown positive effects after receiving adjusted treatment. However, there are currently no reports on the application of mNGS in the identification of pneumonia-type lung cancer.
[0005] For the analysis of various body fluids, the diagnosis of malignant tumors usually relies on cytological analysis as the gold standard. Some studies have used chromosomal instability to identify cancer and applied it to non-invasive prenatal testing, making chromosomal CNV (copy number variation) detection a new method for diagnosing malignant tumors. However, when using mNGS to identify pathogens, for patients with potential malignancies and no infectious pathogens, mNGS may produce negative results, or the preliminary judgment result may be negative by analyzing the CNV of human sequences, which may lead to problems such as misdiagnosis or missed diagnosis. Summary of the invention
[0006] The technical problem to be solved by the present invention is that when mNGS testing is performed on bronchoalveolar lavage fluid samples of clinical patients, patients with pneumonia-type lung cancer cannot be effectively identified at the same time.
[0007] The present application solves the problems raised in the above-mentioned background technology by analyzing alveolar lavage fluid containing a high proportion of human sequences and performing CNV detection on human sequences filtered out by metagenome, thereby achieving auxiliary identification of patients with pneumonia-type lung cancer.
[0008] To achieve the above object, the present invention provides the following technical solutions: A method for detecting copy number variation in a bronchoalveolar lavage fluid sample comprises the following steps: S1. Extract nucleic acid from bronchoalveolar lavage fluid samples of patients and perform quality control; S2. For nucleic acids of qualified quality, add sequencing adapters to construct libraries and perform quality control again; S3, the library that has passed quality control is subjected to mNGS sequencing; S4. After the data is downloaded, it is compared with the human genome database to obtain the number of reads of the human sequence, divide the genome into multiple windows, and calculate the Zscore value of the reads in each window as the result of the nucleic acid copy number variation detection in the sample.
[0009] The human genome database is the hg19 version.
[0010] The window size is 0.5-2M.
[0011] The Zscore value in each window is calculated by the following formula: Z = (x-μ) / σ; x is the Log2 Ratio value of a single exon, μ is the mean Log2 Ratio of the baseline samples, and σ is the standard deviation of the baseline samples.
[0012] A diagnosis system for pneumonia-type lung cancer, comprising: The mNGS sequencing module is used to perform mNGS sequencing on the sequencing library constructed by extracting nucleic acids from the patient's bronchoalveolar lavage fluid sample to obtain sequencing data; The comparison module is used to compare the offline data with the human genome database to obtain the number of reads of the human sequence; The copy number variation calculation module is used to divide the genome into multiple windows and calculate the Zscore value of the reads in each window as the nucleic acid copy number variation detection result in the sample; The classification model building module is used to use the nucleic acid copy number variation detection results as the input features of the model, build a classifier model for samples of pneumonia-type lung cancer and pneumonia patients, and classify the samples to be tested.
[0013] The classifier model is a stacking model.
[0014] The stacking model combines Deep Learning (DL), Extreme Gradient Boosting (XGBoost), Gradient Boosting Machine (GBM), Generalize Linear Model (GLM) and Random Forest (RF).
[0015] A computer readable medium comprising a computer program for executing the following method: Obtain offline data from mNGS sequencing of a sequencing library constructed by extracting nucleic acids from bronchoalveolar lavage fluid samples of patients; Compare the offline data with the human genome database to obtain the number of reads of human sequences; The genome is divided into multiple windows, and the Zscore value of the reads in each window is calculated as the result of the nucleic acid copy number variation detection in the sample; The nucleic acid copy number variation detection results are used as the input features of the model to build a classifier model for pneumonia-type lung cancer and non-cancer samples, and classify the samples to be tested.
[0016] Compared with the prior art, the present invention has the following beneficial effects: The method for identifying pneumonia-type lung cancer of the present invention performs mNGS sequencing on alveolar lavage fluid and distinguishes using the CNV-GLM model, and the sensitivity of the model is 71.8%, the specificity is 84.1%, and the AUC is 0.824. It can provide a reference for early judgment of tumors, and can perform early differential diagnosis and efficacy evaluation, and has good clinical application value. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 Schematic diagram of the analytical workflow for constructing mNGS for identification of pneumonic lung cancer Figure 2 Abnormal results of mNGS-CNV analysis DETAILED DESCRIPTION
[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Example
[0019] 1. Collection of clinical samples and information analysis A total of 1053 bronchoalveolar lavage fluid specimens were included, of which 359 were from tumor patients and 694 were from non-tumor patients according to clinical diagnosis. The samples were divided into a test set and a training set at a ratio of 8:2. The model was trained using the cross-validation method, and the average value of each time was taken as the model performance parameter.
[0020] 2. Performance of mNGS screening for CNVs According to the workflow of mNGS, the specific steps of sample testing are as follows: First, nucleic acids are extracted from the patient's bronchoalveolar lavage fluid sample and quality control is performed. For nucleic acids of qualified quality, the library is constructed after adding sequencing adapters, and quality control is performed again. The library that has passed quality control will be sequenced on the Illumina sequencing platform, and the number of reads generated by sequencing is 20M. After the data is completed and downloaded, the BWA software is used to compare it with the hg19 version of the human genome database to obtain the number of reads of the human sequence. For sequences that cannot be compared to the human genome database, they are compared with the pathogen database to identify pathogenic microorganisms.
[0021] according to Figure 1 The analysis workflow shown in the figure randomly selects 200 bronchoalveolar lavage fluid samples from non-tumor patients as the baseline standard for human genome copy number. The sliding window (bin) is set on the genome sequence, the bin size is set to 1Mb, and the genome sequence is divided into 3113 bins. The copy number features in these bins will be used as input features for subsequent modeling.
[0022] The CNV algorithm is used to analyze the sequencing data to generate a CNV map. The execution process of the algorithm includes the following steps: 1) Count the number of reads in each window and standardize these reads; 2) Determine the average sequencing depth of the sample, implement GC correction, adjust the data to the diploid level, and then calculate the log2 Ratio value; 3) Perform principal component analysis (PCA) on the sample set of non-tumor patients to reduce noise and distinguish the background noise in the sample to be tested from the real CNV signal; 4) Calculate the Z-score value for the sequencing depth of each window based on the Log2 Ratio value after noise reduction.
[0023] The Zscore value of each window is given by: Z = (x-μ) / σ; X is the Log2 Ratio value of a single exon, μ is the mean Log2 Ratio of the baseline samples, and σ is the standard deviation of the baseline samples.
[0024] Since the number of sequencing reads for each sample is only 20M, the 3113 bins cannot be fully covered. Further screening is performed. The standard is that if a bin only appears in less than 25% of the samples, the window is removed, and the remaining windows are retained. The Zscore value of each window is calculated, and finally the remaining 2523 CNV features (Zscore value of each bin) are included. mNGS screening of CNVs in malignant tumor patients found that there are a large number of gains and losses in chromosomes ( Figure 2 ).
[0025] 3. Construction of mNGS-CNV classifier Copy number data (CNV) was used to screen the feature values of malignant and non-malignant groups. Deep learning (DL), extreme gradient boosting (XGBoost), gradient boosting machine (GBM), generalized linear model (GLM) and random forest (RF) were used respectively. The above five algorithms were then integrated through model stacking (Stacked Generalization, also known as Stacking). The classifier model was built in combination with CNV feature values to further evaluate the mNGS detection of CNV for tumor diagnosis prediction, and the prediction ability was evaluated using LOOCV. The input features in the classifier algorithm are the 2523 CNV features obtained from the above screening, and the output values are the classification labels of the tumor group and the non-tumor group. The bootstrap method in statistics was used for analysis in the modeling process, uniform sampling with replacement was performed from the training set, and the performance of each algorithm classifier was constructed, including sensitivity and specificity, as well as the area under the curve (AUC) based on the receiver operating characteristic (ROC) curve. The specific results are shown in Table 1.
[0026]
[0027] 4. Identification of patients with pneumonia-type lung cancer 1,000 bronchoalveolar lavage fluid clinical specimens were collected, and nucleic acid extraction, library construction, machine sequencing and bioinformatics analysis were performed according to the mNGS workflow. The results showed that 15 bronchoalveolar lavage fluid specimens had abnormal CNVs, indicating a high risk of tumors. Follow-up clinical treatment of these 15 patients confirmed that all 15 patients were tumor patients.
[0028] It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be regarded as exemplary and non-limiting from any point of view, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention, and any reference numerals in the claims should not be regarded as limiting the claims involved.
Claims
1. A method for detecting copy number variation in a bronchoalveolar lavage fluid sample, characterized in that: The steps include: S1. Extract nucleic acid from bronchoalveolar lavage fluid samples of patients and perform quality control; S2. For nucleic acids of qualified quality, add sequencing adapters to construct libraries and perform quality control again; S3, the library that has passed quality control is subjected to mNGS sequencing; S4. After the data is downloaded, it is compared with the human genome database to obtain the number of reads of the human sequence, divide the genome into multiple windows, and calculate the Zscore value of the reads in each window as the result of the nucleic acid copy number variation detection in the sample.
2. The method for detecting copy number variation in bronchoalveolar lavage fluid samples according to claim 1, characterized in that: The human genome database is the hg19 version.
3. The method for detecting copy number variation in a bronchoalveolar lavage fluid sample according to claim 1, characterized in that: The window size is 0.5-2M.
4. The method for detecting copy number variation in a bronchoalveolar lavage fluid sample according to claim 1, characterized in that: The Zscore value in each window is calculated by the following formula: Z = (x-μ) / σ; x is the Log2Ratio value of a single exon, μ is the mean Log2Ratio of the baseline sample, and σ is the standard deviation of the baseline sample.
5. A diagnosis system for pneumonia-type lung cancer, characterized in that: include: The mNGS sequencing module is used to perform mNGS sequencing on the sequencing library constructed by extracting nucleic acids from the patient's bronchoalveolar lavage fluid sample to obtain sequencing data; The comparison module is used to compare the offline data with the human genome database to obtain the number of reads of the human sequence; The copy number variation calculation module is used to divide the genome into multiple windows and calculate the Zscore value of the reads in each window as the nucleic acid copy number variation detection result in the sample; The classification model building module is used to use the nucleic acid copy number variation detection results as the input features of the model, build a classifier model for samples of pneumonia-type lung cancer and pneumonia patients, and classify the samples to be tested.
6. The diagnostic system according to claim 5, characterized in that The classifier model is a stacked model.
7. The diagnostic system according to claim 5, characterized in that The stacked model combines deep learning, extreme gradient boosting, gradient boosting machines, generalized linear models, and random forests.
8. A computer readable medium, characterized in that: It includes a computer program for executing the following method: Obtain offline data from mNGS sequencing of a sequencing library constructed by extracting nucleic acids from bronchoalveolar lavage fluid samples of patients; Compare the offline data with the human genome database to obtain the number of reads of human sequences; The genome is divided into multiple windows, and the Zscore value of the reads in each window is calculated as the result of the nucleic acid copy number variation detection in the sample; The nucleic acid copy number variation detection results are used as the input features of the model to build a classifier model for pneumonia-type lung cancer and non-cancer samples, and classify the samples to be tested.