Multi-cancer detection and traceability systems based on DNA high-throughput methylation sequencing
Provided are multi-cancer detection and traceability systems based on DNA high-throughput methylation sequencing. The multi-cancer detection system includes a first sample sequencing module used for sequencing sample DNA, a methylation feature set acquisition module used for comparing methylation levels of sequencing data to obtain a methylation feature set, a terminal motif feature acquisition module used for counting proportions of terminal motif base fragments in all permutations and combinations to obtain a terminal motif feature, a breakpoint motif feature acquisition module used for counting proportions of breakpoint motif base fragments in all permutations and combinations to obtain a breakpoint motif feature, and a model building module used for training a learning classification module by taking the methylation feature set, the terminal motif feature and the breakpoint motif feature as input and whether cancer is suffered from and a probability of the cancer as output to obtain detection and traceability models.
1 . A multi-cancer detection system based on high-throughput methylation sequencing of cell-free deoxyribonucleic acid (DNA), comprising:
a first sample sequencing module used for collecting cancer tissue and pericancerous tissue samples of a target cancer, and plasma samples of a target cancer patient and a healthy person, extracting a first genomic DNA of the cancer tissue and pericancerous tissue samples of the target cancer and a first cell-free DNA of the plasma samples of the target cancer patient and the healthy person, and performing methylation sequencing on the first genomic DNA and the first cell-free DNA respectively to obtain sequencing data of the first genomic DNA and sequencing data of the first cell-free DNA;
a first methylation feature set acquisition module used for performing quality control on the sequencing data of the first genomic DNA and the sequencing data of the first cell-free DNA, the quality control comprising at least one of: ensuring a CGG/TGG ratio greater than 0.6, ensuring a mapping rate greater than 0.4, ensuring a number of reads mapped to a human reference genome is greater than 10 million, ensuring a number of CGI islands covering more than 10× is greater than 10,000, and ensuring a number of CpG sites covering more than 10× is greater than 1 million; and mapping the sequencing data of the first genomic DNA and the sequencing data of the first cell-free DNA respectively to the human reference genome to obtain first coordinates of the sequencing data of the first genomic DNA and the sequencing data of the first cell-free DNA on the human reference genome, wherein the first coordinates comprise a beginning position and an end position of each sequence read in the human reference genome; dividing the human reference genome into a plurality of CpG Island regions according to the first coordinates, and counting methylation levels of the sequencing data of the first genomic DNA about different preset haploid algorithms on the plurality of CpG Island regions; comparing differences in the methylation levels of the sequencing data of the first genomic DNA about different haploid algorithms on the plurality of CpG Island regions to obtain cancer-specific regions of the sequencing data of the first genomic DNA about each of the different haploid algorithms; and counting the methylation levels of the sequencing data of the first genomic DNA about the different preset haploid algorithms on the cancer-specific regions to obtain a methylation feature set of the first cell-free DNA;
a first terminal motif feature acquisition module used for taking all base data of p base pairs of 3′ terminal of the sequencing data of the first cell-free DNA on the human reference genome as a terminal motif base fragment set, and taking a proportion of each base fragment in the terminal motif base fragment set in a permutation and combination of all base sequences of the p base pairs as a terminal motif feature of the first cell-free DNA;
a first breakpoint motif feature acquisition module used for taking all base data of upstream and downstream q base pairs of the 3′ terminal of the sequencing data of the first cell-free DNA on the human reference genome as a breakpoint motif base fragment set, and taking a proportion of each base fragment in the breakpoint motif base fragment set in a permutation and combination of all base sequences of 2*q base pairs as a breakpoint motif feature of the first cell-free DNA; and
a detection model building module used for training a learning classification module by taking the methylation feature set of the first cell-free DNA, the terminal motif feature of the first cell-free DNA, and the breakpoint motif feature of the first cell-free DNA as input vectors, and whether cancer is suffered from as an output vector to obtain a trained multi-cancer detection model which is used for detecting whether a target object suffers from the target cancer.
2 . The multi-cancer detection system based on high-throughput methylation sequencing of cell-free DNA according to claim 1 , wherein the target cancer comprises: lung cancer, gastric cancer, colorectal cancer, liver cancer, breast cancer or thyroid cancer.
3 . The multi-cancer detection system based on high-throughput methylation sequencing of cell-free DNA according to claim 1 , wherein the different preset haploid algorithms comprise Methylation Mean (MM), Methylation Haplotype Load (MHL), Cell Heterogeneity-Adjusted clonal Methylation (CHALM), Partial Methylation Ratio (PDR) and Entropy.
4 . The multi-cancer detection system based on high-throughput methylation sequencing of cell-free DNA according to claim 1 , wherein the first sample sequencing module comprises:
a first sample collection unit used for collecting 10 mL of whole blood samples of the target cancer patient and the healthy person respectively by using cell-free DNA blood collection tubes;
a cell-free DNA extraction unit used for extracting the first cell-free DNA from the whole blood samples after plasma separation by using a plasma DNA extraction kit; and
a cell-free DNA sequencing unit used for building a library and degenerating representative bisulfite sequencing for the first cell-free DNA to obtain the sequencing data of the first cell-free DNA.
5 . The multi-cancer detection system based on high-throughput methylation sequencing of cell-free DNA according to claim 1 , wherein the first sample sequencing module further comprises:
a second sample collection unit used for collecting the cancer tissue and pericancerous tissue samples of the target cancer;
a genomic DNA extraction unit used for extracting the first genomic DNA from the cancer tissue and pericancerous tissue samples of the target cancer by using a genomic DNA extraction kit; and
a genomic DNA sequencing unit used for building a library and degenerating representative bisulfite sequencing for the first genomic DNA to obtain the sequencing data of the first genomic DNA.
6 . The multi-cancer detection system based on high-throughput methylation sequencing of cell-free DNA according to claim 1 , wherein the first methylation feature set acquisition module comprises:
a control group setting unit used for taking samples of each cancer in the target cancer as a positive control group respectively, and taking pericancerous samples corresponding to the target cancer as a negative control group; and
a control group comparison unit used for comparing P values of the methylation levels of the positive control group and the negative control group about the different haploid algorithms after multiple test and correction to obtain the cancer-specific regions of all target cancer samples.
7 . The multi-cancer detection system based on high-throughput methylation sequencing of cell-free DNA according to claim 1 , wherein the learning classification module comprises a first sub-module and a second sub-module connected with the first sub-module, and ensemble algorithms are embedded into the first sub-module; the ensemble algorithms comprise: a logistic regression model algorithm, a support vector machine algorithm, a random forest algorithm, a gradient boosting tree algorithm, a Bayesian model algorithm, a K-nearest neighbor algorithm, an XGBoost algorithm and a CatBoost algorithm; the second sub-module is built-in with a logistic regression model; the ensemble algorithms are used for training the methylation feature set of the first cell-free DNA, the terminal motif feature of the first cell-free DNA, and the breakpoint motif feature of the first cell-free DNA; and the logistic regression model is used for integrating and outputting training results of the ensemble algorithms.
8 . The multi-cancer detection system based on high-throughput methylation sequencing of cell-free DNA according to claim 1 , wherein the p is any integer between 4 and 10, and the q is any integer between 2 and 5.
9 . A multi-cancer traceability system based on high-throughput methylation sequencing of cell-free deoxyribonucleic acid (DNA), comprising:
a second sample sequencing module used for collecting cancer tissue and pericancerous tissue samples of a target cancer, and a plasma sample of a target cancer patient, extracting a second genomic DNA of the cancer tissue and pericancerous tissue samples of the target cancer and a second cell-free DNA of the plasma sample of the target cancer patient, and performing methylation sequencing on the second genomic DNA and the second cell-free DNA respectively to obtain sequencing data of the second genomic DNA and sequencing data of the second cell-free DNA;
a second methylation feature set acquisition module used for mapping the sequencing data of the second genomic DNA and the sequencing data of the second cell-free DNA respectively to a human reference genome after quality control is performed on the sequencing data of the second genomic DNA and the sequencing data of the second cell-free DNA to obtain second coordinates of the sequencing data of the second genomic DNA and the sequencing data of the second cell-free DNA on the human reference genome; dividing the human reference genome into a plurality of CpG Island regions according to the second coordinates, and counting methylation levels of the sequencing data of the second genomic DNA about different haploid algorithms on the plurality of CpG Island regions; comparing differences in the methylation levels of the sequencing data of the second genomic DNA about preset haploid algorithms on the plurality of CpG Island regions to obtain tissue-specific regions of the sequencing data of the second genomic DNA about each of the different haploid algorithms; and counting the methylation levels of the sequencing data of the second genomic DNA about the preset haploid algorithms on the tissue-specific regions to obtain a methylation feature set of the second cell-free DNA;
a second terminal motif feature acquisition module used for taking all base data of p base pairs of 3′ terminal of the sequencing data of the second cell-free DNA on the human reference genome as a terminal motif base fragment set, and taking a proportion of each base fragment in the terminal motif base fragment set in a permutation and combination of all base sequences of the p base pairs as a terminal motif feature of the second cell-free DNA;
a second breakpoint motif feature acquisition module used for taking all base data of upstream and downstream q base pairs of the 3′ terminal of the sequencing data of the second cell-free DNA on the human reference genome as a breakpoint motif base fragment set, and taking a proportion of each base fragment in the breakpoint motif base fragment set in a permutation and combination of all base sequences of 2*q base pairs as a breakpoint motif feature of the second cell-free DNA; and
a traceability model building module used for training a learning classification module by taking the methylation feature set of the second cell-free DNA, the terminal motif feature of the second cell-free DNA, and the breakpoint motif feature of the second cell-free DNA as input vectors, and a probability of suffering from a cancer as an output vector to obtain a trained multi-cancer traceability model which is used for judging types of a detected cancer.