Early-stage prostate cancer diagnosis marker based on circulating free DNA, evaluation system and application

Abstract

The invention relates to a circulating free DNA (cfDNA)-based early-stage prostate cancer diagnosis marker, an evaluation system and application. According to the early-stage prostate cancer diagnosis marker, 25 cfDNA characteristic markers are obtained through fluorescent quantification, nucleic acid protein analysis, cfDNA sequencing and classification model construction, evaluation and screening. The AUC of an ROC curve of an early-stage prostate cancer evaluation model of the cfDNA characteristic marker is as high as 100%, which indicates that the biomarker has a good indication effect in the aspect of diagnosis of early-stage prostate cancer and can be used as a non-invasive biomarker for detection of early-stage prostate cancer. The method is noninvasive, convenient to operate, high in accuracy and capable of effectively distinguishing cfDNA feature differences of prostatic cancer patients and healthy people and efficiently recognizing early prostatic cancer patients.

Description

Technical Field

[0001] The present invention relates to the technical field of tumor screening, and specifically to an early prostate cancer diagnostic marker based on circulating free DNA, an evaluation system and applications. Background Art

[0002] Prostate cancer (PCa) is the most common malignancy in men in Western countries and the second leading cause of cancer-related death. PCa diagnosis is based on histopathology of prostate biopsy, and notably, the implementation of serum prostate-specific antigen (PSA) testing as the initial indication for PCa in the 1990s led to a significant increase in PCa incidence and the number of biopsies performed. Prostate biopsy is associated with a considerable risk of life-threatening sepsis, and a large proportion of initial and repeat biopsies are negative for PCa. Therefore, better minimally invasive or noninvasive PCa biomarkers are needed to replace or complement PSA to ensure earlier and more accurate diagnosis and reduce unnecessary prostate biopsies.

[0003] Most newly diagnosed PCa is localized and usually of relatively low grade. Early diagnosis is crucial for curative treatment, as organ-confined PCa can be cured by radical prostatectomy (RP) or radiotherapy, whereas metastatic prostate cancer (MPC) can only be treated palliatively. However, many patients with relatively indolent early-stage PCa will not benefit from surgery or radiotherapy. In fact, a large number of patients with early-stage PCa may be overdiagnosed and overtreated due to the limited accuracy of currently used clinicopathological prognostic parameters. Emerging molecular biomarkers of PCa aggressiveness have shown great potential to improve risk stratification and may therefore contribute to better and more personalized PCa treatment in the future, but they are not currently implemented in standard clinical practice.

[0004] At present, cfDNA fragment omics features have become an emerging technology for early cancer detection. Its features mainly include fragment size, copy number variation, nucleosome imprinting, whole genome cfDNA fragmentation pattern, and fragment end motif features. In recent years, many studies have shown that cfDNA fragment omics features have potential diagnostic value in various cancer types, such as liver cancer, lung cancer, and colorectal cancer. It is reported that Cristiano et al. developed a whole genome cfDNA fragmentation pattern detection technology based on the ratio of the number of short fragments to long fragments of cfDNA. The technology has an AUC of 0.94, a sensitivity range of 57%-99%, and a specificity of 98% in distinguishing patients with 7 types of cancer from healthy individuals. Subsequently, the team successively reported the application of whole genome cfDNA fragmentation features in lung cancer and liver cancer cohorts. These studies have shown that whole genome cfDNA fragmentation patterns can be used as biomarkers for liver cancer or lung cancer detection. In addition, cfDNA fragment motif features represent another type of biomarker for cancer detection. By analyzing cfDNA fragments from patients with hepatocellular carcinoma and non-cancer patients, researchers revealed the preferred end coordinates of tumor-associated cfDNA, which may be attributed to chromatin accessibility, tumor-specific nucleosome positioning, and nuclease activity. Subsequently, based on the preferential pattern of 4bp (base pair) end motifs exhibited by cancer patients, Jiang et al. constructed diagnostic models using support vector machine (SVM) and logistic regression (LR) algorithms to distinguish cancer from non-cancer patients, with an AUC of 0.89 for both models. Recently, Guo et al. constructed a model based on cfDNA breakpoint motifs that can sensitively detect stage I lung adenocarcinoma with an AUC of 0.985, a sensitivity of 98.0%, and a specificity of 94.7%. These studies further indicate that cfDNA fragment omics features play an important role in cancer detection. Given that detection methods based on single cfDNA features may have insufficient detection capabilities, especially for early-stage cancer patients, recently, multidimensional prediction models combining multiple fragment omics and machine learning methods have improved their detection capabilities for specific cancer types. However, to date, few studies have explored the value of cfDNA fragmentomic characteristics in the early diagnosis of PRAD.

[0005] Patent—CN112086129A introduces a pan-cancer diagnostic model based on cfDNA terminal features and Kmer frequency features. By combining these two biomarkers, the developed diagnostic model achieved a sensitivity and specificity of 0.903 and 0.895, respectively, showing a high degree of accuracy. However, it is worth noting that the model failed to establish a specific diagnostic model for prostate cancer (PRAD), suggesting its limitations in tumor type coverage.

[0006] Patent-CN117373541A proposes a method for identifying the origin of cancer tissue in cfDNA based on the methylation data of plasma cfDNA by calculating the difference interval score values ​​corresponding to different cancer types. This method based on methylation data provides a new perspective for tracing the origin of cancer tissues, and is significantly different from patent CN112086129A in terms of technical route and research diseases, the latter focusing more on the study of prostate cancer (PRAD).

[0007] Patent—CN115376616A demonstrates a multi-classification method based on cfDNA multi-omics data and its implementation device. Using plasma samples from 780 cancer patients, including a variety of common tumor types, the patent achieved the classification of tumor types by establishing a random forest model with an average accuracy of 0.638. Although this accuracy indicator is relatively low and does not include research on prostate cancer (PRAD), it still provides a valuable reference for the application of cfDNA in cancer diagnosis. In summary, the current invention does not have a method for diagnosing prostate cancer based on the terminal features of cfDNA sequencing data. There is no research on using TCGGGG_M6bp, TCGGGG_B6bp, GATGTG_M6bp, GATGTG_B6bp, GTAGTA_B6bp, GTAGTA_M6bp, GAAGTT_M6bp, CGTCGC_B6bp, CGTCGC_M6bp, CGTACC_M6bp, AGCGGA_B6bp, GCTAAT_M6bp, AGCGGA_M6bp, GCTAAT_B6bp, CGACGT_B6bp, CGACGC_M6bp, CGACGT_M6bp, CGCAGT_B6bp, CGCAGT_M6bp, CGACGC_B6bp, CGGGAG_B6bp, CCCCCC_B6bp, CCCCCC_M6bp, CGTACC_B6bp, CGGGAG_M6bp as a combined marker for diagnosing prostate cancer. Based on this, the present invention is specially proposed.

[0008] In view of this, the present invention is proposed. Summary of the invention

[0009] The present invention is achieved through the following technical scheme: a marker suitable for predicting the risk of early prostate cancer, which is based on blood cfDNA sequencing technology and 25 screened cfDNA characteristic markers, and the marker is a blood exosome RNA complex, which can be used as a non-invasive biomarker for early prostate cancer detection. The 25 cfDNA characteristic markers specifically include TCGGGG_M6bp, TCGGGG_B6bp, GATGTG_M6bp, GATGTG_B6bp, GTAGTA_B6bp, GTAGTA_M6bp, GAAGTT_M6bp, CGTCGC_B6bp, CGTCGC_M6bp, CGTACC_M6bp, AGCGGA_B6bp, GCTAAT_M6bp, AGCGGA_M6bp, GCTAAT_B6bp, CGACGT_B6bp, CGACGC_M6bp, CGACGT_M6bp, CGCAGT_B6bp, CGCAGT_M6bp, CGACGC_B6bp, CGGGAG_B6bp, CCCCCC_B6bp, CCCCCC_M6bp, CGTACC_B6bp, and CGGGAG_M6bp.

[0010] In a second aspect, the present invention provides a method for detecting a marker suitable for early prostate cancer risk prediction, which specifically comprises the following steps: S1: Sample collection. Inclusion and exclusion criteria including age, treatment history, pathological diagnosis and functional status; 5 mL of fasting peripheral blood anticoagulated with EDTA was collected from the subjects; S2: Sample pretreatment. After plasma is separated by low-temperature centrifugation, the supernatant is transferred to a new cryotube and stored at -80°C; S3: cfDNA extraction and quantification. Plasma samples were thawed at -80°C, and after lysis, incubation, ice bath, enrichment and washing, the cfDNA concentration was determined using the Qubit dsDNA HS assay kit, and the DNA fragment length was detected using the Bioptic QSEP-100 system; S4: Library construction and quality control, including key steps such as end repair, adapter ligation, PCR amplification and library quality assessment to ensure the quality of library construction. The Qubit4.0 and Bioptic QSEP-100 systems are used to measure library concentration and fragment length; S5: Single-strand circularization and sequencing were performed on the machine, and paired-end sequencing was performed with an average coverage of 1x to obtain high-quality WGS data.

[0011] Wherein, the research sample collection in step S1 is carried out according to the following standards: The study subjects were PRAD patients and healthy controls. The criteria for inclusion in the case group of this study included: (1) age over 18 years old; (2) patients diagnosed for the first time and had not received any tumor-related treatment including immunotherapy, radiotherapy, chemotherapy, anticancer drug therapy, etc.; (3) primary tumors, diagnosed as PRAD by pathological examination, with pathological stage of T1 or T2, no lymph node metastasis and distant metastasis, suitable for surgical resection; (4) normal liver and kidney function; (5) patients had signed informed consent.

[0012] The exclusion criteria for the study included: (1) patients with other malignant tumors, such as lung cancer, liver cancer, or a history of malignant tumors; (2) patients with active infectious diseases, such as active tuberculosis, AIDS, etc.; (3) patients with blood transfusion records in the past 3 months; (4) patients who were unable to understand the study content due to mental illness or other reasons. The inclusion criteria for the healthy control group were: (1) patients over 18 years old; (2) patients in good health; (3) patients who had signed the informed consent form. The exclusion criteria for the healthy control group were: (1) patients with blood transfusion records in the past 3 months; (2) patients who were unable to understand the study content.

[0013] Furthermore, the extraction and quantification of plasma cfDNA in step S3 specifically includes the following steps: (1) Plasma pretreatment: first thaw the frozen plasma sample stored at -80°C in a 37°C water bath. After complete thawing, centrifuge at 16,000 g for 10 min, and then transfer the supernatant to a new centrifuge tube for later use. (2) Lysis process: add 10 μL of proteinase K solution to 500 μL of plasma and vortex for 30 seconds to ensure sufficient mixing; then add 25 μL of 20% SDS and vortex again for 30 seconds. (3) Incubation: Incubate in a 60°C water bath for 30 minutes, flip and mix every 5 minutes to promote uniform reaction; (4) Cooling: After incubation, cool the mixture on ice for 5 minutes; (5) Enrichment: After cooling, place the sample at room temperature to return to room temperature, add 500μL MagMAX cell-free DNA lysis / binding solution and 10μL MagMAX™ cfDNA magnetic beads, vortex on a medium speed oscillator for 10 minutes, and then instantly separate and place on a magnetic stand for 5 minutes until the solution is clear; (6) Washing: Remove the supernatant, add 500μL MagMAX™ cfDNA washing solution and mix thoroughly, then place on the magnetic stand again until the solution is clear, then discard the supernatant; wash twice with 80% ethanol; (7) Elution: After the sample is dried on the magnetic stand at room temperature, add 30μL elution buffer and place on the magnetic stand again for 2 minutes until it is clear, carefully remove the supernatant and store it; (8) Concentration determination: Use Qubit The dsDNA HS detection kit and Qubit4.0 fluorescence quantification instrument were used to measure the cfDNA concentration; (9) Fragment length determination: The Bioptic QSEP-100 system was used to detect the length distribution of cfDNA fragments. Under normal circumstances, the peak value of cfDNA fragments was approximately 167 bp; (10) Storage: The extracted cfDNA was stored at -20°C before library construction.

[0014] Preferably, the step S4 of Chinese library construction and quality control specifically includes the following steps: S41 end repair: The cfDNA extracted from plasma (i.e., input DNA) is blunt-ended, phosphorylated at the 5' end, and a dA tail is added to the 3' end.

[0015] S43 adapter ligation: (1) Dilute the adapter to an appropriate concentration for the amount of input DNA. (2) Thaw the Rapid Ligation buffer, mix thoroughly, and place on ice until needed. (3) Prepare the reaction mixture in the end-repaired PCR tube. (4) Mix by pipetting gently, and centrifuge briefly to concentrate the reaction solution at the bottom of the tube. (5) Place the prepared PCR tube into the PCR instrument, set up and execute the corresponding ligation reaction program. (6) Purify the reaction product using VAHTS DNAClean Beads as follows: A. Bring the beads to room temperature and vortex to mix the VAHTS DNA Clean Beads thoroughly. B. Add 60 μl of VAHTS DNA Clean Beads to 100 μl of the adapter ligation product and mix by vortexing or pipetting 10 times to ensure thorough mixing. C. Incubate at room temperature for 5 minutes. D. Briefly centrifuge the PCR tube and place it on a magnetic rack to separate the beads from the liquid until the solution is clear, which takes about 5 minutes, then carefully remove the supernatant. E. While the PCR tube is still on the magnetic rack, add 200μl of freshly prepared 80% ethanol, incubate at room temperature for 30 seconds, and carefully remove the supernatant. F. Repeat step e once, rinsing twice in total. G. Keep the PCR tube open in the magnetic rack and air dry the beads for 5 to 10 minutes until the ethanol is completely evaporated. H. Remove the PCR tube from the magnetic rack and add 22.5μl of elution buffer (10mM Tris-HCl, pH 8.0 - 8.5) for elution. Vortex or pipette to mix thoroughly, leave at room temperature for 2 minutes, briefly centrifuge again and place it on the magnetic rack. After the solution is clear for about 5 minutes, carefully transfer 20μl of supernatant to a new EP tube, avoiding touching the beads.

[0016] S45 library amplification: (1) After thawing, thoroughly mix PCR Primer Mix 3 and VAHTS HiFi Amplification Mix and prepare the specified mixed reaction system in a sterile PCR tube. (2) Use a pipette to mix gently, avoid shaking, and briefly centrifuge to concentrate the reaction solution at the bottom of the tube. (3) Place the configured PCR tube into the PCR instrument, set up and execute the predetermined amplification program. (4) Use VAHTS DNA Clean Beads to purify the amplified product. The operation method refers to the purification step mentioned above. Note: The sample completed in this step can be stably stored at 4°C for one week. If it needs to be stored for a long time, it should be placed at -20°C and try to avoid repeated freezing and thawing.

[0017] S46 library quality control: (1) Library concentration determination: The concentration of the constructed library was determined using the Qubit 4.0 fluorescence quantitative instrument. (2) Library length distribution detection: The length distribution of DNA fragments was detected using the Bioptic QSEP-100 fully automatic nucleic acid protein analysis system, and qualified cfDNA libraries were stored at -20°C until sequencing.

[0018] Preferably, the single-stranded cyclization and sequencing in step S5 specifically includes the following steps: (1) The double-stranded DNA library is processed, including denaturation, circularization, digestion and other steps, and finally single-stranded circular DNA (ssDNA) is obtained. After that, the rolling circle amplification (RCA) technology is applied to produce DNA nanoballs (DNBs); (2) The prepared DNA is loaded into the microarray chip, and the combined probe anchor polymerization (cPAS) technology is used to perform sequencing; (3) The sequencing primer anchoring and fluorescent probe polymerization reaction are carried out on the DNA nanoball, and the light signal is collected by a high-resolution imaging system, read and identified to obtain single-base sequence information. By repeating this process, comprehensive raw sequencing data is finally obtained. The library that passed the quality control is sequenced on the BGI MGISEQ-2000RS sequencer in a double-end mode, with 100bp at each end and a target average coverage of 1x for whole genome sequencing (WGS).

[0019] The present invention also relates to the use of the detection reagents of the aforementioned markers in the preparation of early prostate cancer diagnostic reagents or diagnostic kits. In some specific embodiments, the detection reagents include exosome extraction reagents and blood exosome RNA detection reagents; in some specific embodiments, the blood exosome RNA detection reagents include upstream and downstream primers and detection probes for PCR.

[0020] On the other hand, the present invention also proposes an evaluation system for early prostate cancer, including a registration and verification module, a sampling module, a sample pretreatment module, a cfDNA extraction and quantification module, a library construction and quality control module, a fluorescence quantification module, a fully automatic nucleic acid protein analysis module, a sequencing module, and a classification model construction and evaluation module; The registration and verification module is connected to the sampling module, and is used to verify and extract the patient's information, and cooperate with the sampling module to match the sample with the patient; The sequencing module is connected to a classification model construction and evaluation module, which includes a classification model construction unit, a model evaluation unit and a model diagnosis evaluation unit. The signal output end of the classification model construction unit is connected to the signal input end of the model evaluation unit, and the signal output unit of the model evaluation unit is connected to the signal input end of the model diagnosis evaluation unit.

[0021] Preferably, the classification model building unit uses one of the algorithms of linear regression, support vector machine, random forest, XGBoost and Stacking to build the classification model.

[0022] Preferably, the model evaluation unit uses a ten-fold cross validation and an independent test set evaluation method to evaluate the model. The independent test set evaluation method is used to verify the performance of the model on unseen data. The scikit-learn library is used to evaluate the model, covering the following core indicators: confusion matrix: a basic evaluation tool for visualizing model performance, ROC curve and AUC value: the ROC curve shows the relationship between the sensitivity and specificity of the model, and the AUC value is used to quantify the model's predictive ability; the closer the AUC value is to 1, the stronger the predictive performance, classification accuracy indicators: including sensitivity, specificity, positive predictive value, negative predictive value and accuracy, and F1 score, which comprehensively reflects the accuracy and balance of the model. The specific formula is as follows: Sensitivity = TP / (TP+FN)×100%; Specificity = TN / (TN+FP) × 100%; Positive predictive value = TP / (TP+FP) × 100%; Negative predictive value = TN / (TN+FN) × 100%; Accuracy = (TP+TN) / (TP+FP+TN+FN)×100%; F1 score = 2 × (precision × recall) / (precision + recall) × 100%; Among them, TP stands for true positive, FP stands for false positive, TN stands for true negative, and FN stands for false negative. Through the above indicators, the performance of the model in the classification task can be comprehensively evaluated.

[0023] Furthermore, the features included in the final diagnostic model of the model diagnosis evaluation unit include the following tumor joint markers: TCGGGG_M6bp, TCGGGG_B6bp, GATGTG_M6bp, GATGTG_B6bp, GTAGTA_B6bp, GTAGTA_M6bp, GAAGTT_M6bp, CGTCGC_B6bp, CGTCGC_M6bp, CGTACC_M6bp, AGCGGA_B6bp, GCTAAT_M6bp, AGCGGA_M6bp, GCTAAT_B6bp, CGACGT_B6bp, CGACGC_M6bp, CGACGT_M6bp, CGCAGT_B6bp, CGCAGT_M6bp, CGACGC_B6bp, CGGGAG_B6bp, CCCCCC_B6bp, CCCCCC_M6bp, CGTACC_B6bp and CGGGAG_M6bp.

[0024] Beneficial Effects The present invention provides a diagnostic marker, diagnostic reagent or diagnostic kit, evaluation system and application of early prostate cancer based on circulating free DNA (cfDNA). Compared with the prior art, it has the following beneficial effects: 1. The present invention establishes a model based on the spectrum of circulating free DNA (cfDNA) in plasma. This system is non-invasive, easy to operate, and highly accurate. It can effectively distinguish the differences in cfDNA characteristics between prostate cancer patients and healthy people, thereby efficiently identifying early prostate cancer patients.

[0025] 2. The present invention calculates the sensitivity, specificity, positive predictive value, accuracy and negative predictive value through the algorithm adopted in the model evaluation, so that the model evaluation can be carried out more comprehensively, which is conducive to improving the accuracy of the evaluation.

[0026] 3. The present invention can quickly read the patient's medical information through registration verification, provide the patient to be sampled with matters that need to be paid attention to before sampling, and reasonably arrange the sampling time for the patient to be sampled, which greatly improves the diagnostic evaluation efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the specific implementation of the present invention or the technical solution in the prior art, the following is a brief introduction to the drawings required for use in the specific implementation or the prior art description. Obviously, the drawings in the following description show some implementations of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0028] Figure 1This is a frequency diagram of the plasma cfDNA fragment terminal motif CCCA of the present invention; Figure 2 is the ROC curve diagram of the MDS of the present invention; Figure 3 is the ROC curve of each model of the present invention in the PRAD VS Healthy group; Figure 4 It is a schematic diagram of the system structure of the present invention; Figure 5 It is a structural principle diagram of the classification model construction and evaluation module of the present invention; Figure 6 This is a structural principle diagram of the registration and verification module of the present invention. DETAILED DESCRIPTION

[0029] The embodiments of the present invention will be described in detail below in conjunction with the examples, but it will be appreciated by those skilled in the art that the following examples are only used to illustrate the present invention and should not be considered as limiting the scope of the present invention. If no specific conditions are specified in the examples, the conditions are carried out according to normal conditions or manufacturer recommendations. If the manufacturer is not specified for the reagents or instruments used, they are all conventional products that can be purchased from the market.

[0030] Example 1 The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings 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.

[0031] See also Figure 1-3 The embodiment of the present invention provides a technical solution: a diagnostic marker for early prostate cancer based on circulating free DNA (cfDNA), an evaluation system and its application, specifically including: (1) Sample collection and processing: Subjects: A total of 36 prostate cancer (PRAD) patients and 64 healthy controls were included in this study. All PRAD patients were in the early stage (T1-T2) and eligible for surgical resection.

[0032] Informed consent: Before the start of the study, all participants were fully informed of the purpose and process of the study, and written consent was obtained.

[0033] Collection process: 5 mL of blood sample was collected from the peripheral blood of the subjects under fasting conditions using EDTA anticoagulation vacuum blood collection tubes, and then immediately inverted 8-10 times to ensure sufficient mixing and prevent coagulation.

[0034] Sample processing: After collection, the samples were transported to the laboratory at low temperature within 2 hours. After centrifugation at 1500g for 10 minutes at 4°C, the supernatant (plasma) was dispensed into cryovials, avoiding contact with the buffy coat, and this operation was performed on a clean bench.

[0035] Storage: The packaged plasma samples were stored in a -80℃ refrigerator for future use.

[0036] (2) cfDNA extraction and quantification: Method selection: Thermo Fisher Scientific's MagMAX Cell-free DNA isolation kit was used to extract cfDNA.

[0037] Operating procedures: Follow the detailed operating instructions provided by the manufacturer to perform cfDNA extraction and quantification, ensuring accurate execution of each step.

[0038] (3) Library construction and quality control: After successful extraction and quality control, the cfDNA samples were used for library construction using the VAHTS UniversalDNA Library Prep Kit for Illumina V3, which included steps such as end repair, phosphorylation, adding "A" tails, and adapter ligation. They were then enriched by PCR to prepare the cfDNA library for whole genome sequencing (WGS).

[0039] (4) Sequencing: The library that has passed quality control confirmation is subjected to 1× coverage WGS analysis on the BGI MGISEQ-2000RS sequencer, using a paired-end 100bp sequencing strategy.

[0040] (5) Application of machine learning algorithms: This study uses linear regression (LR), support vector machine (SVM), random forest (RF), XGBoost and Stacking algorithms to build classification models. These algorithms have shown high accuracy and stability in classification tasks.

[0041] (6) Model evaluation method In machine learning, overfitting is when a model performs well on the training set, but performance degrades on the test set. To evaluate model performance, there are two main approaches: Ten-fold cross validation: performed on the training set to ensure the robustness of model evaluation.

[0042] Independent test set evaluation: Verify the performance of the model on unseen data. Use the scikit-learn library for model evaluation, covering the following core indicators: Confusion matrix: Basic evaluation tool for visualizing model performance. ROC curve and AUC value: The ROC curve shows the relationship between the sensitivity and specificity of the model, while the AUC value (range 0.5-1) is used to quantify the model's predictive ability. The closer the AUC value is to 1, the stronger the predictive performance. Classification accuracy indicators: Including sensitivity (recall rate), specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy, and F1 score, which comprehensively reflect the accuracy and balance of the model. The specific formula is as follows: Sensitivity = TP / (TP + FN) × 100%; Specificity = TN / (TN +FP) × 100%; Positive predictive value = TP / (TP + FP) × 100%; Negative predictive value = TN / (TN + FN) × 100%; Accuracy = (TP + TN) / (TP + FP + TN + FN) × 100%; F1 score = 2 × (precision × recall) / (precision + recall) × 100%. Among them, TP stands for true positive, FP stands for false positive, TN stands for true negative, and FN stands for false negative. Through the above indicators, the performance of the model on the classification task can be comprehensively evaluated.

[0043] (7) Diagnostic model The features included in the final diagnostic model include TCGGGG_M6bp, TCGGGG_B6bp, GATGTG_M6bp, GATGTG_B6bp, GTAGTA_B6bp, GTAGTA_M6bp, GAAGTT_M6bp, CGTCGC_B6bp, CGTCGC_M6bp, CGTACC_M6bp, AGCGGA_B6bp, GCTAAT_M6bp, AGCGGA_M6bp, GCTAAT_B6bp, CGACGT_B6bp, CGACGC_M6bp, CGACGT_M6bp, CGCAGT_B6bp, CGCAGT_M6bp, CGACGC_B6bp, CGGGAG_B6bp, CCCCCC_B6bp, CCCCCC_M6bp, CGTACC_B6bp, and CGGGAG_M6bp. In the PRAD VS Healthy training set, we constructed five classification models, and the results are as follows: Figure 3A and Table 5. In the training set, the prediction performance of the model was evaluated according to the AUC value. The optimal models based on the RF, SVM and Stacking algorithms performed consistently and showed the best prediction performance, with AUC values ​​of 100.0%. The XGBoost and LR models followed closely, with AUC values ​​of 99.9% and 82.6%, respectively. When the sensitivity reached 93.10%, the prediction performance of the model was evaluated according to the F1 score. The RF, SVM, Stacking and XGBoost models performed consistently and showed the best prediction performance, with specificity, accuracy and F1 scores of 100.00%, 97.06% and 96.43%, respectively; the optimal model of LR performed relatively poorly, with a specificity of only 64.10%. In the test set, the prediction performance of these optimal models was evaluated, and the results are shown in Figure 3B and Table 5. RF, SVM, Stacking and XGBoost models can effectively distinguish PRAD patients from healthy controls, and their AUC values, sensitivity, specificity, accuracy and F1 scores all reached 100.00%. Among the five classification models, the LR model still performed the worst, with an AUC value, sensitivity and specificity of 85.7%, 100.00% and 40.00%, respectively. In short, the RF, SVM, Stacking and XGBoost models based on the omics features of cfDNA fragments can effectively identify early PRAD patients. In short, the SVM and Stacking models based on the omics features of cfDNA fragments can effectively identify early PRAD patients.

[0044] In some specific embodiments, if the detection information of the marker exceeds a threshold, the subject is determined to have early prostate cancer, otherwise the subject is determined to not have early prostate cancer.

[0045] In some specific embodiments, the system further includes a detection module and / or a display module.

[0046] In some specific embodiments, the detection module is used to detect the aforementioned markers, and the detection module uses the aforementioned diagnostic reagents or diagnostic kits for detection.

[0047] In some specific implementations, the display module is used to display the diagnosis results by display, printing or broadcasting.

[0048] Statistical methods used in the above steps: Clinical data in this study were analyzed using SPSS 26.0 software. For normally distributed data, mean ± standard deviation was used and differences were compared by t test; for non-normally distributed data, median and interquartile range were used and Mann-Whitney U test was used for comparison. Count data were expressed as number and percentage and compared by chi-square test. R software (version 4.2.1) and “limma” package were used to analyze the data differences between cancer and healthy groups, with thresholds set as |Log2 FC|≥2 and adjusted P value <0.05. Heat maps, box plots, and ROC curves were drawn using multiple packages of R software. Python software (version 13.9.13) was used to construct classification models and calculate indicators such as sensitivity, specificity, and accuracy. Bilateral P <0.05 was considered statistically significant.

[0049] Example 2 See also Figures 1 to 3 This embodiment provides a technical solution: a diagnostic marker for early prostate cancer based on circulating free DNA (cfDNA), an evaluation system and its application, specifically including: 1. Study sample collection. Inclusion and exclusion criteria including age, treatment history, pathological diagnosis and functional status. Collect 5 mL of fasting peripheral blood anticoagulated with EDTA from the subjects. See above for detailed criteria; 2. Sample pretreatment: After plasma is separated by low-temperature centrifugation, the supernatant is transferred to a new cryotube and stored at -80°C.

[0050] 3. cfDNA extraction and quantification. Plasma samples were thawed at -80°C, and after lysis, incubation, ice bath, enrichment and washing, the cfDNA concentration was determined using the Qubit dsDNA HS detection kit, and the DNA fragment length was detected using the Bioptic QSEP-100 system. See above for detailed steps; 4. Library construction and quality control. This includes key steps such as end repair, adapter ligation, PCR amplification, and library quality assessment to ensure the quality of library construction. Finally, the Qubit4.0 and Bioptic QSEP-100 systems are used to determine the library concentration and fragment length. The library construction process is described above; 5. Single-strand circularization and sequencing. Double-end sequencing was performed using the BGISEQ-2000RS sequencer, with the average coverage set to 1x to obtain high-quality WGS data. See above for detailed steps; 6. lcWGS data analysis process: First, the original sequencing data (raw reads in fastq format) were subjected to preliminary quality control using fastp software, including: (1) deleting reads with an N base content of more than 10%; (2) removing adapter sequences; (3) removing reads with low-quality bases (Q≤15) accounting for more than 40% in single-end reads. After quality control, the obtained clean data (clean reads) will be evaluated for quality using FastQC software. Then, the clean reads were aligned with the human reference genome hg19 using BWA software, and the mapping rate, duplication rate, and genome coverage were analyzed using Samtools software. Sambamba software was used to remove duplicate reads and reads with a mapping rate of less than 30%.

[0051] 7. Whole genome fragmentation feature analysis: hg19 autosomes were divided into 572 regions of 5M in size. After removing known low mapping and Duke blacklist regions, 473 regions were finally determined for analysis. Based on the reference genome alignment results, the insertion size of cfDNA fragments was calculated to distinguish short fragments of 100-150bp and long fragments of 151-220bp. The ratio of the number of short fragments to long fragments in each 5M region is the fragment size ratio (FSR).

[0052] 8. Analysis of terminal motifs and breakpoint motifs: The end repair step in library preparation is designed to process the 3' end of the cfDNA fragments to promote complementary alignment of the 3' and 5' ends. After all DNA molecules are processed, the 5' end is retained. After the double-end sequencing results are aligned with the reference genome, the 4bp and 6bp terminal motifs at the 5' end of the cfDNA fragments and the corresponding breakpoint motifs are defined. The types of terminal motifs and breakpoint motifs and their frequencies of occurrence are statistically analyzed for further bioinformatics evaluation.

[0053] 9. Terminal Motif Diversity Score (MDS): To quantify the diversity of terminal motifs in cfDNA fragments, the MDS of each sample is calculated using the normalized Shannon entropy formula based on the frequency of occurrence of each terminal motif. The MDS value ranges from 0 to 1, with higher values ​​indicating more terminal motif types in cfDNA, and vice versa.

[0054] 10. Classification model construction and evaluation: Combine cfDNA fragmentation features (such as FSR) and motif features (including terminal motifs and breakpoint motifs) to build a prediction model for early diagnosis of cancer. Five machine learning algorithms, including LR, RF, SVC, xgboost and stacking, were used for predictive analysis of urinary tract malignancies. A binary classification model was constructed by performing ten-fold cross-validation on the training set, and the model robustness was verified in an independent test set. At the same time, indicators such as AUC, sensitivity, specificity, positive predictive value, negative predictive value and accuracy were used to evaluate the model performance.

[0055] In summary, the present invention provides a diagnostic marker for early prostate cancer based on circulating cell-free DNA (cfDNA), an evaluation system and its application in combination with circulating cell-free DNA (cfDNA). The marker was combined with TCGGGG_M6bp, TCGGGG_B6bp, GATGTG_M6bp, GATGTG_B6bp, GTAGTA_B6bp, GTAGTA_M6bp, GAAGTT_M6bp, CGTCGC_B6bp, CGTCGC_M6bp, CGTACC_M6bp, AGCGGA_B6bp, GCTAAT_M6bp, AGCGGA_M6bp, GCTAAT_B6bp, CGACGT_B6bp, CGACGC_M6bp, CGACGT_M6bp, CGCAGT_B6bp, CGCAGT_M6bp, CGACGC_B6bp, CGGGAG_B6bp, CCCCCC_B6bp, CCCCCC_M6bp, CGTACC_B6bp, and CGGGAG_M6bp to establish a diagnostic model for early PRAD, so as to perform sensitive and specific evaluation and diagnosis of PRAD. Through low-coverage whole genome sequencing (lcWGS) technology, the present invention analyzed the fragmentation characteristics and terminal motifs of cfDNA. Based on the differences in cfDNA fragment characteristics, LR, SVC, random forest (RF), XGboost and Stacking algorithms were used to construct models respectively. RF, SVM, Stacking and XGBoost models can effectively distinguish PRAD patients from healthy controls, and their AUC values, sensitivity, specificity, accuracy and F1 scores all reached 100.00%. The LR model still performed the worst, with an AUC value, sensitivity and specificity of 85.7%, 100.00% and 40.00%, respectively. In short, the RF, SVM, Stacking and XGBoost models based on the omics characteristics of cfDNA fragments can effectively identify early PRAD patients and healthy people.

[0056] Table 1 Overview of sequencing data of study samples Table 1 Summary of sequencing data of study samples Table 2 Basic characteristics of PRAD patients and healthy controls Table 2 General characteristics of the PRAD patients and healthy controls Note: a The measurement data were expressed as median and interquartile range (P25, P75), and the count data were expressed as number of cases (n) and percentage (%). b When comparing the two groups, the measurement data were tested by Mann-Whitney U test, and the count data were tested by chi-square test. PRAD: prostate adenocarcinoma; BMI: body mass index; cfDNA: circulating free DNA; TPSA: total prostate-specific antigen; FPSA: free prostate-specific antigen.

[0057] Table 3 Relationship between 4b terminal motif diversity scores and clinically relevant characteristics in PRAD patients Table 3 Relationship between 4bp end motif scores and clinically relevant diversity characteristics in PRAD patients Note: The comparison between the two groups was performed using the Mann-Whitney U test. cfDNA: circulating free DNA; MDS: motif diversity score; Table 4 Diagnostic value of MDS in each case-control group Table 4 Diagnostic value of MDS in each case control group Table 5 Prediction performance of each model in PRAD VS Healthy dataset Table 5 Predictive performance of each model in the PRAD VS Healthydataset Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, it should be understood by those skilled in the art that the technical solutions described in the above embodiments may still be modified, or some or all of the technical features therein may be replaced by equivalents. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A diagnostic marker for early prostate cancer, characterized in that: This marker is a set of 25 cfDNA characteristic markers obtained by screening circulating free DNA in the blood based on fluorescence quantification, nucleic acid protein analysis, and cfDNA sequencing. It can be used as a non-invasive biomarker for early prostate cancer detection. The 25 cfDNA characteristic markers screened include TCGGGG_M6bp, TCGGGG_B6bp, GATGTG_M6bp, GATGTG_B6bp, GTAGTA_B6bp, GTAGTA_M6bp, GAAGTT_M6bp, CGTCGC_B6bp, CGT CGC_M6bp, CGTACC_M6bp, AGCGGA_B6bp, GCTAAT_M6bp, AGCGGA_M6bp, GCTAAT_B6bp, CGACGT_B6bp, CGACGC_M6bp, CGACGT_ M6bp, CGCAGT_B6bp, CGCAGT_M6bp, CGACGC_B6bp, CGGGAG_B6bp, CCCCCC_B6bp, CCCCCC_M6bp, CGTACC_B6bp, CGGGAG_M6bp.

2. A method for detecting a marker suitable for early prostate cancer risk prediction, characterized in that: The specific steps include: S1, cfDNA sequencing of early prostate cancer; S2. Screening of blood cfDNA characteristic markers for early prostate cancer based on cfDNA sequencing; S3. Construct an early prostate cancer diagnosis model based on characteristic molecules screened by cfDNA sequencing; S4. Assess patient risk based on early prostate cancer diagnostic model.

3. The method for detecting a marker suitable for early prostate cancer risk prediction according to claim 2, characterized in that: The step S1 specifically includes the following steps: A1: Sample collection; A2: Sample pretreatment; A3: cfDNA extraction and quantification; A4: Library construction and quality control; A5: Single-strand circularization and sequencing.

4. An evaluation system for early prostate cancer based on circulating free DNA, characterized in that: It includes registration and verification module, sampling module, sample pretreatment module, cfDNA extraction and quantification module, library construction and quality control module, fluorescence quantification module, fully automatic nucleic acid protein analysis module, sequencing module and classification model construction and evaluation module; The registration and verification module is connected to the sampling module, and is used to verify and extract the patient's information, and cooperate with the sampling module to match the sample with the patient; The sequencing module is connected to a classification model construction and evaluation module, which includes a classification model construction unit, a model evaluation unit and a model diagnosis evaluation unit, wherein the signal output end of the classification model construction unit is connected to the signal input end of the model evaluation unit, and the signal output unit of the model evaluation unit is connected to the signal input end of the model diagnosis evaluation unit; The cfDNA extraction and quantification module and the library construction and construction module are both connected to the fully automatic nucleic acid protein analysis module, and the fully automatic nucleic acid protein analysis module is connected to the sequencing module.

5. The early prostate cancer assessment system based on circulating free DNA according to claim 4, characterized in that: The classification model building unit uses one of the algorithms of linear regression, support vector machine, random forest, XGBoost and Stacking to build a classification model.

6. The early prostate cancer assessment system based on circulating free DNA according to claim 4, characterized in that: The model evaluation unit uses ten-fold cross validation and independent test set evaluation methods to evaluate the model. The independent test set evaluation method is used to verify the performance of the model on unseen data. The scikit-learn library is used to evaluate the model, covering the following core indicators: confusion matrix: basic evaluation tool for visualizing model performance, ROC curve and AUC value: ROC curve shows the relationship between the sensitivity and specificity of the model, and the AUC value is used to quantify the prediction ability of the model; the closer the AUC value is to 1, the stronger the prediction performance, classification accuracy indicators: including sensitivity, specificity, positive predictive value, negative predictive value and accuracy, and F1 score, which comprehensively reflects the accuracy and balance of the model. The specific formula is as follows: Sensitivity = TP / (TP+FN)×100%; Specificity = TN / (TN+FP) × 100%; Positive predictive value = TP / (TP+FP) × 100%; Negative predictive value = TN / (TN+FN) × 100%; Accuracy = (TP+TN) / (TP+FP+TN+FN)×100%; F1 score = 2 × (precision × recall) / (precision + recall) × 100%; Among them, TP stands for true positive, FP stands for false positive, TN stands for true negative, and FN stands for false negative. Through the above indicators, the performance of the model in the classification task can be comprehensively evaluated.

7. An early prostate cancer assessment system based on circulating free DNA according to any one of claims 4 to 6, characterized in that: The features included in the final diagnostic model of the model diagnostic evaluation unit include the following tumor joint markers: TCGGGG_M6bp, TCGGGG_B6bp, GATGTG_M6bp, GATGTG_B6bp, GTAGTA_B6bp, GTAGTA_M6bp, GAAGTT_M6bp, CGTCGC_B6bp, CGTCGC_M6bp, CGTACC_M6bp, AGCGGA_B6bp, GCTAAT_M6bp, AGCGGA_M6bp, GCTAAT_B6bp, CGACGT_B6bp, CGACGC_M6bp, CGACGT_M6bp, CGCAGT_B6bp, CGCAGT_M6bp, CGACGC_B6bp, CGGGAG_B6bp, CCCCCC_B6bp, CCCCCC_M6bp, CGTACC_B6bp and CGGGAG_M6bp.

Images