Molecular marker detection system based on fusion gene multi-omics analysis

Abstract

The invention relates to the technical field of molecular marker detection, and discloses a molecular marker detection system based on fusion gene multi-omics analysis. The system provided by the invention constructs a '36 fusion genes + 48 single genes' multi-detection system based on fusion gene multi-omics analysis, combines a minimal residual focus signal enhancement mechanism and deep learning model analysis, realizes more sensitive detection of molecular markers of thyroid cancer, and can obtain rich molecular information in single detection. Meanwhile, a detection method for various thyroid cancer related molecular markers is realized, and high sensitivity and high specificity can still be achieved under the condition that the ctDNA content is extremely low.

Description

Technical Field

[0001] The present invention relates to the technical field of molecular marker detection, and in particular to a molecular marker detection system based on fusion gene multi-omics analysis. Background Art

[0002] Thyroid cancer is a clinically common endocrine malignancy, and its early detection and intervention are crucial for improving prognosis. Traditional thyroid cancer screening relies primarily on physical examination and palpation, as well as imaging (thyroid ultrasound). However, ultrasound examinations have limited sensitivity for small lesions (e.g., diameters <5 mm, particularly <2 mm) and are susceptible to missed diagnoses due to operator experience. Furthermore, given the high prevalence of thyroid nodules in the general population, distinguishing early-stage malignant nodules from these is a major challenge.

[0003] Detecting tumor-associated molecular markers (such as ctDNA, circulating RNA, and exosomes) in blood and other bodily fluids can noninvasively reveal tumor information in vivo. Preliminary research has yielded promising results, particularly for thyroid cancer: detecting specific gene mutations in plasma ctDNA can provide earlier warnings than traditional methods. However, due to the extremely low tumor burden in early-stage thyroid cancer, the amount of ctDNA released into the bloodstream is often negligible (perhaps only one ten-thousandth of the cell-free DNA or even less). Single-marker assays often lack sensitivity. For example, detecting only the BRAF V600E mutation is ineffective if the mutation is not present in a patient's tumor. Relying solely on a single fusion gene can also miss cases that do not carry the fusion. Therefore, multi-omics and multi-marker combined testing strategies are gaining increasing attention. Simultaneously analyzing up to dozens of genetic variants associated with thyroid cancer promises to improve detection rates and ensure the reliability of results.

[0004] Furthermore, incorporating artificial intelligence into early screening systems can uncover hidden cancer signals from multi-dimensional data, improving the accuracy of identifying early lesions. For example, building AI models by combining gene mutation profiles, fusion gene profiles, and other patient clinical data can help improve the sensitivity and specificity of early screening.

[0005] In summary, the development of a low-cost, high-throughput multi-marker detection system combined with a deep learning algorithm is expected to achieve more sensitive early detection and risk assessment of molecular markers of thyroid cancer. Summary of the Invention

[0006] In order to solve the technical problem of more sensitive detection of molecular markers of thyroid cancer, the present invention provides a molecular marker detection system based on fusion gene multi-omics analysis.

[0007] The specific technical solutions of the present invention are: In a first aspect, the present invention provides a molecular marker detection system based on fusion gene multi-omics analysis, comprising: (a) Detection kit: Contains: Primer pairs specifically targeting the breakpoints of 36 thyroid cancer fusion genes, probe sets specifically targeting 48 single-gene mutation sites, and capture probes carrying dual barcode sequences, each containing a sample ID tag and a molecule ID tag; Among them, the fusion genes include RET / PTC1, RET / PTC3, PAX8-PPARG, and the single genes include BRAFV600E and TERT C228T / C250T; (b) Signal noise reduction module: This module has a built-in background noise database and a statistical filtering unit. The filtering unit is configured to output a valid signal when the detected variation frequency is greater than 3 times the standard deviation of the background of a healthy population; (c) Deep learning model: The input is the effective signal and clinical feature data, and the output is a 0-1 risk score.

[0008] The system provided by this invention, based on multi-omics analysis of fusion genes, constructs a "36 fusion genes + 48 single genes" multiplex detection system. Combining a minimal residual lesion signal enhancement mechanism with deep learning model analysis, it enables more sensitive detection of thyroid cancer molecular markers, acquiring rich molecular information in a single test. Simultaneously, it implements a method for detecting multiple thyroid cancer-related molecular markers with high sensitivity and specificity, even at extremely low ctDNA levels.

[0009] As a preferred embodiment of the above detection system, in the detection kit, the sample ID tag of the dual barcode sequence is an 8-12 bp unique code, and the molecule ID tag is a 10-15 bp random nucleotide sequence.

[0010] As a preferred embodiment of the above detection system, the capture probe has a 12 bp sample ID tag at the 5' end and a 15 bp random molecular identifier (UMI) at the 3' end.

[0011] As a preferred embodiment of the above detection system, in the detection kit, the 36 fusion genes also include ETV6-NTRK3 and BRAF fusion, and the single gene mutation sites include NRAS Q61R and TP53 R175H.

[0012] As a preferred embodiment of the above detection system, in the signal noise reduction module, the statistical filtering unit is configured to trigger a joint positive determination when the frequencies of ≥2 independent markers exceed the background threshold.

[0013] As a preferred embodiment of the above detection system, the deep learning model adopts a Transformer architecture, including 4 attention heads and a 128-dimensional hidden layer.

[0014] In a second aspect, the present invention provides an integrated detection device for thyroid cancer molecular markers, comprising: Sample processing unit: integrated microfluidic chip for plasma cfDNA enrichment and library construction; Amplification sequencing unit: built-in high-fidelity PCR module and nanopore sequencing chip; Analysis control unit: runs the above-mentioned signal noise reduction module and deep learning model.

[0015] As a preferred embodiment of the above-mentioned detection device, the microfluidic chip integrates a cfDNA adsorption membrane, a UMI labeling amplification chamber, and a sequencing library purification channel.

[0016] In a third aspect, the present invention provides a storage medium storing computer-executable instructions, wherein when the instructions are executed, the following are implemented: Receive molecular marker detection data and call the above-mentioned signal noise reduction module to filter background noise; The denoised data is input into the above deep learning model to generate a risk score.

[0017] Compared with the prior art, the present invention has the following technical effects: The system provided by this invention, based on multi-omics analysis of fusion genes, constructs a "36 fusion genes + 48 single genes" multiplex detection system. Combining a minimal residual lesion signal enhancement mechanism with deep learning model analysis, it enables more sensitive detection of thyroid cancer molecular markers, acquiring rich molecular information in a single test. Simultaneously, it implements a method for detecting multiple thyroid cancer-related molecular markers with high sensitivity and specificity, even at extremely low ctDNA levels. DETAILED DESCRIPTION

[0018] The present invention will be further described below with reference to the following embodiments. Those skilled in the art will be able to implement the present invention based on these descriptions. Furthermore, the embodiments of the present invention described below are generally only a portion of the embodiments of the present invention, rather than all of the embodiments. Therefore, all other embodiments derived by those skilled in the art based on the embodiments of the present invention without inventive effort should fall within the scope of protection of the present invention.

[0019] Example 1 Provided is a molecular marker detection system based on fusion gene multi-omics analysis, comprising: (a) Detection kit: Contains: Primer pairs specifically targeting the breakpoints of 36 thyroid cancer fusion genes, probe sets specifically targeting 48 single-gene mutation sites, and capture probes carrying dual barcode sequences, each containing a sample ID tag and a molecule ID tag; Among them, the fusion genes include RET / PTC1, RET / PTC3, PAX8-PPARG, and the single genes include BRAFV600E and TERT C228T / C250T; (b) Signal noise reduction module: This module has a built-in background noise database and a statistical filtering unit. The filtering unit is configured to output a valid signal when the detected variation frequency is greater than 3 times the standard deviation of the background of a healthy population; (c) Deep learning model: The input is the effective signal and clinical feature data, and the output is a 0-1 risk score.

[0020] The system provided in this example utilizes multi-omics analysis of fusion genes to construct a "36 fusion genes + 48 single genes" multiplex detection system. This system, combined with a minimal residual lesion signal enhancement mechanism and deep learning model analysis, enables more sensitive detection of thyroid cancer molecular markers, acquiring rich molecular information in a single test. Simultaneously, it implements a method for detecting multiple thyroid cancer-related molecular markers, achieving high sensitivity and specificity even in extremely low ctDNA levels.

[0021] In this embodiment, as a preferred embodiment of the above detection system, in the detection kit, the sample ID tag of the dual barcode sequence is an 8-12 bp unique code, and the molecule ID tag is a 10-15 bp random nucleotide sequence.

[0022] In this embodiment, as a preferred embodiment of the above detection system, the capture probe has a 12 bp sample ID tag at the 5' end and a 15 bp random molecular identifier (UMI) at the 3' end.

[0023] In this embodiment, as a preferred embodiment of the above detection system, in the detection kit, the 36 fusion genes also include ETV6-NTRK3 and BRAF fusion, and the single gene mutation sites include NRAS Q61R and TP53 R175H.

[0024] In this embodiment, as a preference for the above detection system, in the signal noise reduction module, the statistical filtering unit is configured to trigger a joint positive determination when the frequencies of ≥2 independent markers exceed the background threshold.

[0025] In this embodiment, as a preferred embodiment of the above-mentioned detection system, the deep learning model adopts a Transformer architecture, including 4 attention heads and a 128-dimensional hidden layer.

[0026] Example 2 Provided is a molecular marker detection system based on fusion gene multi-omics analysis, comprising: Primer pairs specifically targeting the breakpoints of 36 thyroid cancer fusion genes, probe sets specifically targeting 48 single-gene mutation sites, and capture probes carrying dual barcode sequences, each containing a sample ID tag and a molecule ID tag; Among them, the fusion genes include RET / PTC1, RET / PTC3, PAX8-PPARG, and the single genes include BRAFV600E and TERT C228T / C250T; (b) Signal noise reduction module: This module has a built-in background noise database and a statistical filtering unit. The filtering unit is configured to output a valid signal when the detected variation frequency is greater than 3 times the standard deviation of the background of a healthy population; (c) Deep learning model: The input is the effective signal and clinical feature data, and the output is a 0-1 risk score.

[0027] (1) Multiple molecular marker detection panel detection kit: Contains primer pairs that specifically target 36 thyroid cancer fusion gene breakpoints, probe sets that specifically target 48 single gene mutation sites, and capture probes carrying dual barcode sequences. The 36 fusion genes cover common fusion events in thyroid cancer, such as RET gene fusion RET / PTC1, RET / PTC3, and other fusions of RET with different partner genes, NTRK1 / 3 fusion (such as ETV6-NTRK3), ALK fusion, and PPARG fusion (PAX8-PPARG), and may also include rare BRAF fusions. The 48 single genes include hotspot mutation sites and thyroid-specific genes that are closely related to the occurrence and development of thyroid cancer. For example, BRAF V600E, TERT promoter mutation (C228T / C250T), RAS family genes (NRAS Q61R, etc.), TP53, PIK3CA, AKT1, PTEN and other driver gene mutations, as well as some thyroid-related gene expressions or thyroid-specific thyroglobulin (Tg) gene sequences.

[0028] The above panels can use probe capture or high-multiplex PCR to achieve simultaneous detection of all targets. Preferably, target capture probes with unique dual barcode sequence tags are used to uniquely label each molecule during library preparation, improving the accuracy of multiplex detection.

[0029] (2) Signal noise reduction module: This module embodies the signal enhancement and noise reduction mechanism of the present invention. This module introduces a signal enhancement mechanism at both the experimental and algorithmic levels to address the problems of weak signal and high noise that may be caused by trace amounts of ctDNA. At the algorithmic level, a background model and statistical noise filtering are introduced: a large number of healthy individual samples are used to establish a background noise distribution model for each site. For the detected variation, its frequency is required to be significantly higher than the background noise threshold before it is judged as positive, thereby reducing false positive results. In addition, the multi-marker combination signal is comprehensively considered. For example, if multiple independent markers in the same patient sample are marginally positive, they are comprehensively judged as true positive, thereby improving the detection sensitivity; on the contrary, if only a single site slightly exceeds the threshold, it needs to be treated with caution or re-examined for confirmation.

[0030] Furthermore, at the experimental level, the amount of ctDNA collected was maximized by increasing the equivalent plasma sampling volume (e.g., processing more than 20 mL of plasma per test) and employing highly efficient cfDNA extraction and enrichment methods. Furthermore, Unique Molecular Identifier (UMI) tags and high-fidelity DNA polymerases were used during library amplification to correct for PCR amplification bias and identify sequencing errors.

[0031] (3) Deep learning model: This includes a set of trained deep learning models for intelligent interpretation and risk assessment of test results. The model can use an integrated multi-layer perceptron (MLP) or a Transformer-based architecture, and its input is the data of 84 molecular markers detected (including the qualitative results and quantitative abundance of each mutation / fusion) and some basic clinical information of the subjects (such as age, gender, etc., optional). After being trained on a large number of data from known thyroid cancer patients and healthy individuals, the model can output a risk score or classification result, such as a 0-1 risk score, or a classification result such as "suspected positive" or "negative". Specifically, the model learns the combination of molecular features of early thyroid cancer, such as the co-occurrence of multiple trace mutations often indicates the presence of a tumor, while the absence of abnormalities may be negative. In addition, the model can also identify the characteristics of benign nodules corresponding to certain marker patterns, thereby avoiding misjudging harmless mutations carried by benign nodules as cancer.

[0032] Through AI models to mine and analyze complex data, this system can increase the sensitivity to 96% and the specificity to over 99%.

[0033] Based on the above detection system, this embodiment also provides an integrated detection device for thyroid cancer molecular markers, which includes: Sample processing unit: integrated microfluidic chip for plasma cfDNA enrichment and library construction; Amplification sequencing unit: built-in high-fidelity PCR module and nanopore sequencing chip; Analysis control unit: runs the above-mentioned signal noise reduction module and deep learning model.

[0034] Preferably, the microfluidic chip integrates a cfDNA adsorption membrane, a UMI labeling amplification chamber, and a sequencing library purification channel.

[0035] Based on the above detection system, this embodiment further provides a storage medium storing computer-executable instructions, which, when executed, implement: Receive molecular marker detection data and call the above-mentioned signal noise reduction module to filter background noise; The denoised data is input into the above deep learning model to generate a risk score.

[0036] Unless otherwise specified, the raw materials and equipment used in the present invention are commonly used in the art; the methods used in the present invention are conventional methods in the art unless otherwise specified.

[0037] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any simple modification, change and equivalent transformation made to the above embodiment based on the technical essence of the present invention still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A molecular marker detection system based on fusion gene multi-omics analysis, characterized by: include: (a) Detection kit: Contains: Primer pairs specifically targeting the breakpoints of 36 thyroid cancer fusion genes, probe sets specifically targeting 48 single-gene mutation sites, and capture probes carrying dual barcode sequences, each containing a sample ID tag and a molecule ID tag; Among them, the fusion genes include RET / PTC1, RET / PTC3, PAX8-PPARG, and the single genes include BRAFV600E and TERT C228T / C250T; (b) Signal noise reduction module: This module has a built-in background noise database and a statistical filtering unit. The filtering unit is configured to output a valid signal when the detected variation frequency is greater than 3 times the standard deviation of the background of a healthy population; (c) Deep learning model: The input is the effective signal and clinical feature data, and the output is a risk score.

2. The detection system according to claim 1, wherein: In the detection kit, the sample ID tag of the dual barcode sequence is an 8-12 bp unique code, and the molecule ID tag is a 10-15 bp random nucleotide sequence.

3. The detection system according to claim 2, wherein: In the capture probe, the 5' end is a 12 bp sample ID tag, and the 3' end is a 15 bp random molecular identifier.

4. The detection system according to claim 1, wherein: In the detection kit, the 36 fusion genes also include ETV6-NTRK3 and BRAF fusion, and the single gene mutation sites include NRAS Q61R and TP53 R175H.

5. The detection system according to claim 1, wherein: In the signal noise reduction module, the statistical filtering unit is configured to trigger a joint positive determination when the frequencies of ≥2 independent markers exceed a background threshold.

6. The detection system according to claim 1, wherein: The deep learning model adopts the Transformer architecture, which includes 4 attention heads and 128-dimensional hidden layers.

7. An integrated detection device for thyroid cancer molecular markers, characterized by: include: Sample processing unit: integrated microfluidic chip for plasma cfDNA enrichment and library construction; Amplification sequencing unit: built-in high-fidelity PCR module and nanopore sequencing chip; Analysis control unit: runs the signal noise reduction module and deep learning model as described in claim 1.

8. The detection device according to claim 7, wherein: The microfluidic chip integrates a cfDNA adsorption membrane, a UMI labeling amplification chamber, and a sequencing library purification flow channel.

9. A storage medium storing computer-executable instructions, characterized in that: When executed, the instructions: receiving molecular marker detection data, and calling the signal noise reduction module as claimed in claim 1 to filter background noise; The denoised data is input into the deep learning model as described in claim 1 to generate a risk score.