System and method for detecting hepatocellular carcinoma microRNA group

Abstract

The invention provides a system for detecting hepatocellular carcinoma micro-RNA (Ribonucleic Acid) based on blood molecular spectrum analysis. The system includes a sample preparation module for processing a blood sample of a subject and separating a plasma portion and a serum portion therefrom. And the nucleic acid extraction module is used for extracting the microRNA expression spectrum from the plasma part. The real-time polymerase chain reaction module is used for detecting and quantifying the expression quantity of the micro RNA group in the micro RNA expression spectrum, and the real-time polymerase chain reaction module comprises miR-361-5p, miR-130a-3p, miR-27a-3p, miR-30d-5p and miR-193a-5p. Processing by a trained classifier according to the quantitative result to generate a hepatocellular carcinoma risk classification result so as to grade the subjects according to the hepatocellular carcinoma risk; and the trained classifier processing can be further incorporated into a result generated by combining the alpha fetoprotein expression quantity detected and quantified from the serum part with the expression quantity of the microRNA group, so that the accuracy is improved.

Description

Technical Field

[0001] This invention relates to the fields of molecular biology and computational pathology. More specifically, this invention relates to a system and method for generating hepatocellular carcinoma reports from subjects using circulating microRNA biomarkers and machine learning-based classification methods. Background Technology

[0002] Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer, accounting for over 90% of all liver cancer cases. Approximately 80% to 90% of newly diagnosed HCC patients also have cirrhosis. Among various causes, chronic hepatitis B virus infection is a major risk factor, especially in Asia, accounting for over 50% of HCC cases. Globally, HCC is the third leading cause of cancer death. Despite this, early-stage HCC is potentially curable; after interventions such as liver ablation, surgical resection, or transplantation, the 5-year survival rate can exceed 70%. However, due to the lack of reliable and effective early screening methods, the prognosis of HCC remains unfavorable in many cases, with approximately 50% of patients diagnosed at an advanced stage, making radical treatment unsuitable. Unlike other solid tumors, to avoid bleeding and metastasis, the diagnosis of HCC typically relies on non-invasive imaging rather than tissue biopsy. Current clinical guidelines from the American Association for the Study of Liver Diseases recommend liver ultrasound combined with serum alpha-fetoprotein (AFP) testing every two years for high-risk individuals, particularly adults with cirrhosis. Serum alpha-fetoprotein (AFP) is the most widely used serum biomarker for hepatocellular carcinoma screening worldwide. If abnormal results are found, such as liver nodules ≥10 mm or serum AFP levels higher than 20 ng / mL (AFP20), further diagnostic evaluation, such as multiphasic computed tomography (CT) or magnetic resonance imaging (MRI), should be performed.

[0003] In recent years, liquid biopsy has emerged as a novel minimally invasive method for early cancer screening and longitudinal disease monitoring. MicroRNAs are small non-coding RNAs, approximately 22 nucleotides in length, that regulate post-transcriptional gene expression by binding to the 3' untranslated region (3'UTR) of target messenger RNA (mRNA), thereby inducing mRNA degradation or inhibiting translation. Circulating microRNAs exhibit extremely high stability in blood, making them ideal biomarkers for liquid biopsy applications. Although many studies have attempted to analyze microRNA characteristics for the diagnosis of hepatocellular carcinoma, different studies have yielded inconsistent results. For example, the dysregulation of circulating microRNA miR-122-5p in hepatocellular carcinoma has been inconsistent across different studies; some studies have shown upregulation, while others have reported downregulation or no significant change.

[0004] Therefore, this invention provides a method and system for analyzing a subset of hepatocellular carcinoma microRNAs in subject samples. This method, through predetermined molecular processing steps and data analysis workflows, assists professionals in making clinical decisions. This invention also relates to a workflow or platform for detecting and analyzing microRNA expression profiles, thereby enhancing the application value of circulating biomarkers in the management of liver disease progression. Summary of the Invention

[0005] The purpose of this invention is to provide a system and method to solve the above-mentioned technical problems.

[0006] According to a first aspect of the invention, a system for detecting hepatocellular carcinoma microRNA profiles is provided. More specifically, the system includes: a sample preparation module for processing a blood sample and separating a plasma fraction and a serum fraction therefrom; a nucleic acid extraction module for extracting a microRNA expression profile from the plasma fraction; a real-time polymerase chain reaction (PCR) module for detecting and quantifying the expression levels of a microRNA group (miR-361-5p, miR-130a-3p, miR-27a-3p, miR-30d-5p, and miR-193a-5p) in the microRNA expression profile; and a computing device including a processor and a memory, wherein the memory stores instructions that, when the processor executes the instructions, perform the following steps: taking a quantitative value of the expression level of the microRNA group as input; processing the input using a trained classifier to generate a hepatocellular carcinoma risk classification result based on the expression level of the microRNA group, and rating the subject according to hepatocellular carcinoma risk (i.e., low risk, intermediate risk, and high risk).

[0007] According to one embodiment of the invention, the computing device can further process the alpha-fetoprotein (AFP) expression level detected and quantified from the serum fraction and compare it with a threshold of 20 ng / mL. Regardless of whether the AFP level is below or above the threshold, the classifier incorporates both the AFP expression level and the microRNA expression level as binary inputs into the hepatocellular carcinoma risk classification result. Alternatively, when the AFP expression level data is not applicable, a separate classifier using only the microRNA group can be used.

[0008] According to another embodiment of the invention, the trained classifier includes a logistic regression model, wherein the logistic regression model is trained on an annotated training dataset, and the training dataset includes microRNA expression profiles and ground truth annotations for hepatocellular carcinoma diagnosis. This training dataset can be obtained from patient medical records.

[0009] According to another embodiment of the invention, the system further includes a training module to update the trained classifier based on a new annotated training dataset.

[0010] According to another embodiment of the present invention, the subject is a patient with cirrhosis.

[0011] According to a second aspect of the present invention, a computer-implemented method for detecting hepatocellular carcinoma microRNA profiles is provided. More specifically, the method includes: obtaining a blood sample from a subject; separating a plasma fraction from the blood sample; extracting a microRNA expression profile of the subject from the plasma fraction; detecting and quantifying the expression level of a microRNA group in the microRNA expression profile, the microRNA group including miR-361-5p, miR-130a-3p, miR-27a-3p, miR-30d-5p, and miR-193a-5p; processing the quantified value of the expression level using a machine learning classifier trained on a computer processor, the classifier generating a hepatocellular carcinoma risk classification result based on the expression level of the microRNA group, rating the subject according to hepatocellular carcinoma risk (i.e., low risk, intermediate risk, and high risk).

[0012] According to one embodiment of the invention, the method further includes separating a serum fraction from the blood sample, detecting and quantifying alpha-fetoprotein (AFP) expression levels therein, and comparing the measured AFP expression levels with a threshold of 20 ng / mL. Regardless of whether the AFP level is below or above the threshold, the classifier incorporates both the AFP expression level and the microRNA group expression level as binary inputs into the hepatocellular carcinoma risk classification result. Alternatively, when AFP expression level data is not applicable, a separate classifier using only the microRNA group is used.

[0013] According to another embodiment of the present invention, the subject is a patient with cirrhosis.

[0014] According to another embodiment of the invention, the classifier includes a logistic regression model trained on a LASSO-regularized annotated training dataset.

[0015] According to another embodiment of the invention, the method further includes generating a hepatocellular carcinoma report via the computer processor, the hepatocellular carcinoma report including hepatocellular carcinoma risk classification information. The hepatocellular carcinoma report also includes one or more recommendations, including repeat testing, multi-contrast magnetic resonance imaging (MRI), or oncology consultation. Attached Figure Description

[0016] Referring to the accompanying drawings, embodiments of the present invention will be described in more detail below, wherein:

[0017] Figures 1A-1C This study demonstrates the upregulation of 18 microRNAs in patients with hepatocellular carcinoma, including: Figure 1A This illustrates the workflow according to an embodiment of the present invention; Figure 1BA volcano plot showing the differences in microRNA expression levels between hepatocellular carcinoma patients and cirrhosis patients during the screening phase, with 16 validated microRNA targets identified; and Figure 1C The results of unsupervised hierarchical clustering of 18 microRNA targets in 354 paired samples during the validation phase 1 are presented.

[0018] Figures 2A-2D The detection performance and risk scoring effect of microRNAome are demonstrated in different sample sets, including: Figure 2A Show the training set; Figure 2B Show the test set; Figure 2C Display the validation set; and Figure 2D The distribution of risk scores for microRNA groups in patients with different stages of hepatocellular carcinoma is shown, indicating that the risk score effect of this microRNA group is superior to that of serum alpha-fetoprotein levels above 20 ng / mL in all sample sets.

[0019] Figures 3A-3E The correlation between microRNA expression levels and clinical parameters was demonstrated, among which... Figures 3A-3C Showing its relationship with different stages of hepatocellular carcinoma (HCC) Figure 3A ), tumor size ( Figure 3B ) and tumor invasion ( Figure 3C The correlation of ); Figure 3D Demonstrating the interaction network between 5 microRNAs and their target genes; and Figure 3E The results of Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of the target genes of these five microRNAs are presented.

[0020] Figure 4 A system for detecting the performance levels of a subject's microRNA genome according to an embodiment of the present invention is shown. Detailed Implementation

[0021] In the following description, systems and / or methods for detecting hepatocellular carcinoma microRNA genomes are listed as preferred embodiments. It will be apparent to those skilled in the art that modifications, including additions and / or substitutions, can be made without departing from the scope and spirit of the invention. Specific details may be omitted below to avoid obscuring the invention; however, this document is prepared to enable those skilled in the art to practice the techniques disclosed herein without excessive experimentation.

[0022] As used herein, the term "annotated training dataset" refers to a set of biological and / or clinical data obtained from human subjects, where each data point is annotated with relevant clinical annotations. These datasets typically include molecular or biomarker profiles, which serve as true annotations for supervising the training and validation of machine learning models.

[0023] In some embodiments, the annotated training dataset may be derived from prospective or retrospective clinical studies, multicenter clinical trials, hospital biobanks, or publicly available databases such as the Gene Expression Database (GEO), The Cancer and Tumor Genome Atlas (TCGA), or the International Cancer Genome Consortium (ICGC). This dataset may be preprocessed to remove interference or artifacts and may be standardized to ensure consistent comparisons across samples. In some embodiments, the annotated training dataset includes at least the expression levels of the microRNA group (e.g., miR-361-5p, miR-130a-3p, miR-27a-3p, miR-30d-5p, and miR-193a-5p) and the true labeling of diagnostic results (e.g., confirmed diagnosis of hepatocellular carcinoma versus non-hepatocellular carcinoma).

[0024] Annotated training datasets can be used to train, test, and validate classifiers, including logistic regression models, neural networks, random forests, and other supervised learning algorithms, to assess a patient's risk or likelihood of developing hepatocellular carcinoma. The term can also encompass processed datasets, including interplate controls, normalization factors, and metadata that associates each sample with a clinical cohort or outcome category.

[0025] According to a first aspect of the invention, a system is provided for detecting the hepatocellular carcinoma microRNA genome of a subject by integrating molecular profiling analysis with computational processing. See also Figure 4 According to an embodiment of the present invention, the system 10 is designed to simplify and refine the early detection and classification process of hepatocellular carcinoma risk by analyzing specific circulating microRNAs in blood samples, utilizing high-throughput, real-time quantitative polymerase chain reaction and machine learning-based assessment results.

[0026] In one embodiment, system 10 includes a sample preparation module 101 for processing blood samples obtained from a subject (preferably via venipuncture). Upon obtaining the blood sample, the sample preparation module 101 processes the blood sample, separating it into plasma and serum fractions using centrifugation or other conventional separation techniques. These different components are then used for subsequent molecular profiling analysis and biomarker detection.

[0027] Total RNA, including microRNAs, is analyzed from this plasma fraction using nucleic acid extraction module 102. Nucleic acid extraction module 102 can employ a spinning column method, magnetic bead technology, or automated cassette method to ensure efficient and high-quality RNA extraction for subsequent nucleic acid analysis. The focus of extraction is on the microRNA expression profile in the plasma, which includes stable circulating microRNA biomarkers relevant to tumor biology.

[0028] The extracted RNA was then fed into a real-time polymerase chain reaction (RT-PCR) module 103 to detect and quantify the expression levels of a predetermined microRNA group, including miR-361-5p, miR-130a-3p, miR-27a-3p, miR-30d-5p, and miR-193a-5p. This microRNA group can serve as a specific indicator to differentiate liver lesions from hepatocellular carcinoma and non-hepatocellular carcinoma, especially in high-risk populations. The R-PCR module 103 includes thermal cycling capabilities and a fluorescence detection optics system to support locked nucleic acid (LNA) probe quantification, thereby enabling sensitive and specific amplification of the target microRNA.

[0029] The system 10 also includes a computing device 104 having a processor 104a and a memory 104b storing executable instructions. When the processor 104a executes instructions, it receives as input the quantitative values ​​of microRNA expression from the real-time polymerase chain reaction module 103. The processor 104a then processes the input using a trained classifier (e.g., a logistic regression model) and generates a hepatocellular carcinoma risk classification, rating the subject according to hepatocellular carcinoma risk (i.e., low risk, intermediate risk, and high risk).

[0030] It is worth noting that the classifier is trained using an annotated training dataset, which includes microRNA expression profiles and corresponding clinical hepatocellular carcinoma diagnostic information, thereby enabling predictive modeling and risk stratification.

[0031] The computing device can also output a hepatocellular carcinoma report, which is displayed via a user interface. This report may include hepatocellular carcinoma risk classification information and the relative expression levels of each microRNA within the microRNA group. This hepatocellular carcinoma report can assist clinicians in making decisions regarding further diagnostic evaluation, treatment initiation, or monitoring strategies.

[0032] In some embodiments, the computing device further integrates the expression data of the microRNA genome with the expression level of alpha-fetoprotein (AFP) measured from the serum fraction. AFP expression is a widely used biomarker for liver cancer screening, and combining it with the microRNA genome can further improve the accuracy of detection. Specifically, the system includes comparing AFP concentrations to clinically accepted thresholds (typically 20 ng / mL) as part of a classification consideration, thus simultaneously introducing molecular and serological markers to achieve multimodal risk assessment.

[0033] To enhance the adaptability and continuous improvement of detection performance, the system may also include a training module to provide new annotated training datasets for retraining or updating the classifier. This dynamic learning capability allows the system to stay in sync with emerging clinical data and changes in population-specific biomarkers.

[0034] This system is particularly suitable for high-risk groups for hepatocellular carcinoma, such as patients diagnosed with cirrhosis. Since patients with cirrhosis typically require monitoring for hepatocellular carcinoma progression over time, the system of this invention provides a minimally invasive and highly accurate tool for early detection and monitoring.

[0035] By combining blood molecular diagnostics, machine learning-based classification, and real-time clinical decision support, the system of this invention can improve the detection and monitoring of hepatocellular carcinoma in high-risk populations and support timely and personalized clinical intervention.

[0036] According to a second aspect of the invention, a computer-implemented method for detecting the microRNA genome in hepatocellular carcinoma is provided, with the option to selectively integrate serum biomarkers. This method employs machine learning to generate predictive models for early detection and risk stratification of hepatocellular carcinoma in clinical settings.

[0037] In this method, a blood sample is first obtained from the subject via routine venipuncture, and the plasma fraction is separated from the blood sample using standard centrifugation or plasma separation procedures. Nucleic acids are extracted from the separated plasma, with a focus on extracting small molecular weight RNA classes, especially microRNAs, which can serve as indicators of tumorigenesis.

[0038] MicroRNA expression profiles were extracted from the subjects' plasma fractions, and the expression levels of a specific microRNA group, including miR-361-5p, miR-130a-3p, miR-27a-3p, miR-30d-5p, and miR-193a-5p, were quantified. These microRNAs can differentiate hepatocellular carcinoma from non-hepatocellular carcinoma liver diseases (such as cirrhosis). Quantification methods could employ real-time quantitative polymerase chain reaction (RT-qPCR) with locked nucleic acid (LNA) probes, or any quantitative method that provides suitable quantitative values ​​as input.

[0039] Once quantification is complete, the performance quantification values ​​of the five microRNAs are processed by a machine learning classifier trained on a computer processor. In one embodiment, the classifier includes a logistic regression model trained using Least Absolute Shrinkage and Selection Operator (LASSO) regularization, thus enabling the model to select the most informative features while avoiding overfitting. The classifier is trained on an annotated training dataset that includes ground truth labels for diagnosis and matching molecular profiles, allowing the model to learn discriminative patterns associated with hepatocellular carcinoma risk.

[0040] Based on the processed input data, the classifier generates a hepatocellular carcinoma (HCC) risk classification, rating subjects according to their HCC risk (i.e., low, intermediate, and high). Based on this classification, the computer's processor further generates a HCC report, which may include one or more of the following outputs: HCC risk classification, a report showing the relative performance of microRNAs, and diagnostic interpretations or recommendations. Recommendations may include clinical actions such as repeat testing, scheduling multi-contrast MRI, or referring the patient to an oncology department to support timely clinical intervention and treatment planning.

[0041] In another embodiment, the raw quantitative values ​​of the real-time polymerase chain reaction (typically the cycle threshold Ct value) are normalized relative to internal control microRNAs (such as miR-16-5p, which exhibits stable expression levels across different samples and diseases) to determine the relative expression level of each microRNA. Therefore, the resulting normalized values ​​(e.g., using 2) -ΔCt Method, where ΔCt=Ct 目标 -Ct 内参 It can accurately reflect the abundance of each target microRNA in the subject's plasma relative to the internal reference microRNA in the same sample.

[0042] Hepatocellular carcinoma reports can be displayed via a user interface that can be integrated into clinical dashboards, electronic health record systems, or standalone websites. This interface is easy for healthcare providers to use and interpret.

[0043] In some embodiments, the method further includes detecting and quantifying pseudoalpha-fetoprotein (AFP) expression levels from a serum fraction of the same blood sample. The quantified AFP expression levels can be combined with microRNA expression levels and used in hepatocellular carcinoma risk classification reports to improve accuracy. In these embodiments, AFP measurements can be compared to a threshold of 20 ng / mL (clinically commonly used as a cutoff for elevated hepatocellular carcinoma risk).

[0044] Preferably, the subjects being examined are patients with cirrhosis, because individuals with cirrhosis have a higher risk of developing hepatocellular carcinoma and are usually the primary target population for hepatocellular carcinoma surveillance.

[0045] Overall, this method offers a scalable and clinically significant new approach for the early detection of hepatocellular carcinoma using circulating biomarkers and machine learning models. It enables personalized clinical recommendations and, through integration with traditional serological markers and imaging follow-up procedures, improves clinical workflow.

[0046] Example

[0047] Example 1: Identification of upregulated microRNAs in hepatocellular carcinoma patients

[0048] A total of 522 non-hemolyzed plasma samples were collected from Chongqing and Beijing. In short, in this multicenter cohort study, plasma samples were collected from 522 patients (male:female = 344:178; median age 56 years, range 17–86 years), covering three phases: screening, validation phase 1, and validation phase 2. In the screening phase, microarray analysis was performed on 90 samples (50 from hepatocellular carcinoma patients and 40 from cirrhosis patients) to analyze microRNAs showing differential expression levels between hepatocellular carcinoma and cirrhosis patients. In validation phase 1, reverse transcription real-time quantitative polymerase chain reaction (RT-PCR) was used to validate significant microRNAs selected in the screening phase from 354 paired samples (including 72 samples from the screening phase and an additional 282 samples). Blood samples from both the screening and validation phases 1 were collected from the Second Affiliated Hospital of Chongqing Medical University (N=227) and the General Hospital of the Chinese People's Liberation Army (N=151) between November 2020 and April 2022. Additionally, from September to December 2022, 150 samples were recruited from the Second Affiliated Hospital of Chongqing Medical University for validation phase 2; most of these patients had a history of chronic hepatitis B virus infection. Hepatocellular carcinoma was diagnosed using computed tomography (CT) or magnetic resonance imaging (MRI), with some patients undergoing histopathological interpretation by at least two independent pathologists according to the American Association for the Study of Liver Diseases (AAS). Tumor staging was based on the Barcelona Hepatocellular Carcinoma Staging System, classifying stages 0 and A as early-stage hepatocellular carcinoma and the remaining stages as late-stage hepatocellular carcinoma. All patients with hepatocellular carcinoma had not undergone surgery, chemotherapy, or radiotherapy prior to blood collection. Cirrhosis was diagnosed by ultrasound, and patients with cirrhosis were included in the trial as a high-risk control group. Chronic hepatitis B virus infection was defined as chronic liver disease caused by persistent hepatitis B virus infection (HBsAg positive for >6 months and detectable hepatitis B virus DNA in serum). Patients' clinical characteristics were also documented for subsequent analysis; and all patients signed informed consent forms. This trial was approved by the Ethics Committee of the Second Affiliated Hospital of Chongqing Medical University (2022-80) and followed the principles of the Declaration of Helsinki.

[0049] Peripheral blood from all subjects was collected using 4 mL BD Vacutainer EDTA blood collection tubes (Becton Dickinson) and stored at 4°C. Within 8 hours of collection, plasma was separated by centrifugation at 4°C and 4,000 rpm for 15 minutes. Plasma samples were then aliquoted into sterile microtubes and stored at -80°C until subsequent RNA extraction. Additionally, the absorbance (A414) of all plasma samples was measured at 414 nm using a NanoDrop ND-2000 to assess the degree of hemolysis; plasma samples with an A414 value greater than 0.2 were excluded from subsequent analyses.

[0050] For microarray analysis, microRNA was extracted from 400 μl of plasma using the mirVana microRNA isolation kit according to the manufacturer's instructions. Total RNA was quantified using a NanoDrop ND-2000 instrument, and RNA integrity was assessed using an Agilent Bioanalyzer 2100 instrument. For quantitative polymerase chain reaction (PCR) analysis, microRNA was extracted from 200 μl of plasma using the miRNeasy serum / plasma kit. The extraction process was performed according to the manufacturer's instructions with only minor modifications: 1) To standardize extraction differences, an exogenous control was added to the sample during RNA extraction; 2) To improve RNA extraction efficiency, phage MS2 vector RNA (1 μg per mL) was added to the sample during RNA extraction. The extracted RNA was stored at -80°C for later use.

[0051] Clinical and demographic data of the participants are summarized in Table 1. 354 matched patients were included in Validation Phase 1, after adjusting for confounding factors such as age, sex, and hepatitis B virus status. No sample matching was used in Validation Phase 2 to more accurately reflect actual clinical conditions, resulting in a higher proportion of males, older patients, and patients with hepatitis B virus status.

[0052] Table 1 The data is presented as n (%) or average (SD).

[0053] like Figure 1A As shown, the method for identifying microRNAs associated with hepatocellular carcinoma includes three key stages: a screening stage, a first validation stage, and a second validation stage.

[0054] During the screening phase, the Agilent Human MicroRNA Microarray Kit (version 21.0, 8×60K, ID: 070156) was used, and data analysis was performed by OE Biotechnology Co., Ltd. (Shanghai) following the operating procedures of Agilent's microRNA microarray system and microRNA labeling and hybridization kit, using an Agilent G2505C scanner to scan the chip slides. Raw data were extracted using feature extraction software (version 10.7.1.1, Agilent Technologies). Only microRNAs with at least 50% detectable signal in any one sample group were included for subsequent data analysis, and the included data were normalized using quantiles. Ultimately, a total of 2,549 microRNAs were analyzed, of which 188 microRNAs showed statistically significant differences in expression levels between hepatocellular carcinoma patients and cirrhosis patients (Pad). j <0.05)(see Figure 1B ).

[0055] The results were cross-referenced with the publicly available microRNA expression profiling dataset GSE106817, and a total of 16 microRNAs with consistently upregulated expression levels were identified in different cohorts (see Table 2). In addition, two more microRNAs with literature reports were included for further validation. Therefore, a total of 18 candidate microRNAs were finally selected, including: miR-122-5p, miR-1260b, miR-130a-3p, miR-193a-5p, miR-21-5p, miR-22-3p, miR-24-3p, miR-27a-3p, miR-29a-3p, miR-29c-3p, miR-30d-5p, miR-320b, miR-320c, miR-320d, miR-320e, miR-328-3p, miR-361-5p, and miR-92a-3p (see [link to study]). Figure 1C ).

[0056] In validation phase 1, 18 candidate microRNAs screened from the 354 clinically matched hepatocellular carcinoma and cirrhosis samples were quantified using reverse transcription quantitative polymerase chain reaction (RT-PCR). In short, using the miRCURY LNA real-time reagent kit (Qiagen), 3 μl of microRNA was reverse transcribed in a 10 μl reaction system. The resulting complementary DNA was diluted 20-fold before RT-PCR. The RT-PCR was performed using the miRCURY LNA microRNA probe polymerase chain reaction assay (Qiagen), with amplification conditions of 95°C for 2 minutes of thermal activation, followed by cycles of 95°C for 5 seconds and 56°C for 30 seconds, for a total of 40 cycles. The RT-PCR was performed on a QuantStudio 7Pro real-time polymerase chain reaction system (Applied Biosystems). Undetected reactions were entered as Ct values ​​of 40, and results were normalized using miR-16-5p as an internal control. Relative quantification of microRNAs was performed using a 2-1... -ΔCt Method for calculating (ΔCt=Ct) 目标 -Ct 内参 To ensure consistent quantitative results across reactions, three inter-plate controls were added to each polymerase chain reaction to correct for inter-plate differences, and the results were standardized based on the inter-plate controls.

[0057] MicroRNA expression levels obtained from real-time quantitative polymerase chain reaction (RT-PCR) were standardized using z-scores, and the differences in expression levels were analyzed using a two-tailed Student's t test or a Wilcoxon rank-sum test. A p-value < 0.05 and a fold change > |2| were used as the statistical significance criterion. The p-values ​​were corrected using the Benjamini-Hochberg method for multiple hypothesis testing. In validation phase 1, propensity score matching (PSM) was used, with gender, age, and hepatitis B virus infection status as key confounding factors to screen for matched samples. A total of 177 pairs of matched hepatocellular carcinoma and cirrhosis patients were obtained. To ensure that the sample size was sufficient to identify the degree of difference in microRNA expression levels, statistical power analysis was performed using G*Power software. Considering the average effect size of significantly expressed microRNAs in validation phase 1, the effect size Cohen's D was set to 0.69. The results show that the sample size of 177 pairs exceeded the minimum requirement of 47 pairs required for 80% power and 5% type 1 error.

[0058] To ensure the accuracy of relative quantification, selecting an appropriate internal control is crucial. Among the candidate microRNAs tested, miR-16-5p consistently showed high expression in all samples and was strongly positively correlated with the overall microRNA expression level (Pearson's r = 0.82, P < 0.001). More importantly, miR-16-5p did not show a significant difference in expression levels between the hepatocellular carcinoma group and the cirrhosis group, indicating that it is more suitable as a standardized reference than traditional exogenous controls. After standardization with miR-16-5p, all 18 candidate microRNAs identified in validation phase 1 showed significant upregulation in hepatocellular carcinoma samples, including 16 microRNAs initially identified in the microarray analysis during the screening phase (such as...). Figure 1B These microRNAs (miR-122-5p, miR-320d, miR-1260b, miR-92a-3p, miR-21-5p, miR-130a-3p, miR-22-3p, miR-574-3p, miR-328-3p, miR-29c-3p, miR-27a-3p, miR-361-5p, miR-29a-3p, miR-320e, miR-24-3p, miR-423-5p, miR-148a-3p, miR-320b, miR-30d-5p, miR-193a-5p, and miR-320c) were detected in over 92% of the samples, and except for miR-122-5p, none of them were significantly associated with hepatitis B virus infection status. Notably, miR-361-5p, miR-130a-3p, and miR-24-3p were the most significantly upregulated microRNAs. adj The values ​​are all below 10 -15 Furthermore, unsupervised hierarchical clustering showed that the samples were clustered primarily based on the expression profiles of microRNAs, and were unrelated to alpha-fetoprotein expression levels or cohort origin (e.g., Figure 1C (As shown). These results demonstrate that the 18-microRNA genome has the potential to serve as a robust biomarker for distinguishing hepatocellular carcinoma patients from those with cirrhosis based on molecular performance characteristics.

[0059] Example 2: Validating microRNA targets through meta-analysis

[0060] A total of 221 articles related to circulating microRNAs for the diagnosis of hepatocellular carcinoma were obtained through searches in PubMed and Web of Science. Before December 14, 2023, this embodiment systematically searched PubMed and Web of Science for studies on circulating microRNAs related to hepatocellular carcinoma. The search process used combinations of keywords such as "hepatocellular carcinoma," "circulating microRNAs," "diagnosis," and "hepatitis B virus," along with medical themes. The search was supplemented manually by examining the references of the retrieved articles and related review articles. Only articles published in English and in peer-reviewed journals were included. After removing duplicate results, the abstracts were screened to identify relevant articles. Finally, the inclusion criteria were: (1) original research; (2) clinical studies evaluating the use of circulating microRNAs for the diagnosis of hepatocellular carcinoma; and (3) microRNA analysis studies using serum or plasma as samples. If the following conditions are not met, the study will be excluded: (1) it is not a case-control study; (2) it does not use real-time quantitative polymerase chain reaction as a quantitative method; (3) the control group includes patients with co-infection or viral infection other than hepatitis B virus.

[0061] Data extracted from eligible research literature will include: first author, publication year, region, sample type, sample size, sample characteristics, study design, standardized internal control, type of microRNA detected, dysregulation type, effect size, and p-value. For studies with multiple control groups, only data comparing the control group with the highest risk of hepatocellular carcinoma will be extracted (Hepatitis B virus - cirrhosis > cirrhosis > Hepatitis B virus). Synthetic analyses will include both significant and non-significant results. If effect size data are not directly presented in the literature, performance data will be extracted from the plot using PlotDigitizer. A conservative conversion will be used for the predefined significance p-value thresholds: p ≥ 0.05 and p ≥ 0.01 will be converted to p = 0.5; p < 0.05 will be p = 0.025; p < 0.01 will be p = 0.005; p < 0.001 will be p = 0.0005; p < 0.0001 or p = 0.0000 will be p = 0.00005. All relevant studies were evaluated by two independent researchers, and any disagreements were resolved through consultation with the principal investigator.

[0062] After evaluating the complete literature, 41 articles met the inclusion criteria, disclosing 93 microRNAs with differential expression levels. However, most of these appeared only individually in a single study. Eleven microRNAs had expression data from three or more studies, making them suitable for meta-analysis, including miR-122-5p, miR-21-5p, miR-192-3p, miR-223-3p, miR-27a-3p, miR-26a-5p, miR-29a-3p, miR-29c-3p, miR-193a-5p, miR-125b-5p, and miR-214-3p. Of the 18 microRNAs identified in this invention, five were previously disclosed as being upregulated in hepatocellular carcinoma.

[0063] Specifically, effect sizes and their corresponding variance values ​​(p-value, 95% confidence interval [95% CI], or standard deviation) were extracted from the included studies. If the mean and standard deviation were not provided, they were estimated based on the corresponding variance measure or from the median and interquartile range. When the effect size of a microRNA came from three or more studies, Hedges'g value was calculated as the pooled standardized mean difference (SMD) between hepatocellular carcinoma and cirrhosis patients for meta-analysis. Cochran's Q test and Higgin's inconsistency index (I²) were used to assess heterogeneity among studies; a p-value less than 0.05 in the Cochran's Q test indicated significant heterogeneity. For microRNAs with significant heterogeneity, a DerSimonian-Laird random-effects model was used; otherwise, a fixed-effects model was used. Meta-analysis and heterogeneity testing were performed using the "metafor" data package (version 4.2.0) in R software. The standardized mean difference and 95% CI of the syncytial analysis were presented as a forest plot; studies with extreme effect sizes in the syncytial analysis were identified using Baujat plots and influence analysis. Subgroup analyses were performed based on differences in experimental design. Standardized internal controls commonly used in multiple studies were considered as common internal controls, with RNU6 as the independent group and the rest classified as uncommon internal controls. Potential publication bias was assessed by visually examining funnel plots and Egger's test. Unless otherwise stated, two-sided p-values ​​less than 0.05 were considered statistically significant. Bonferroni multiple test adjustment was used for the significance of the syncytial analysis (α = 0.05 / 11 = 4.55 × 10⁻⁶). -3 ).

[0064] Among the identified microRNAs, miR-122-5p and miR-21-5p were the most frequently reported and thus served as prime candidates for the meta-analysis. However, a preliminary meta-analysis incorporating all available research revealed significant heterogeneity in both microRNAs, rendering the results inconclusive. No evidence of publication bias was detected for either microRNA. Baujat plot analysis identified the primary sources of heterogeneity, mainly attributable to studies using less commonly used standardized internal controls. Specifically, the study by Xu et al. contributed significantly to the heterogeneity of miR-122-5p, while the study by Guo et al. contributed significantly to the heterogeneity of miR-21-5p.

[0065] To address this issue, research literature was divided into multiple subgroups based on the standardized internal controls used. Subgroup differentiation significantly reduced heterogeneity within each group. In studies using miR-16-5p or exogenously added internal controls, miR-122-5p was observed (effect size = 0.66, 95% CI: 0.53–0.78; P = 1.17 × 10⁻⁶). -24 Both miR-21-5p (effect size = 0.41, 95% CI: 0.04–0.78; P = 0.028) showed consistent upregulation.

[0066] When the same subgroup analysis method was applied to other microRNAs, miR-192-3p and miR-29a-3p, which showed upregulated expression levels, were identified in studies using conventional standardized internal controls; while in studies using mixed or non-standardized internal controls, one microRNA (miR-193a-5p) with consistent upregulated expression levels and two microRNAs (miR-26a-3p and miR-223-3p) with consistent downregulated expression levels were identified.

[0067] In summary, the combined analysis results validated the four previously identified upregulated microRNAs (miR-122-5p, miR-21-5p, miR-29a-3p, and miR-193a-5p), thus further confirming the robustness and reproducibility of identifying dysregulated microRNAs in hepatocellular carcinoma using microRNA expression profiling.

[0068] Example 3: Establishment and validation of a combination of 5-microRNA genome and alpha-fetoprotein 20 for detecting hepatocellular carcinoma.

[0069] First, univariate logistic regression analysis was used to evaluate the detection performance of individual microRNAs in differentiating hepatocellular carcinoma from cirrhosis. In short, matched samples from validation phase 1 were randomly assigned to a 75% training set and 25% test set. In the training set, LASSO regression analysis was performed after 100 replicates of 5-fold cross-validation using the caret software package to establish an optimized microRNA group containing alpha-fetoprotein 20. The microRNA group with the highest area under the curve (AUC) was selected as the final microRNA group and further validated in the test set and additional validation sets. Multicollinearity among variables within the group was detected using the variance inflation factor (VIF), with a score less than 5 considered acceptable. Risk scores were calculated for all patients and defined as the sum of the expression levels of each microRNA multiplied by its corresponding logistic regression coefficient in the model. The Youden index was used to determine the cutoff value for the risk scores, classifying patients into high-risk and low-risk groups. Subgroup analyses were further performed in multiple sample subsets to evaluate the robustness and detection performance of the established microRNA groups. The detection performance, sensitivity, and specificity of the model were analyzed and evaluated using the area under the curve (AUC) and receiver operating characteristic (ROC) curves. PSM was performed using the `matchit` function from the `MatchIt` package, and the Delong test from the `pROC` package was used for statistical analysis and comparison of the ROC curves. All statistical analyses were performed in R software (version 4.4.1).

[0070] Among the 18 upregulated microRNAs, miR-361-5p had the highest area under the curve (AUC) at 0.801, while alpha-fetoprotein (AFP) 20 had an AUC of 0.767. Several top-ranking microRNAs exhibited detection performance comparable to AUC 20, while the overall accuracy of the remaining microRNAs was lower. Notably, none of the 18 microRNAs showed a significant correlation with AFP expression (Pearson correlation, all P values ​​were greater than 0.05), suggesting that the expression of these microRNAs and AFP may occur through independent biological pathways. This independence supports the possibility that circulating microRNAs can serve as complementary biomarkers to AFP in the detection of hepatocellular carcinoma.

[0071] To further improve detection accuracy, this embodiment employs an interpretable machine learning approach to integrate multiple microRNA signals. A Logistic Minimum Absolute Contraction and Selection Operator (LASSO) regression model was trained using 75% of 354 matched samples to identify an optimized microRNA set comprising miR-361-5p, miR-130a-3p, miR-27a-3p, miR-30d-5p, and miR-193a-5p, which can be combined with the alpha-fetoprotein 20 index. The area under the curve for this combination on the training set was 0.873 (95% CI: 0.831–0.915) (see [link to training data]). Figure 2A In the remaining 25% of the test set, the area under the curve remained at 0.879 (95% CI: 0.809–0.950) (see...). Figure 2B Its robustness was further evaluated and confirmed in an independent validation set, and the results showed that its area under the curve reached 0.957 (95% CI: 0.924-0.990) (see...). Figure 2C In contrast, using alpha-fetoprotein 20 alone performed poorly across the datasets (Area under the curve for training set = 0.747 [95% CI: 0.698–0.795]; Area under the curve for test set = 0.790 [95% CI: 0.708–0.870]; Area under the curve for validation set = 0.786 [95% CI: 0.687–0.886]; see [link to relevant documentation]). Figure 2A-2C ).

[0072] Further subgroup analyses were conducted within different clinically relevant hepatocellular carcinoma subgroups, including early-stage hepatocellular carcinoma, hepatitis B virus-associated hepatocellular carcinoma, hepatitis B virus-negative hepatocellular carcinoma, and alpha-fetoprotein-negative hepatocellular carcinoma. The results showed that the combined group (microRNA group + alpha-fetoprotein 20 indicator) outperformed alpha-fetoprotein 20 indicator alone in all subgroups. Notably, the combined group also outperformed continuous detection of alpha-fetoprotein expression in the training set. Table 2 summarizes the detection indicators for the overall and early-stage hepatocellular carcinoma.

[0073] Table 2. Comparison of the detection performance of the combined group (microRNA group + alpha-fetoprotein 20 index) and the alpha-fetoprotein 20 index in identifying patients with hepatocellular carcinoma or early-stage hepatocellular carcinoma in the test and validation sets. Early-stage hepatocellular carcinoma was included in patients at Barcelona hepatocellular carcinoma stage 0 and A. Comparisons were made within the same patient group. Samples lacking alpha-fetoprotein (AFP) expression were excluded. The area under the curve (AUC) for the combined group and the AFP 20 index was compared using the DeLonghi test.

[0074] Further calculation of the combined risk score for each subject was performed to stratify the subjects' risk. Results showed that the risk score gradually increased with the progression of hepatocellular carcinoma stage (e.g., ...). Figure 2D No significant differences were observed when stratified by age, sex, or hepatitis B virus status. Using the Youden index as the classification threshold (0.5572), both the combined group and the alpha-fetoprotein 20 (AFP20) assay showed specificities exceeding 90% in the detection of early-stage hepatocellular carcinoma. However, compared to using AFP20 alone, the combined group exhibited significantly improved sensitivity (63% vs. 50%), highlighting the enhanced screening efficiency and early-stage hepatocellular carcinoma detection rate when microRNA assays are used in combination with AFP20.

[0075] Example 4: Association analysis of microRNAs with clinical parameters and related genes of their targets

[0076] Further analysis was conducted to determine the association between microRNA expression levels and key tumor-related clinical features in hepatocellular carcinoma patients. Three platforms—TargetScan 8.0, miRDB, and miRWalk—were used to predict the genes targeted by microRNAs. The target genes predicted by all three platforms were then subjected to functional enrichment analysis in the Kyoto Encyclopedia of Genetics and Genomes (KEGG) database. Interaction networks for three or more microRNAs predicting common target genes were constructed using Cytoscape v3.10.2.

[0077] Consistent with the excellent detection performance of the microRNA genome provided by this invention for early-stage hepatocellular carcinoma, a total of 17 microRNAs showed significant upregulation in patients with early-stage hepatocellular carcinoma (see...). Figure 3A This further supports its potential application value as a biomarker for early detection. In addition, 11 microRNAs showed increased levels in patients with larger tumor volumes (see...). Figure 3B Furthermore, five microRNAs showed increased levels in cases of tumor invasion (see...). Figure 3C ).

[0078] Among the five microRNAs in the microRNA group provided by this invention, three (miR-130a-3p, miR-361-5p, and miR-27a-3p) showed positive correlations with tumor stage and tumor volume, indicating that these microRNAs are not only related to the detection of hepatocellular carcinoma, but may also have clinical significance in disease stratification.

[0079] To explore the underlying molecular mechanisms of these microRNA dysregulations, three bioinformatics tools—TargetScan 8.0, miRDB, and miRWalk—were used to predict the genes associated with their targets. Repeated target-related genes identified by the three platforms were integrated to construct an interaction network (e.g., Figure 3D (As shown). Pathway enrichment analysis of these repetitive target genes, according to the Kyoto Encyclopedia of Genes and Genomes, revealed significant enrichment in multiple key cancer signaling pathways, including the PI3K-Akt, MAPK, and hepatocellular carcinoma pathways (see...). Figure 3E This indicates that these dysregulated microRNAs can play a functional role in tumor progression through the aforementioned signaling cascade.

[0080] This invention represents a significant advancement in hepatocellular carcinoma (HCC) detection in high-risk populations using liquid biopsy to detect circulating microRNAs associated with HCC, including alpha-fetoprotein (AFP) 20. Compared to existing standard screening methods that rely solely on AFP, this invention significantly improves detection performance. Because AFP has limited sensitivity, especially in early-stage HCC, using AFP alone as a screening criterion often leads to delayed detection and poor clinical outcomes.

[0081] In this invention, 18 circulating microRNAs were significantly upregulated in hepatocellular carcinoma (HCC) patients, with four of them further validated through comprehensive meta-analysis. Based on this, this invention provides a microRNA group containing five specific microRNAs (miR-361-5p, miR-130a-3p, miR-27a-3p, miR-30d-5p, and miR-193a-5p), which can be used in conjunction with or alone as an alpha-fetoprotein (AFP) marker. This microRNA group demonstrated excellent performance in differentiating HCC from cirrhosis (including early-stage HCC). Notably, this microRNA group can still be used independently for HCC detection even in the absence of AFP data. This invention represents one of the largest multicenter, high-throughput studies of circulating microRNA detection to date, highlighting the potential of liquid biopsy using microRNAs as a minimally invasive tool for early cancer detection. While this method is not intended to replace imaging techniques, its use as a first-line screening method in conjunction with AFP markers can significantly enhance HCC surveillance.

[0082] Among the microRNAs analyzed, miR-122-5p and miR-21-5p were the most frequently observed in previous literature and meta-analyses, further enhancing their detection potential. As a liver-specific microRNA, miR-122-5p has been extensively studied, but its expression levels in the bloodstream of hepatocellular carcinoma (HCC) patients have yielded conflicting results. Meta-analysis results indicate that this discrepancy is primarily due to different standardization strategies used in various studies. Furthermore, circulating miR-122-5p expression is not specific to HCC and is also associated with other chronic liver diseases (including hepatitis B virus infection, hepatitis C, non-alcoholic fatty liver disease, and drug-induced liver injury); conversely, in HCC tissue, the expression levels of miR-122-5p in hepatocytes are consistently downregulated.

[0083] This invention reveals an inverse relationship between the expression level of circulating miR-122-5p and the expression level of miR-122-5p in liver tissue. Upregulation of circulating miR-122-5p showed almost no correlation with traditional liver injury markers, such as aspartate aminotransferase (AST) or alanine aminotransferase (ALT) (AST: r = 0.10, P = 0.03; ALT: r = 0.03, P > 0.05), suggesting that its elevation in the blood is not solely due to hepatocyte death. A similar phenomenon has been observed in non-alcoholic fatty liver disease; Pirola et al. proposed that the decrease in miR-122-5p expression level in liver tissue may stem from its active efflux into the circulatory system.

[0084] Other key microRNAs in the microRNAe of this invention (including miR-361-5p, miR-130a-3p, and miR-27a-3p) have not been widely disclosed in previous hepatocellular carcinoma studies. Notably, miR-27a-3p has been shown to mediate transgenerational susceptibility to hepatocellular carcinoma in animal models; for example, upregulation of serum miR-27a-3p expression during pregnancy in obese individuals can increase offspring liver expression and the risk of hepatocellular carcinoma, highlighting its potential role in tumorigenesis.

[0085] To achieve standardization, this invention selected miR-16-5p as an internal control due to its stable and high expression levels across the entire study cohort. Meta-analysis results also showed that using miR-16-5p as an internal control yielded consistent quantitative results. However, it should be noted that some studies have reported downregulation of miR-16-5p expression levels in hepatocellular carcinoma patients, indicating a need for further evaluation and consensus on the optimal standardization strategy for circulating microRNA assays.

[0086] In summary, this invention systematically provides and validates a microRNA matrix that can be used in combination with the alpha-fetoprotein 20 (AFP20) indicator. The combined use of both demonstrated excellent performance in various clinical subgroups (including early-stage hepatocellular carcinoma, hepatitis B virus-associated hepatocellular carcinoma, and AFP-negative hepatocellular carcinoma), highlighting the potential value of this microRNA matrix as a reliable, non-invasive, and scalable tool for population-based hepatocellular carcinoma screening and early cancer detection.

[0087] According to various embodiments of the present invention, the functional units and modules of the system and method can be implemented by computing devices, computing processors, or electronic circuits, including but not limited to application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), microcontrollers, central processing units (CPUs), graphics processing units (GPUs), and other programmable logic devices configured or programmed according to the teachings of the present invention. Those skilled in the art can readily develop computer instructions or software code executable in computing devices, processors, or programmable logic devices based on the disclosure of the present invention.

[0088] All or part of the steps of the methods in the various embodiments of the present invention can be executed on one or more computing devices, including server computers, personal computers, laptops, smartphones, tablets and other mobile computing devices.

[0089] The embodiments may include computer storage media, volatile or non-volatile storage devices storing computer instructions or software code, which can be used to program or configure the aforementioned computing device, processor, or electronic circuit to perform any of the processes of the present invention. The storage media and storage devices may include, but are not limited to, floppy disks, optical disks, Blu-ray discs, DVDs, CD-ROMs, magneto-optical disks, read-only memory (ROM), random access memory (RAM), flash memory devices, or any medium or device suitable for storing instructions, code, and / or data.

[0090] According to various embodiments, each functional unit and module may also be implemented in a distributed computing environment and / or cloud computing environment, wherein all or part of the machine instructions are executed in a distributed manner by one or more processing devices interconnected by a communication network, including but not limited to: intranet, wide area network (WAN), local area network (LAN), Internet and other forms of data transmission media.

[0091] The above description is provided for the purpose of illustrating and describing the invention, and is not intended to be exhaustive or to limit the invention to the precise forms disclosed. Many modifications and variations will be apparent to those skilled in the art.

[0092] The above embodiments were chosen and described in order to best explain the principles of the invention and its practical application, thereby enabling other skilled in the art to understand the various embodiments of the invention and the various modifications suitable for the intended particular use.

Claims

1. A system for detecting a hepatocellular carcinoma microRNA panel, comprising: comprising: a sample preparation module for processing a blood sample of a subject and separating a plasma fraction and a serum fraction therefrom; a nucleic acid extraction module for extracting a microRNA expression profile from the plasma fraction; a real-time polymerase chain reaction module for detecting and quantifying expression levels of a microRNA panel in the microRNA expression profile, the microRNA panel comprising miR-361-5p, miR-130a-3p, miR-27a-3p, miR-30d-5p, and miR-193a-5p; and a computing device comprising a processor and a memory, wherein the memory has instructions stored therein that, when executed by the processor, perform the steps of: inputting quantitative values of the expression levels of the microRNA panel; processing the input using a trained classifier to generate a hepatocellular carcinoma risk classification result from the expression levels of the microRNA panel to rank the subject according to hepatocellular carcinoma risk.

2. The system of claim 1, wherein the hepatocellular carcinoma risk comprises a low hepatocellular carcinoma risk, a hepatocellular carcinoma moderate risk, and a high hepatocellular carcinoma risk.

3. The system of claim 1, wherein the computing device further: processes detected and quantified alpha-fetoprotein expression levels from the serum fraction; and compares the alpha-fetoprotein expression levels to a threshold value of 20 ng / mL. regardless of whether the alpha-fetoprotein expression levels are below or above the threshold value, the trained classifier inputs the alpha-fetoprotein expression levels and the expression levels of the microRNA panel as two input values and generates the hepatocellular carcinoma risk classification result; wherein, in the absence of the alpha-fetoprotein expression levels, the classifier generates the hepatocellular carcinoma risk classification result based only on the expression levels of the microRNA panel.

4. The system of claim 3, wherein, 5. The system of claim 1, wherein the trained classifier comprises a logistic regression model, wherein the logistic regression model is trained on an annotated training dataset comprising microRNA expression profiles and true labels of hepatocellular carcinoma diagnoses.

6. The system of claim 1, wherein the system further comprises a training module to update the trained classifier according to a new annotated training dataset.

7. The system of claim 1, wherein the subject is a patient with cirrhosis. comprising:

8. A computer-based method for detecting microRNA genomes in hepatocellular carcinoma, characterized in that, providing a blood sample of a subject; separating a plasma fraction from the blood sample; extracting a microRNA expression profile of the subject from the plasma fraction; detecting and quantifying expression levels of a microRNA panel in the microRNA expression profile, wherein the microRNA panel comprises miR-361-5p, miR-130a-3p, miR-27a-3p, miR-30d-5p, and miR-193a-5p; processing quantitative values of the expression levels of the microRNA panel using a trained machine learning classifier on a computer processor, the classifier generating a hepatocellular carcinoma risk classification result from the expression levels of the microRNA panel to rank the subject according to hepatocellular carcinoma risk. ​ 9. The method of claim 8, further comprising: isolating a serum fraction from the blood sample and detecting and quantifying an alpha-fetoprotein expression level therein; and comparing the alpha-fetoprotein expression level to a threshold of 20 ng / mL.

10. The method of claim 9, wherein, regardless of whether the alpha-fetoprotein expression level is below or above the threshold, the trained classifier takes the alpha-fetoprotein expression level and the microRNA panel expression level as two input values and generates the hepatocellular carcinoma risk classification result; wherein when the alpha-fetoprotein expression level is absent, the classifier generates the hepatocellular carcinoma risk classification result based on the microRNA panel expression level only.

11. The method of claim 8, wherein the subject is a patient with cirrhosis.

12. The method of claim 8, wherein the trained classifier comprises a logistic regression model, wherein the logistic regression model is trained with an annotated training dataset.

13. The method of claim 8, wherein the hepatocellular carcinoma report comprises hepatocellular carcinoma risk classification information, wherein the hepatocellular carcinoma risk classification result further comprises one or more recommended actions comprising repeat testing, multiphase contrast-enhanced magnetic resonance imaging examination, and oncology consultation.

Images