CN117316278A - Cancer noninvasive early screening method and system based on cfDNA fragment length distribution characteristics - Google Patents

Abstract

The invention discloses a non-invasive early screening method and system for cancer based on the length distribution characteristics of cfDNA fragments. This method uses lower-depth whole-gene sequencing to count the differences in fragment length distribution characteristics of tumor-derived cfDNA and healthy individual-derived cfDNA to establish an early cancer screening model and achieve non-invasive early screening of cancer. The solution of the present invention focuses on the characteristic of blood cfDNA fragment size to distinguish ctDNA from non-tumor-derived cfDNA, does not rely on mutation detection of oncogenes or tumor suppressor genes, and eliminates interference caused by clonal hematopoietic mutations; secondly, embodiments of the present invention The data shows that the size characteristics of blood cfDNA fragments can be used to distinguish healthy people from early-stage tumor patients; finally, due to the use of low-depth whole-gene sequencing technology, the solution of the present invention involves a significant reduction in detection costs, which is beneficial to future applications in the field of early screening of malignant tumors.

Description

A non-invasive early screening method for cancer based on cfDNA fragment length distribution characteristics and system

Technical field

The invention belongs to the field of medical detection technology, and specifically relates to a non-invasive early screening method and system for cancer based on the length distribution characteristics of cfDNA fragments.

Background technique

Most of the mortality from human cancer worldwide is due to late diagnosis that results in poor therapeutic intervention, so early diagnosis of tumors is particularly important. Traditional biomarker imaging techniques play an important role in tumor diagnosis; however, the specificity of traditional serum biomarkers is not satisfactory for treatment guidance. Furthermore, imaging technology cannot be used for "real-time" detection due to radiation exposure and economic concerns. From the FDA's approval of the first plasma cfDNA (cell free DNA) liquid biopsy product based on EGFR gene mutations in 2016 to bTMB (blood tumor mutation burden) being proven to predict the effectiveness of immunotherapy, liquid biopsy has shone in the field of cancer treatment. . Liquid biopsies are increasingly being considered for early tumor diagnosis, treatment guidance, and recurrence monitoring. It can provide information about the tumor. Additionally, liquid biopsies provide a non-invasive alternative to traditional "solid biopsies" that cannot be performed consistently in certain circumstances or in "real time." Despite these numerous advantages, there are several limitations, such as a lack of consensus on detection methods, difficulty in analyzing large amounts of sequencing information, and insufficient evidence-based medicine.

Currently, the conventional method for studying liquid biopsy and early cancer screening is to identify cfDNA released by tumors through mutation detection of oncogenes or tumor suppressor genes. Unfortunately, the amount of circulating tumor DNA (ctDNA) molecules in the blood is often much lower than the amount of non-cancer-related DNA fragments in the blood, making their detection very difficult, especially in the early stages of cancer. Previous studies have found that accurate tumor information can only be obtained when the abundance of ctDNA in cfDNA is 10% or more. However, except for some advanced tumors that release large amounts of ctDNA, the abundance of ctDNA in most tumor patients does not meet this standard. At present, the sensitivity and accuracy of ctDNA detection are mainly improved by increasing the sequencing depth. However, increasing the sequencing depth may lead to false positives, because non-tumor-derived DNA may also contain various tumor-related mutations. High-depth sequencing is also extremely expensive and cannot be widely used clinically. Moreover, mutation research is only targeted at late-stage patients whose cancer has metastasized. Literature Razavi, P., Li, B.T., Brown, D.N. et al. High intensity sequencing reveals the sources of plasma circulating cell-free DNA variants. Nat Med 25, 1928–1937 (2019). Pedram et al. analyzed 124 metastases The peripheral blood of cancer patients and 47 healthy people underwent ultra-deep detection of 508 specific genes, more than 60,000 times, and found a large number of cfDNA mutations, 81.6% from healthy people and 53.2% from cancer patients, all of which are white blood cells. Caused by clonal hematopoiesis rather than tumor cells. This will lead to a high misdiagnosis rate in identifying cancer patients based on mutations in specific target genes. And because cancer patients are very different, after defining special target genes, some special cancer patients may be missed. Therefore, there are certain limitations in specific panel mutation detection of cfDNA. These problems have always limited the application of liquid biopsy.

Another discovery by scientists may solve this problem. Previous studies have shown that the length of cfDNA released into the blood by different cells will vary. For example, the length of many cfDNA is 167bp (similar to the DNA length of a nucleosome), which may be half the length of cfDNA released into the blood during apoptosis. Related to caspase-dependent DNA cleavage. The baby's cfDNA fragment is significantly shorter than the mother's cfDNA fragment, and this feature is used in prenatal diagnosis. These studies indicate that cfDNA derived from different cells exhibits unique patterns in length, which may serve as signals to differentiate cfDNA sources. For a variety of reasons, cancer genomes become disorganized in the way they are packaged. This means that when cancer cells die, they release their DNA into the blood in a disorganized manner, resulting in differences in the length distribution characteristics of cfDNA fragments that are likely to be present in the circulation. Tumor cell DNA and non-tumor DNA cells. Research reports have found that cfDNA fragment length can be used as a feature to distinguish tumor cells from non-tumor cells, which can make up for the low sensitivity and false positives in detecting cfDNA mutations. However, this hypothesis has been contradicted by a few previous studies in this area. The conflicting results have stalled research in this area, so scientific and systematic research on cfDNA fragmentation systems urgently needs to be improved and optimized.

At present, there are a few reports on the use of cfDNA fragment size in research on liquid biopsy and early cancer screening. Among them, the two studies most similar to the study of this invention (PMID: 30404863 and PMID: 31142840):

Literature Mouliere, F., et al., Enhanced detection of circulating tumor DNA byfragment size analysis. Sci Transl Med, 2018.10(466). In Mouliere et al., Mouliere et al. collected 344 plasma samples (including 18 different types) from 200 cancer patients. type of cancer) and 65 healthy people's plasma samples were subjected to lower depth (0.4×) whole-genome sequencing to detect the fragment length characteristics of cfDNA. Analysis found that ctDNA fragments with cancer mutations are generally 20-40 bp shorter than nucleosomal DNA fragments (167 bp), and are enriched in the 90-150 bp range. Researchers then increased the abundance of ctDNA by enriching short fragments and selective sequencing. , using the fragment ratios of different length intervals as features to distinguish tumor blood samples and healthy blood samples through machine learning algorithms. However, because this method selectively enriches fragments, other length fragments also contain effective information such as tumor mutations, and only taking 90-150 bp fragments will lose some information. In addition, the study used a combination of the total proportion of fragment distribution and mutation characteristics to train the model. There was a lack of systematic consideration of the fragment size proportion information of specific functional sites in the entire genome. Therefore, this method uses cfDNA fragment characteristics as an early diagnosis of cancer. There is still a lot of room for improvement in the indicators.

In the literature Cristiano, S., et al., Genome-wide cell-free DNA fragmentation inpatients with cancer. Nature, 2019.570(7761): p.385-389. In the literature, Cristiano et al. detected 208 patients with cancer through whole-genome sequencing. Blood samples from patients with different stages of breast, colorectal, lung, ovarian, pancreatic, gastric and bile duct cancer and 215 healthy people found circulating tumor DNA in 57% to more than 99% of the patients. This method is based on lower coverage whole-genome sequencing methods to detect cfDNA. The number of short and long cfDNA sequences mapped to different regions of the genome was analyzed in non-overlapping windows covering the genome to build a model for predicting early cancer types. Moreover, most of the patients enrolled in the above-mentioned research institutes are late-stage cancer patients with limited cancer types, and the fragment size statistics adopt a "one-size-fits-all" standard for multiple cancer types (short fragments are defined as 100 to 150 bp, and long fragments are defined as 151 bp). to 220 bp), there is a lack of more in-depth large-sample statistics and research on determining the optimization parameters of cancer-specific fragments for single cancer types.

Based on this, designing a non-invasive cancer early screening method that can achieve low-depth whole-genome sequencing can predict early-stage cancer patients, and can significantly reduce the cost of early cancer screening and improve the screening accuracy. This is a great addition to the technology in this field. It is very necessary for personnel.

Contents of the invention

The purpose of the present invention is to provide a non-invasive early screening method for cancer based on the length distribution characteristics of cfDNA fragments. It mainly solves the technical problems of low specificity and high false positive rate in early screening of bile and pancreatic malignant tumors in the existing technology.

The technical solutions adopted by the present invention to solve the above technical problems are as follows:

A non-invasive early screening method for cancer based on the fragment length distribution characteristics of cfDNA. The method is to use lower-depth whole-gene sequencing to count the difference in fragment length distribution characteristics of tumor-derived cfDNA and healthy individual-derived cfDNA to establish early screening of cancer. Check model to achieve non-invasive early screening of cancer. The lower-depth whole-gene sequencing method refers to a sequencing depth of 2X to 4X.

As a preferred embodiment, the normalized z-score of the number of short fragments and total fragments in 504 5Mb length regions of the whole genome is calculated as the feature input value for model training. The total number of fragments is the sum of the defined number of long fragments and the number of short fragments.

The cfDNA fragments include short fragments and long fragments, where the length range of the short cfDNA fragment is [130,177] bp and the length range of the long fragment is [177,237] bp.

As a preferred embodiment, the LinearSVC algorithm is used, and the 5-fold cross-validation method is repeated 30 times to obtain the model coefficients and establish an early cancer screening model.

As a preferred embodiment, the cancer is a malignant tumor of the bile and pancreas. The biliopancreatic malignant tumors include pancreatic cancer, gallbladder cancer and cholangiocarcinoma.

The present invention also provides a non-invasive early screening system for cancer based on the length distribution characteristics of cfDNA fragments. The system includes:

cfDNA fragment feature extraction module, used to obtain cfDNA fragment size feature data in the sample;

The machine learning classification model building module is used to establish an early cancer screening model based on the statistical difference in size characteristics of cfDNA derived from tumors and cfDNA derived from healthy individuals;

The independent validation queue evaluation module is used to verify the prediction performance of the established machine learning classification model through the independent validation queue.

As a preferred embodiment, the cfDNA fragment feature extraction module includes:

The sequencing data comparison unit is used to compare the sequencing data to the human reference genome hg19 after removing the sequencing adapters from the sequencing data;

The cfDNA fragment statistics unit is used to count cfDNA fragment length data information; the hg19 autosomal chromosome is divided into 504 adjacent, non-intersecting window fragments, each window fragment length is 5Mb; the statistical length in each window area is greater than 130bp and less than The ratio of the number of 177bp cfDNA to the number of cfDNA with a length greater than 177bp and less than 237bp; finally, the number of long and short cfDNA fragments in each 5Mb interval is obtained;

The cfDNA fragment feature determination unit is used to determine the interval with the largest difference in fragment distribution between cancer patients and healthy controls based on the difference in fragment distribution between cancer patients and healthy controls; define short fragment ranges [130,177] and long fragment ranges [177,237] , and then calculate the standardized z-score of the short fragment cfDNA and the total number of fragments in each of the 504 windows as the feature input value for model training.

As a preferred embodiment, the machine learning classification model building module includes:

The sample data classification unit is used to divide the samples into training sets and test sets in a ratio of 4:1, and to keep the distribution proportions of healthy controls and various cancer samples in the two sets consistent;

The model parameter acquisition unit is used to process the sample data in the training set; in the training queue, the model parameters are obtained using the 5-fold cross-validation method repeated 30 times;

The model performance evaluation unit is used to draw the receiver operating characteristic curve of the training queue based on the model prediction value and pathological detection results of each sample in the training queue.

Compared with the prior art, the beneficial effects of the present invention are as follows:

The solution of the present invention first focuses on the characteristic of blood cfDNA fragment size to distinguish ctDNA from non-tumor-derived cfDNA, and does not rely on mutation detection of oncogenes or tumor suppressor genes, which eliminates interference caused by clonal hematopoietic mutations; secondly, the present invention The data of the examples show that the size characteristics of blood cfDNA fragments can be used to distinguish healthy people from early-stage tumor patients, and the ctDNA signal of tumor patients cannot be detected because of the low ctDNA content in early-stage tumor patients; finally, due to the use of low-depth whole-gene Sequencing technology, the solution of the present invention involves a significant reduction in detection costs compared to other full-depth or ultra-depth NGS detection methods. These advantages are conducive to the future application of this solution in the field of early screening of malignant tumors.

The study of this invention uses plasma samples of 60 cases of bile and pancreatic tumors and 31 cases of healthy controls detected in clinical practice to conduct low-depth (2X~4X) whole-genome detection of cfDNA. Considering the position distribution of cfDNA fragment size in the entire genome, it is important to distinguish tumor cells. Establish an analysis system for DNA factors of non-tumor cells. Through systematic statistical analysis of the size and quantity of DNA fragments covering different regions of the whole genome in blood, we trained and tested the cfDNA fragment size characteristics in the research cohort, which can be used to establish the early detection of bile and pancreatic malignant tumors. Screening biomarker diagnostic model. Furthermore, the independent verification team of this study included 94 patients with bile and pancreatic tumors and 40 healthy people, and successfully verified the efficiency of the diagnostic model based on the length distribution characteristics of blood free DNA fragments. This method uses a more precise analysis of the length of cfDNA in blood to find clues for early tumor screening, providing more solid and reliable data support for precise clinical applications.

Description of drawings

Figure 1 is a fragment length distribution map of cfDNA at the whole genome level in Example 1 of the present invention.

Figure 2 is a difference distribution diagram of cfDNA fragments between cancer patients and healthy individuals in Example 1 of the present invention.

Figure 3 is a ROC curve chart of the training set in Embodiment 2 of the present invention.

Figure 4 is a ROC curve chart of the test set in Embodiment 2 of the present invention.

Figure 5 is the ROC curve of independent validation cohort subjects in Example 2 of the present invention.

Detailed ways

The technical solution of the present invention will be described in detail below with reference to examples. The reagents and biological materials used below are all commercial products unless otherwise specified.

Example 1

(1)Research cohort and clinical information

This study included a total of 154 patients with biliopancreatic tumors (pancreatic cancer, gallbladder cancer, and bile duct cancer) diagnosed by tumor markers, imaging examinations (such as ultrasound, abdominal CT scan, etc.) and pathological examination results, and 71 healthy controls. Blood samples from patients and healthy controls were collected before surgery. Each enrolled patient was given an accurate diagnosis based on pathological examination results after surgery.

(2) Blood collection, separation and storage

Whole blood from preoperative cancer patients and healthy controls was collected in 10 ml cell-free nucleic acid storage tubes (REF43803, BD, USA) and transported at room temperature. The received whole blood samples were separated into plasma using a two-step centrifugation method. First, separate the plasma and cell components by centrifuging at 4°C and 1600g for 10 minutes. Carefully aspirate the supernatant, taking care not to aspirate to the white blood cell layer. At the same time, record the hemolysis grade of the plasma. Samples with hemolysis grade ≥5 will not be included in subsequent studies. The plasma was then centrifuged again at 16,000g for 15 minutes at 4°C to remove any remaining cells or cell debris. Transfer the supernatant to centrifuge tubes and divide into 1 ml tubes. The separated plasma samples are stored in a -80°C refrigerator.

(3)cfDNA extraction

Take the plasma sample out of the -80°C refrigerator and place it in a water bath. Incubate it statically at 37°C for about 5 minutes. Transfer the plasma to a low-temperature refrigerated centrifuge and centrifuge it at 1600g and 4°C for 10 minutes. Carefully draw the supernatant into a centrifuge tube. Plasma cfDNA was extracted using the QIAamp Circulating Nucleic Acid Kit (55114, Qiagen, Shanghai, China) kit to extract cfDNA from 1 ml of plasma. The specific operating steps were referred to the product instructions. Finally, 30 μl EB was used to elute cfDNA. The total amount of extracted cfDNA was quantified using a Qubit fluorescence quantifier and corresponding reagents (Q32854, Thermo Fisher, USA). The Agilent 2100 bioanalyzer and the corresponding Agilent High Sensitivity DNAKit&Reagents (5067-4626, Agilent, USA) were used to detect the distribution of cfDNA fragments.

(4) cfDNA library construction and WGS sequencing

Use cfDNA quality control qualified samples for cfDNA library construction and WGS sequencing. The library was prepared using the KAPA DNA Hyper Prep kit (KK8504, KAPA, USA). For detailed operating procedures, refer to the product manual. The input amount of each cfDNA sample is 10ng, and then the base ends are filled and A tails are added, and then the adapter is connected, adapter purified, and PCR amplified for 7 cycles to enrich the library. After purification, the DNA is finally eluted with 25 μl of eluent. Qubit measures the concentration of plasma cfDNA library, and 4150 measures the fragment distribution of plasma cfDNA library. Libraries that passed the quality inspection were selected for whole-genome sequencing using the NovoSeq 6000 platform, with a sequencing strategy of 2x150bp and a sequencing volume of ~10G (~3×).

(5) cfDNA fragment size feature extraction

Based on low pass Whole Genome Sequencing (LP-WGS) detection technology, the sequence information of cfDNA in the plasma of patients and healthy controls is obtained. The analysis process of sequencing data is as follows:

1) Sequencing data comparison. After removing the adapters (fastq) from the original fastq data obtained by LP-WGS sequencing, use BWA software (version: 0.7.12-r1039) to align the sequencing data to the human reference genome hg19 (genome download link: ftp://ftp -trace.ncbi.nih.gov/1000genomes/ftp/technical/reference/human_g1k_v37.fasta.gz), remove low-quality sequences and filter out duplicate sequences from the resulting BAM files.

2) Statistics of cfDNA fragment length.

3) Whole-genome fragment size distribution map. Exclude low-coverage regions of the hg19 reference genome and Duke black box regions; then divide the hg19 autosomal chromosome into 504 contiguous, non-intersecting window segments, each window segment is 5Mb in length; statistics are calculated within each window region The ratio of the number of cfDNA with a length greater than 130bp and less than 177bp to the number of cfDNA with a length greater than 177bp and less than 237bp; finally, the number of long and short cfDNA fragments in each 5Mb interval is obtained, and finally the ratio is used to visualize the cfDNA fragment size map of the entire genome, see Figure 1 , is the fragment length distribution map of cfDNA at the whole genome level.

(6) Establishment of machine learning classification model

1) Determine the characteristics of large and small fragments. According to the difference distribution of fragment distribution between cancer patients and healthy controls, determine the interval with the largest difference in fragment distribution between cancer patients and healthy controls. See Figure 2, which is a difference distribution diagram of cfDNA fragments between cancer patients and healthy individuals. Figure 2 The ordinate refers to the difference in frequency of cfDNA fragments between cancer patients and healthy individuals. Define the short fragment range [130, 177] and the long fragment range [177, 237], and then calculate the standardized z-score of the short fragment cfDNA and the total number of fragments in each of the 504 windows as the feature input value for model training.

2) Divide the samples into training sets and test sets. Divide all samples into training sets and test sets at a ratio of 4:1, and keep the distribution ratios of healthy controls and various cancer samples in the two sets consistent.

3) Process the sample data in the training set. In the training queue, use the 5-fold cross-validation method repeated 30 times to obtain the model coefficients.

4) Evaluate the performance of the model. Based on the model prediction value and pathological detection results of each sample in the training set, the receiver operating characteristic curve (ROC curve) of the training set is drawn. Based on the predicted value, a series of thresholds are set to divide the training set into healthy people and cancer patients, and then the pathological test results are used as the true value to evaluate the prediction performance of the model. Model prediction performance evaluation methods, including area under the ROC curve (AUC, Area UnderCurve, value range 0 ~ 1), positive predictive value (PPV, Positive Predictive Value, value range 0 ~ 1), specificity (value range 0 ~1), accuracy (value range 0~1) and sensitivity (value range 0~1), the higher the value, the better the effect.

(7) Verification of classification model prediction efficiency

In the independent validation cohort, the effectiveness of the model in predicting classification is verified based on the classification model and prediction values determined in the training cohort. The process is as follows:

1) Confirm the variables. In the independent validation cohort, the standard z-score of 504 windows of cfDNA short fragments and the number of total fragments in the whole genome was used as the variable.

2) Model effectiveness verification. Based on the molecular marker expression and pathological detection results of each sample in the test set, the ROC curve of the test set is drawn. Based on the predicted value, the independent validation cohort was divided into healthy people (same as training set and test set) and cancer group, and the prediction performance of the model was evaluated, including specificity, sensitivity and accuracy. The higher the value, the better the effect.

Example 2

(1)Research cohort and clinical information

This study included two research cohorts, a total of 154 patients with bile and pancreatic malignant tumors diagnosed by tumor markers, imaging examinations (such as ultrasonic examination, abdominal CT scan, etc.) and pathological examination results, and 71 healthy individuals, who were collected before surgery. Blood samples were collected from patients, while blood samples from healthy controls were collected. A total of 60 patients (29 cases of pancreatic cancer, 15 cases of gallbladder cancer, and 16 cases of cholangiocarcinoma) and 31 healthy individuals were included in the training set and test set (Table 1). Divide the samples into training sets and test sets at a ratio of 4:1, and keep the distribution ratios of healthy controls and various cancer samples in the two sets consistent. Table 1 shows the grouping information of healthy controls and patients in the training set and test set. The analysis results show that there is no significant difference in the gender ratio of the training set and the test set samples, as well as the distribution ratio of the number of healthy controls and cancer patients.

Table 1: Training set and test set information

A total of 94 patients with biliopancreatic tumors (37 cases of pancreatic cancer, 17 cases of gallbladder cancer, and 40 cases of cholangiocarcinoma) and 40 healthy subjects were enrolled in the independent validation cohort. Table 2 shows the participant information in the training set and independent validation cohort. The analysis results show that there is no significant difference in the gender ratio and the number distribution ratio of healthy controls and cancer patients in the training set and independent validation cohort samples.

Table 2: Training set and independent validation cohort information

(2) Health and cancer classification scoring model

Using the training set, combined with the pathological detection results, the LinearSVC algorithm is used to construct a scoring model for cancer patients and healthy people. The model consists of three parts: variables, model formulas and predicted values. The process is as follows:

①Model variables and parameters. In the training queue, the model used the standardized z-score of 1008 feature variables (Table 3) for the number of short fragments and total fragments in 504 5Mb length regions, and used the 5-fold cross-validation method with 30 repetitions to obtain the model coefficients.

Table 3: Example of model input variables

serial number model variables number of segments 1 Bin1 short fragment N1 2 Bin1 all fragments N2 … … … 1007 Bin504 short clip N1007 1008 Bin504 all fragments N1008

②Scoring model. The scoring model formula is as follows:

Among them, x _i is the input variable, and the calculation formula of model parameters w and b is as follows:

Among them, λ is the penalty parameter, n is the number of samples, _yi is the true value of the sample, 1 is cancer, and -1 is health.

Using the classification model and the fragment distribution of different regions at the genome-wide level of each sample, the class prediction results of each sample can be obtained.

③Evaluation of model effectiveness.

In order to build an early screening classification model that distinguishes cancer patients from healthy individuals, the training set samples are divided into healthy individuals and cancer patients using predicted values. Taking the pathological test results as the true value, the ROC curve of the training set is drawn based on the predicted value and pathological results in the queue. The AUC value of the training set reaches 1. See Figure 3, which is the ROC curve of the training set. The PPV (accuracy), specificity and sensitivity predicted by the trained model were 100%, 100% and 100% respectively (Table 4). The results show that in the training set, this risk prediction model has high sensitivity and NPV, and the model has better prediction performance for early diagnosis of cancer.

④Verification of the prediction performance of the discriminant model.

In order to verify the performance of the discriminant model, the test set patients were divided into healthy and cancer patient groups (same as the training set) using the threshold set by the predicted value. Taking the pathological detection results as the true value, the efficiency of the model is verified based on the classification model and prediction value determined in the training queue, and the ROC curve of the test set is drawn, and its AUC value reaches 1. See Figure 4, which is the ROC curve chart of the test set. And evaluate the model prediction performance, including accuracy, specificity and sensitivity, which were 100%, 100% and 92% respectively (Table 4). The results show that: in the test set, this risk prediction model also has high specificity, sensitivity and accuracy, that is, the model prediction performance is better.

Table 4

(3) Independent verification queue verification

In order to further verify the performance of the discriminant model, patients in the independent validation cohort were divided into healthy and cancer patient groups at a threshold set by the predicted value. Taking the pathological test results as the true value, the efficiency of the model was verified based on the classification model and predicted values determined in the training set and test set, and the ROC curve of the independent validation cohort was drawn. The AUC value of the independent validation cohort was as high as 0.94. Figure 5 shows the ROC curve of subjects in the independent validation cohort. And evaluate the model prediction performance, including accuracy, specificity and sensitivity, which were 90.1%, 77.5% and 87.2% respectively (Table 5). The results show that in the independent validation cohort, this risk prediction model also has high specificity, sensitivity and accuracy, that is, the model prediction performance is better.

Table 5: Independent Validation Cohort Model Performance Verification

The above are only some preferred embodiments of the present invention, and the present invention is not limited to the contents of the embodiments. For those skilled in the art, various changes and modifications can be made within the scope of the technical solution of the present invention, and any changes and modifications made are within the protection scope of the present invention.

Claims (9)

1. A non-invasive early screening method for cancer based on the fragment length distribution characteristics of cfDNA. The method is to use lower-depth whole-gene sequencing to collect statistics on the difference in fragment length distribution characteristics of tumor-derived cfDNA and healthy individual-derived cfDNA to establish a diagnosis of cancer. Early screening model enables non-invasive early screening of cancer. 2. The non-invasive early screening method for cancer based on cfDNA fragment length distribution characteristics according to claim 1, characterized in that: calculating the standardized z- score, used as the feature input value for model training. 3. The non-invasive early screening method for cancer based on the length distribution characteristics of cfDNA fragments according to claim 1, characterized in that: the cfDNA fragments include short fragments and long fragments, wherein the length range of the short cfDNA fragment is [130,177] bp, and the long fragment is [130,177] bp. The fragment length range is [177,237]bp. 4. The non-invasive early screening method for cancer based on cfDNA fragment length distribution characteristics according to claim 1, which is characterized by: adopting the LinearSVC algorithm and using 30 times repeated 5-fold cross-validation method to obtain model coefficients and establish early screening of cancer. Model. 5. The non-invasive early screening method for cancer based on cfDNA fragment length distribution characteristics according to any one of claims 1 to 4, characterized in that: the cancer is a malignant tumor of the bile and pancreas. 6. The non-invasive early screening method for cancer based on cfDNA fragment length distribution characteristics according to any one of claims 1 to 4, characterized in that: the bile and pancreatic malignant tumors include pancreatic cancer, gallbladder cancer and cholangiocarcinoma. 7. A non-invasive early screening system for cancer based on cfDNA fragment length distribution characteristics, characterized in that the system includes: cfDNA fragment feature extraction module, used to obtain cfDNA fragment size feature data in the sample; The machine learning classification model building module is used to establish an early cancer screening model based on the statistical difference in size characteristics of cfDNA derived from tumors and cfDNA derived from healthy individuals; The independent validation queue evaluation module is used to verify the prediction performance of the established machine learning classification model through the independent validation queue. 8. The non-invasive early screening system for cancer based on cfDNA fragment length distribution characteristics according to claim 7, wherein the cfDNA fragment feature extraction module includes: The sequencing data comparison unit is used to compare the sequencing data to the human reference genome hg19 after removing the sequencing adapters from the sequencing data; The cfDNA fragment statistics unit is used to count cfDNA fragment length data information; the hg19 autosomal chromosome is divided into 504 adjacent, non-intersecting window fragments, each window fragment length is 5Mb; the statistical length in each window area is greater than 130bp and less than The ratio of the number of 177bp cfDNA to the number of cfDNA with a length greater than 177bp and less than 237bp; finally, the number of long and short cfDNA fragments in each 5Mb interval is obtained; The cfDNA fragment feature determination unit is used to determine the interval with the largest difference in fragment distribution between cancer patients and healthy controls based on the difference in fragment distribution between cancer patients and healthy controls; define short fragment ranges [130,177] and long fragment ranges [177,237] , and then calculate the standardized z-score of the short fragment cfDNA and the total number of fragments in each of the 504 windows as the feature input value for model training. 9. The non-invasive early screening system for cancer based on cfDNA fragment length distribution characteristics according to claim 7, wherein the machine learning classification model establishment module includes: The sample data classification unit is used to divide the samples into training sets and test sets in a ratio of 4:1, and to keep the distribution proportions of healthy controls and various cancer samples in the two sets consistent; The model parameter acquisition unit is used to process the sample data in the training set; in the training queue, the model parameters are obtained using the 5-fold cross-validation method repeated 30 times; The model performance evaluation unit is used to draw the receiver operating characteristic curve of the training queue based on the model prediction value and pathological detection results of each sample in the training queue.