Clinical and Molecular Prognostic Markers for Liver Transplantation

Abstract

The present invention relates to a method for predicting the outcome of liver transplantation for the treatment of hepatocellular carcinoma (HCC) or a method for allocating HCC patients to receive liver transplantation, the method comprising the steps of determining the expression level of an indicator gene selected from dermatopontin, clusterin, calpain small subunit 1, F-box and WD repeat-containing protein 7 and Sprouty RTK signaling antagonist 2 in a liver sample obtained from a patient suffering from HCC, and comparing the expression level of the indicator gene with the expression level of an internal control gene to predict a positive prognosis for the liver transplant recipient using a linear support vector machine algorithm in combination with the variable total tumor volume.

Description

[Technical Field]

[0001] This application claims priority to European Patent Application No. EP19201218.5 dated 2 October 2019, which is incorporated herein by reference.

[0002] This invention relates to a method for determining the prognosis or potential positive outcomes of patients who are scheduled to undergo or are considering liver transplantation during the course of treatment for hepatocellular carcinoma. Compared to current criteria for patient selection, the selection of suitable candidates for liver transplantation can be improved by integrating both clinical and molecular markers. This application will contribute to improving the selection of patients who require liver transplantation due to hepatocellular carcinoma. [Background technology]

[0003] explanation Hepatocellular carcinoma (HCC) is a common disease with a significant impact on mortality and quality of life. Understanding its multifactorial and complex etiology is crucial for providing information for stratifying patients in current treatments and for designing personalized medicine optimized for each patient's genome. Liver transplantation (LT) is the best treatment for HCC in cirrhosis, but organ availability is limited due to the high likelihood of poor outcomes for this disease. Furthermore, expanding current criteria is expected to lead to a further increase in demand for transplants, thus lengthening pre-transplant waiting times, increasing dropout rates, and worsening intention-to-treat outcomes. Even when organs are available, the benefits of LT must be balanced against the donor risks.

[0004] Therefore, optimal selection of HCC patients is essential. The current clinical morphology model, namely the Milan Criteria (MC), may exclude excellent candidates, such as patients with tumors discovered late in the disease course that are progressive but currently "benign." Conversely, the MC may not exclude early HCC patients characterized by highly invasive tumor behavior. Identifying the best candidates for liver transplantation in HCC patients in the context of cirrhosis could potentially increase the number of candidates for liver transplantation (LT) and thus impact the demand for the donor pool.

[0005] Molecular biomarkers can provide relevant information about the biological behavior of tumors. However, currently, there are few studies dealing with patient HCC molecular biomarkers, and therefore no available consensus-based prognostic biomarkers that would enable appropriate patient selection for transplantation. Since criteria based on tumor morphology can only partially identify patients with good / poor prognoses, we propose that it is important to focus on genes or genetic signatures that influence prognosis. Of these, genes expressed in HCC patients with good prognoses (i.e., genes whose upregulation leads to good outcomes) can complement currently accepted criteria for selecting subgroups of patients who are likely to benefit from LT. Of course, negative predictor genes also play an important role, as they can identify patients who are likely to experience disease relapse.

[0006] Dermatopontin (DPT) DPT, also known as TRAMP (tyrosine-rich acidic matrix protein) (gene ID GC01M168664), is an extracellular matrix protein that can function in cell-matrix interactions and matrix assembly. This protein is present in various tissues and is thought to be expressed in mesenchymal cells (fibroblasts and myofibroblasts) and macrophages. This molecule is crucial for extracellular matrix assembly, cell adhesion, and wound healing. It also accelerates collagen fibrillation and modifies the behavior of TGF-beta through interaction with decorin in the extracellular matrix microenvironment in vivo. DPT can inhibit the formation of the decorin-TGF-beta1 complex, increasing the cellular response to TGF-beta and enhancing its biological activity. Furthermore, DPT has been identified as a downstream target of the vitamin D receptor. The vitamin D receptor then mediates downstream signaling of 1,25-dihydroxyvitamin D3, exerting an antiproliferative effect against HCC. DPT correlates with cell adhesion and tumor invasion. Positive expression of DPT is associated with suppression of metastasis in oral cancer and giant cell tumors of bone. Downregulation of DPT is associated with the oncogenesis and progression of HCC through possible interactions with TGF-beta1 and other potential mechanisms. DPT expression levels are significantly lower in HCC tissue than in healthy liver.

[0007] Calpain small subunit 1 (CAPNS1) Calpain, a family of proteases, is involved in cell migration and invasion by altering the structure of cell adhesion molecules and cytoskeletal components, or by interfering with intracellular signaling pathways. Its regulatory subunits, CAPNS1 and CAPN4, may play important roles in calpain activity. Knockdown of CAPNS1 in malignant endothelial cells can reduce their ability to spread, while upregulation of CAPNS1 correlates with increased tumor size, number, and alpha-fetoprotein (AFP) levels after HCC resection in animals.

[0008] Clusterin (CLU) CLU is a protein encoded by the CLU gene chaperone (gene ID GC08M027596) and exists in both cytoplasmic and secretory forms. When CLU forms a complex with EIF3I, it activates the Akt pathway, which in turn promotes the metastasis of HCC cells.

[0009] F-box and WD repeat containing protein 7 (FBXW7, gene ID GC04M152321) The gene “F-box and WD repeat domain-containing protein 7 (FBXW7),” also known as Sel10, hCDC4, or hAgo, encodes a member of the F-box protein family that functions as a substrate recognition component of SCF E3 ubiquitin ligase. FBXW7 protein is an important tumor suppressor and one of the most commonly disregulated ubiquitin-proteasome system proteins in human cancers. FBXW7 regulates the proteasome-mediated degradation of oncogeneic proteins such as cyclin E, c-Myc, Mcl-1, mTOR, Jun, Notch, and AURKA. Mutations in this gene have been detected in ovarian and breast cancer cell lines, suggesting a potential role for this gene in the pathogenesis of human cancer. In vitro studies have shown that FBXW7 can be used as a prognostic marker for hepatocellular carcinoma, with low levels of FBXW7 expression associated with reduced survival in HCC patients. Thus, FBXW7 plays a crucial role in the progression of HCC. In other words, FBXW7 inhibits the migration and invasion of HCC cells via the Notch1 signaling pathway.

[0010] Sprouty RTK signaling antagonist 2 (SPRY2, gene ID GC13M080335) The Drosophila Spry (dSPRY) gene family, which includes four homologs SPRY1-4, is thought to be involved in the negative feedback loop of the RAF / MEK / ERK pathway associated with HCC carcinogenesis. In particular, SPRY2 antagonizes growth factor-mediated cell proliferation, migration, and differentiation by regulating receptor tyrosine kinase (RTK) signaling and repressing the RAF / MEK / ERK pathway. This protein is a key regulator of essential pathways involved in cancer development, including angiogenesis, cell proliferation, invasion, migration, and cytokinesis. [Overview of the Initiative] [Problems that the invention aims to solve]

[0011] Based on the cutting-edge technology described above, the objective of the present invention is to provide a molecular marker that helps in accurately selecting the best candidate for LT from among all HCC patients. [Means for solving the problem]

[0012] Molecular markers need to be able to predict tumor behavior and invasiveness. This objective is achieved by the claims herein, and further advantageous embodiments are provided herein in whole. [Brief explanation of the drawing]

[0013] [Figure 1] Figure 1 shows the RT-qPCR results of the differences in expression of selected genes corresponding to relapse in the validation set. Relapse present (R: left) and no relapse (noR: right). Wilcoxon (Mann-Whitney) test (confidence level = 0.95). [Figure 2] Figure 2 shows the receiver operating characteristic (ROC) analysis for DPT expression and tumor recurrence. The area under the curve (AUC) is 0.77. A ΔCt cutoff value of 7 was selected, corresponding to 84% sensitivity and 63% specificity. [Figure 3] Figure 3 shows disease-free survival (DFS) based on DPT expression. [Figure 4A] Figure 4 shows the survival rate based on DPT expression in patients both within and outside the Milan criteria (MC). [Figure 4B] Figure 4 shows the survival rate based on DPT expression in patients both within and outside the Milan criteria (MC). [Figure 5A] Figure 5 shows disease-free survival based on DPT expression in patients with total tumor volume (TTV) exceeding 115 cm³ (A), poor differentiation (B), or microvascular infiltration (C). [Figure 5B]Figure 5 shows disease - free survival by DPT expression in patients with total tumour volume (TTV) exceeding 115 cm3 (A), poor differentiation (B), or microvascular invasion (C). [Figure 5C] Figure 5 shows disease - free survival by DPT expression in patients with total tumour volume (TTV) exceeding 115 cm3 (A), poor differentiation (B), or microvascular invasion (C). [Figure 6] Figure 6 shows survival by DPT expression in a subgroup of HCV patients. [Figure 7] Figure 7 is a figure showing disease - free survival by CLU expression. [Figure 8A] Figure 8 is a figure showing disease - free survival by CLU expression in the "poor prognosis" subgroup. A: Patients exceeding MC.<00已翻译内容00088> [Figure 8B] Figure 8 is a figure showing disease - free survival by CLU expression in the "poor prognosis" subgroup. B: Patients with TTV exceeding 115 cm3. [Figure 8C] Figure 8 is a figure showing disease - free survival by CLU expression in the "poor prognosis" subgroup. C: Patients with microvascular invasion. [Figure 8D] Figure 8 is a figure showing disease - free survival by CLU expression in the "poor prognosis" subgroup. D: Patients with poorly differentiated tumours. [Figure 9] Figure 9 is a figure showing survival by CLU expression in hepatitis C virus (HCV) patients. [Figure 10] 已翻译内容 Figure 10 is a figure showing disease - free survival by the combined DPT / CLU score. [Figure 11A] Figure 11 shows DFS using the combined gene score in patients outside MC (A), with TTV exceeding 115 cm3 (B), with poorly differentiated tumours (C), and with microvascular invasion (D). [Figure 11B] Figure 11 shows DFS using the combined gene score in patients outside MC (A), with TTV exceeding 115 cm3 (B), with poorly differentiated tumours (C), and with microvascular invasion (D). [Figure 11C]Figure 11 shows the disease-free survival (DFS) using a composite gene score in patients with extra-MC (A), TTV greater than 115 cm³ (B), poorly differentiated tumor (C), and microvascular invasion (D). [Figure 11D] Figure 11 shows the disease-free survival (DFS) using a composite gene score in patients with extra-MC (A), TTV greater than 115 cm³ (B), poorly differentiated tumor (C), and microvascular invasion (D). [Figure 12] Figure 12 shows the DFS of the composite gene score in patients with a TTV of 115 cm³ or less. [Figure 13] Figure 13 shows disease-free survival based on the expression of DPT and CLU separately or in combination. Compared to CLU, DPT shows superior function in predicting long-term survival. [Figure 14] Figure 14 shows the confusion matrix of the prediction algorithm. [Figure 15] Figure 15 shows the Kaplan-Meier curves for disease-free survival based on an algorithm using CLU, DPT, and TTV values ​​for the entire cohort population. [Figure 16] Figure 16 shows the Kaplan-Meier curves for disease-free survival based on an algorithm using CLU, DPT, and TTV values ​​for patients outside the Milan criteria. [Figure 17] Figure 17 shows the Kaplan-Meier curves for disease-free survival based on an algorithm using CLU, DPT, and TTV values ​​for patients within the Milan criteria. [Modes for carrying out the invention]

[0014] The inventors improved the clinical criteria for selecting patients with liver cirrhosis (LT) in HCC and evaluated the role of selected biomarkers in the population. They examined the role of current clinical markers to specifically determine the best candidates for LT in subgroups of patients exceeding the current absolute clinical criteria of the Milan criteria, integrating both clinical and molecular features to accurately address tumor biology.

[0015] Using clinical data from public repositories, a systematic review of the subject of molecular prognostic biomarkers for HCC, and in-house clinical data available to the inventors, a set of potential candidate genes was identified. The inventors identified single-gene changes using prognostic values ​​and combined them to create predictive multivariate signatures. The systematic review allowed for the identification of genes associated with HCC progression that were expected to provide prognostic information after LT, thereby contributing to patient selection for LT.

[0016] The inventors hereby demonstrate that DPT, CLU, CAPNS1, FBXW7, and SPRY2 show differential expression between patients with and without HCC recurrence after LT. Furthermore, DPT and CLU effectively identify a subgroup of patients who do not experience HCC recurrence after LT and therefore have a positive prognosis, either individually or in combination.

[0017] The terms “gene expression” or “expression,” or “gene product,” can refer to the process of generating nucleic acids (RNA) or peptides or polypeptides, and either or both of their products, which also refer to any intermediate processes that regulate the transcription and translation, or the processing of genetic information, for the production of polypeptide products. The term “gene expression” can also be applied to the transcription and processing of RNA gene products, e.g., regulatory RNA or structural (e.g., ribosomal) RNA. When the expressed polynucleotides originate from genomic DNA, expression may include the splicing of mRNA in eukaryotic cells. Expression can be assayed at the transcriptional and translational levels, i.e., both mRNA and / or protein products.

[0018] In the context of this invention, the term "favorable prognosis" refers to the absence of recurrence of HHC disease within 5 years after LT. Favorable prognosis is measured in the examples by its direct correlation with patient survival, and therefore overall survival (OS) or disease-free survival (DFS) are substantially synonymous.

[0019] In the context of this invention, the terms “support vector machine (SVM),” “linear kernel SVM,” or “SVM algorithm” refer to a supervised machine learning model that enables the classification and / or regression analysis of data. This is sometimes called a support vector network and, in the context of this invention, is used as a form of binary linear classification algorithm. In the context of this invention, the algorithm uses a training step to build a model in which samples of patient data are associated with a set of variables, including but not limited to gene expression levels or tumor volume measurements, so that the algorithm can assign samples to one category or another, for example, surviving or not surviving at 5 years.

[0020] In one embodiment, the present invention relates to a method for predicting outcomes of liver transplantation for the treatment of HCC. Alternative to this embodiment, the present invention relates to a method for treating HCC by liver transplantation. In another alternative embodiment, the present invention relates to a method for stratifying HCC patients into different groups of those who are more or less likely to benefit from receiving a liver transplant, in other words, into different groups of those who are more likely to show longer overall survival after LT without disease recurrence.

[0021] A method according to this aspect of the present invention comprises a determination step, wherein the expression level (particularly mRNA level) of a gene biomarker or indicator gene in a liver sample (also referred to herein as indicator gene expression level) is obtained from a patient suffering from HCC, and the indicator gene is: a. Dermatopontin b. Clasterin c. Calpain small subunit 1 d. F-box and WD repeat-containing protein 7 e. Sprouty RTK signaling antagonist 2 Selected from a list that includes this item.

[0022] In certain embodiments, the indicator gene is CLU or DPT.

[0023] Alternative to this first aspect of the present invention is a method for predicting the outcome of liver transplantation for the treatment of hepatocellular carcinoma (HCC), wherein the method is: - In the determination process, the expression levels of each of the group of indicator genes, including CLU and DPT, in liver samples obtained from patients suffering from HCC, - In the classification process, a step of assigning patients to a favorable prognosis as an outcome for liver transplantation based on the expression level of the indicator gene, Includes.

[0024] In a particular embodiment, the group of indicator genes is: - Calpain small subunit 1 - F-box and WD repeat-containing protein 7 - Sprouty RTK signaling antagonist 2 It further includes at least one of the following.

[0025] In certain embodiments, overexpression of any of the indicator genes is associated with a favorable prognosis.

[0026] In certain embodiments, overexpression of DPT and / or CLU leads to a favorable prognosis.

[0027] In certain embodiments, the indicator gene is overexpressed above a threshold. In alternative embodiments, the expression level of the indicator gene is compared to a control sample selected from representative patients for a subset of outcomes for each disease.

[0028] In certain embodiments, the expression level of an indicator gene is determined by using a quantitative polymerase chain reaction (PCR) sensitive to the level of mRNA encoding the indicator gene present in the sample. Specific primers for amplifying the target indicator gene region identified by the present invention are provided in Table 6, and these can be used in particular to determine the expression level of the indicator gene according to the method provided by the present invention. Global RNA sequencing is an alternative method that can generate the gene expression levels used according to the present invention.

[0029] The expression level of an indicator gene can be compared to the expression level of internal control genes, particularly housekeeping genes. In certain embodiments, the expression level of an indicator gene is compared to the expression level of ribosomal protein L13A (RPL13A) (gene ID 23521). Other possibilities include other housekeeping genes, or combinations of genes, such as GADPH and / or TBP.

[0030] In a particular embodiment, the expression level of an indicator gene is determined by PCR. The expression value of the indicator gene relative to the threshold is determined as the difference between the threshold cycle number of the indicator gene and the threshold cycle number of the internal control gene. The threshold cycle number is the number of PCR cycles in which the product (the indicator gene and the internal control gene) is detected.

[0031] In certain embodiments, the following calculations are used to generate values ​​that reflect the expression levels of indicator genes: Target quantity = ΔCt ΔCt(R) = Ct(DPT in R) - Ct(reference gene in R) ΔCt(nonR) = Ct(nonR DPT) - Ct(nonR reference gene) In the formula, R represents samples with disease recurrence, and nonR represents samples without disease recurrence.

[0032] In certain embodiments, the indicator gene is DPT, and its threshold is a difference in gene expression greater than 7, or ΔCt.

[0033] In certain embodiments, the indicator gene is CLU, and its threshold is a difference in gene expression greater than -0.54, or ΔCt.

[0034] In certain embodiments, the difference in the threshold cycle number of an indicator gene is said to indicate gene overexpression if it exceeds that threshold: The ΔCt value for DPT expression levels is higher than 7, and the ΔCt value for CLU expression levels is higher than -0.54.

[0035] The thresholds provided herein are examples of ΔCt values ​​that are useful for classifying gene expression levels according to the present invention. These thresholds reflect gene expression with reference to housekeeping genes. Gene expression levels with positive values ​​indicate that the gene is expressed more than the housekeeping reference gene, while negative values ​​indicate that the gene is expressed less than the housekeeping reference gene. Overexpression of an indicator gene is defined as expression exceeding the ΔCt threshold for separating patients with a poor prognosis from patients with a good prognosis.

[0036] In a particular embodiment, 115 cm 3 DPT overexpression accompanied by the determination of total tumor volume as described below indicates a favorable prognosis.

[0037] In a particular embodiment, 115 cm 3 CLU overexpression accompanied by the determination of total tumor volume as described below indicates a favorable prognosis.

[0038] In certain embodiments, patient outcomes are predicted using multiple patient factors or variables, such as CLU and DPT expression levels and tumor volume. In certain embodiments of this aspect of the present invention, expression levels of CLU and DPT, which are predictive index genes, exceeding a certain threshold indicate a favorable prognosis. In related embodiments, index gene expression and HCC tumor volume ≤ 115 cm² 3The additional non-genetic variables indicate a favorable prognosis.

[0039] In certain embodiments, the expression levels of one or more indicator genes and / or tumor volume measurements in a patient sample are incorporated into the algorithm to provide a value that reflects the likelihood of disease recurrence, and in particular the algorithm is a support vector machine algorithm, and more particularly the algorithm is a linear kernel support vector machine algorithm.

[0040] In certain embodiments, indicator gene expression levels and tumor size variables are used in predictive algorithms, particularly machine learning algorithms, and more specifically, linear kernel support vector machine (SVM) learning algorithms, to classify patients into subsets that are likely or unlikely to have a good prognosis, particularly a good prognosis in the following five years after LT.

[0041] The data presented in the examples classified CLU overexpression into a Ct range of 27.36–40.5 in this cohort, and DPT into a range of 33.49–35.69. The change in Ct compared to the control (also called delta or ΔCt) was expressed, and -0.54 was determined to be a useful threshold for CLU overexpression. Similarly, a ΔCt threshold of 7 was identified as a useful threshold for DPT expression in patient samples. Furthermore, the tumor volume was 115 cm³. 3 Smaller tumors have also been found to indicate a better prognosis. The tumor volume value refers to the sum of all tumors contained within the patient's liver, and useful methods that can be used to define tumor volume can be selected from computed tomography and magnetic resonance imaging.

[0042] To assist in classifying patient outcomes, thresholds such as those specified in this embodiment of the invention can be used in particular to binarize patient values. The data shown in Example 5 demonstrate useful thresholds for CLU, DPT, and tumor volume. Samples with expression levels or volume below these thresholds are assigned a score of 0, and samples with measurements above the thresholds are assigned a score of 1. To create a classification system in which the resulting scores indicate patient outcomes after LT, this embodiment incorporates binarized multivariate data from a cohort of patients into an SVM algorithm.

[0043] In certain embodiments, the determination step includes determining the expression of DPT and the expression of a gene selected from CLU, CAPNS1, FBXW7, and SPRY2, particularly the expression of DPT and CLU.

[0044] In certain embodiments, overexpression of CLU or DPT alone indicates a favorable prognosis. In certain embodiments, predictive classifications based on the expression levels of indicator genes, total tumor volume (TTV), or a combination thereof, as provided by the present invention, are combined with additional prognostic factors, such as assessment according to the Milan criteria. The data from the examples demonstrate that CLU expression levels can accurately predict disease outcomes in a subset of HCC patients classified outside the Milan criteria.

[0045] The Milan criteria, introduced by Mazzaferro in 1996 (Mazzaferro et al., N Engl J Med. 1996 Mar 14;334(11):693-9), restrict transplantation of HCC in adults as follows: (1) the diameter of a single tumor is less than 5 cm; (2) there are three or fewer tumor lesions, each not exceeding 3 cm; (3) there is no vascular invasion; and (4) there is no extrahepatic involvement.

[0046] Furthermore, the present invention relates to a system for detecting the high expression of liver transplant biomarkers. This system includes means for determining the expression of DPT and the expression of genes selected from CLU, CAPNS1, FBXW7, and SPRY2, particularly the expression of DPT and CLU.

[0047] In another embodiment, the present invention encompasses the use of primers for amplification and detection of the expression of DPT and CLU, as well as additional biomarkers optionally selected from CLU, CAPNS1, FBXW7, and SPRY2, in a kit for analyzing biomarkers to predict outcomes of liver transplantation for the treatment of HCC.

[0048] Furthermore, the present invention includes a method for treating a patient previously diagnosed with HCC who is classified as having a good prognosis, using a liver transplant procedure, according to any one of the above aspects and embodiments. [Examples]

[0049] Example 1: Study of molecular prognostic biomarkers The inventors analyzed the performance of several current selection criteria based on morphological characteristics and showed that these may exclude good candidates and erroneously include poor candidates. The inventors tested genes previously identified as predictive biomarkers for HCC prognosis after transplantation. The genes DPT and CLU, either alone or in combination, were able to effectively identify a subgroup of patients with a very low probability of relapse after long-term retention of HCC.

[0050] Example 2: Molecular prognostic markers for hepatocellular carcinoma Studies using biomarkers in liver resection (LR) were included in the analysis because important information can be obtained from a cohort of patients who underwent LR. Unlike single genes, overlap between signatures is rare in the literature, and gene signatures are often not reproducible. Therefore, the inventors focused particularly on single-gene biomarkers rather than gene signatures. Furthermore, since many biomarkers have not been further tested and validated outside the context of their original studies, their use in the current study represents at least an effort to externally validate them.

[0051] The inventors searched for the following data: data types (mRNA, miRNA, and protein), prognostic information, specific genes involved, genes associated with good or poor prognosis, types of changes (overexpression, downregulation, high / low methylation, mutations), patient samples, statistical data, and author observations.

[0052] Example 3: Pilot set This first group was run to better trim the biomarkers associated with a better prognosis identified through a literature review, and to further reduce the number of putative biomarkers associated with early relapse. From the initially proposed 20 patients (see Example 6, Sample Collection), nine additional patients were included in the test set due to difficulties in RNA extraction from the first samples belonging to older cases. Samples from three patients were not used due to unsuitability. Thus, the test set consisted of 26 patients (6 patients with MCs beyond relapse; 7 patients with MCs beyond relapse; 7 patients within MCs without relapse; 6 patients within MCs with relapse). After RNA extraction, cDNA was obtained and RT-qPCR was performed. Finally, differences in gene expression evaluated correlated with clinical data. Table 1 shows the results of the test set according to relapse within or beyond MC.

[0053] [Table 1]

[0054] In patients within the Milan Continence (MC), clathelin (CLU, p=0.09), CAPNS1 (p=0.05), and FBXW7 (p=0.04) showed significant differences in expression depending on relapse. In patients outside the MC, clathelin (p=0.02), dermatopontin (p=0.06), and SLC16A4 (p=0.08) also showed significant differences in expression depending on relapse. Claathelin was the best marker to distinguish all relapses (p=0.01). Next, subsets of selected biomarkers that met the inclusion criteria for the experimental study—CAPNS1, DPT, CLU, and FBXW7—were applied to the validation set. Furthermore, the inventors included SPRY2 for further downstream analysis of this validation set because, although not significant in the other two groups, the p-value for the “Milan Continence” group in the experimental set was borderline significant (p=0.13). MUC15, another gene showing a borderline p-value (p=0.14) for statistical significance in the "all criteria" group, was not included in the group of genes tested in the validation set. This exclusion is likely due to its reduced function in RT-qPCR (too high Ct and frequent sample breakage), which is associated with very low expression levels.

[0055] Example 4: Validation set The complete population of 301 patients who were selected for LT for HCC between September 1992 and February 2014 was considered. According to the inclusion and exclusion criteria (see Example 6, Study Population), a total of 275 patients with histologically confirmed HCC were identified in this analysis. Of the 44 patients excluded, 32 had perioperative death and 12 had residual and / or extrahepatic disease. Patients who had previously been included in the experimental set (26), as well as patients less than 5 years post-LT (38), were also excluded. The initial study population consisted of 167 patients. In 33 patients (19.7%), sample analysis was impossible due to extensive tumor necrosis (14 / 167; 8.3%), poor quality RNA extraction (12 / 167; 7.2%), or unavailability of formalin-fixed paraffin-embedded tissue (7 / 167; 4.2%). As a result, 20 additional patients less than 5 years post-LT were added, bringing the validation set to a total of 154 patients. A total of 180 patients had their samples analyzed for molecular markers. Table 2 shows the general characteristics of the experimental set (n=26), the validation set (n=154), and the overall population (n=180).

[0056] As expected, the experimental set exhibited more invasive tumor characteristics compared to the validation set, as patients within and outside the microclavus (MC) were evenly distributed in this first set. Patients in the experimental set more frequently had vascular invasion, increased tumor size, increased TTV, and a lower proportion of patients within the MC. In the overall population, the median donor age was 38 years, consistent with the high proportion of FAP donors (patients with familial amyloid polyneuropathy are treated with living donor transplants, but the patients' livers become available for the treatment of other patients). VHC was prevalent, with 46% of patients being infected. A key observation was the short waiting time to LT, with a median of 1 month and a mean waiting time of 2.2 months (0–18). The median tumor size was 1, with a mean of 2.17 (1–11), and the median maximum tumor size was 2.75 cm (mean 3.25 cm, variable from 0.3–20 cm). Only 120 patients (66.7%) had MC (microclavicular function), and 154 patients (85.6%) had TTV < 115 cm.3 It was inside.

[0057] [Table 2]

[0058] Expression levels of CAPNS1, DPT, CLU, FBXW7, SLC16A4, and SPRY2 were analyzed in a validation set. In many samples, SLC16A4 expression was too low to be determined, and further analysis of this specific biomarker was abandoned. Figure 1 shows the expression differences of selected genes associated with relapse, using the Wilcoxon (Mann-Whitney) test.

[0059] Calpain small subunit 1 (CAPNS1) CAPNS1 was included in the validation set because it showed a slight difference in expression (p=0.05) for recurrence in patients within Milan criteria in the experimental set. As a result, in the validation set, CAPNS1 expression levels in tumor tissue differed significantly with recurrence (OR 1.448, CI 1.140~1.840, p=0.002). However, this was not reflected in the corresponding significant difference in the hazard ratio for either DFS (HR 1.073, CI 0.978~1.178, p=0.136) or OS (HR 1.014, CI 0.913~1.125, p=0.796). ROC curves were used to determine the cutoff value for recurrence. With an AUC of 0.63, the optimal cutoff value was considered to be a delta-CT value of 0.20, corresponding to a sensitivity of 65% and a specificity of 45%. However, when this variable was classified, no effect on DFS (p=0.378) or OS (p=0.655) was observed. While this marker is associated with relapse, its distribution shows considerable overlap between patients with relapse and those without.

[0060] Dermatopontin (DPT) DPT was included in the validation set because its difference in recurrence was nearly statistically significant in patients exceeding the Milan criteria (p=0.06). DPT showed a strong association with recurrence in the validation set (OR 1.283, CI 1.103~1.494, p=0.001). The association with DFS (HR 1.048, CI 0.977~1.123, p=0.192) or OS (HR 1.036, CI 0.961~1.116, p=0.357) was not initially clear. ROC curve analysis revealed an AUC of 0.77 (Figure 2), allowing for a delta CT cutoff value of 7, which corresponds to a sensitivity of 84% and a specificity of 63%. Using this value, an HR of 0.480 (CI 0.295~0.782, p=0.003) was observed for DFS. The hazard ratio (HR) for OR was 0.503 (CI 0.305~0.831, p=0.007). Strong DPT expression reduced the risk of relapse by approximately nine times (OR 0.116, CI 0.37~0.363, p<0.001).

[0061] The above findings indicate that DPT overexpression is associated with a favorable prognosis after LT. In patients with strong DPT expression, the risk of relapse was reduced fivefold, disease-free survival increased by 50%, and overall survival increased by 40%. DPT expression was able to differentiate patients with a favorable prognosis after LT from those with a poor prognosis. A possible synergistic effect with CLU expression was found to be associated with microvascular invasion. This association is thought to be related to the function of extracellular matrix assembly and cell adhesion, ultimately influencing microvascular invasion. This is the first report to link DPT expression with the prognosis after LT for HCC, and the first to demonstrate that this gene is an independent predictor of a favorable prognosis.

[0062] Sprouty RTK signaling antagonist 2 (SPRY2) This protein is a key regulator of essential pathways involved in cancer development, including angiogenesis, cell proliferation, invasion, migration, and cytokinesis. Song et al. (Hepatobiliary & Pancreatic Diseases International 2012;11(2):177-184) investigated SPRY2 expression in tissue microarrays using samples from 240 randomly selected HCC patients who underwent hepatectomy. 207 patients (86.3%) showed downregulation of SPRY2 expression, which correlated with decreased survival (p=0.002) and increased recurrence (p=0.003), acting as an independent predictor of postoperative recurrence in HCC patients (HR=1.47; 95% CI, 1.02~2.08; p=0.037). Our study is the first to correlate SPRY2 overexpression with increased recurrence in a population of patients who underwent LT for HCC. However, the inventors were unable to determine a cutoff value that could effectively distinguish between relapsed and non-relapsed patients due to significant overlap between the two groups. Therefore, the clinical relevance of SPRY2 still needs to be tested in further studies.

[0063] SPRY2 showed some difference in expression for relapse within the Milan criteria in the experimental set (p=0.13). In the validation set, SPRY2 was significantly associated with relapse, with an OR of 1.342 (CI 1.106~1.628, p=0.003). However, no association was found with DFS (HR 1.050, CI 0.962~1.145, p=0.273) or OS (HR 1.030, CI 0.943~1.125, p=0.514). Since the AUC was 0.68, the optimal cutoff value was determined to be a delta-CT value of 7.6, corresponding to a sensitivity of 69% and a specificity of 65%. However, no correlation was found for DFS (HR 1.420, CI 0.897~2.248, p=0.134) or OS (HR 1.305, CI 0.814~2.094, p=0.269).

[0064] Clasterin In the experimental set, CLU expression revealed a difference in relapse rates among patients exceeding the Milan criteria (P=0.02), and this was also true for samples from the entire set (P=0.01). In the validation set, the Wilcoxon test indicated that CLU expression was not associated with relapse (Figure 36). However, logistic regression analysis revealed that CLU expression was correlated with relapse in the validation set (OR 1.219, CI 1.059~1.403, p=0.006). While the Wilcoxon test is a very robust statistical test, it is not sensitive to subtle differences between groups, and in this case, this was reflected as a subtle difference in expression between the relapsed and non-relapsed groups. No association was initially observed with DFS (HR 1.073, CI 0.978~1.178, p=0.136) or OS (HR 1.014, CI 0.913~1.125, p=0.796). Using the ROC curve, a cutoff value of -0.54 for ΔCT was determined, corresponding to an AUC of 0.59 (sensitivity 77%, specificity 35%). At this cutoff, the HR for DFS was 1.568 (CI 0.951~2.584, p=0.078), and the HR for OS was 1.550 (CI 0.926~2.595, p=0.096). Although this association appears weak, it was considered sufficient to warrant further investigation with multiple logistic regression analysis and Cox regression analysis.

[0065] F-box and WD repeat-containing protein 7 (FBXW7) FBXW7 levels appeared to differ in response to relapse in patients within the Milan criteria in the experimental set (p=0.04). However, this difference was not observed in the validation set. Logistic regression revealed an OR of 1.238 for relapse (CI 0.894–1.052, p=0.545). Cox regression analysis yielded an HR of 1.023 for DFS (CI 0.929–1.128, p=0.640) and an OS of 1.000 (CI 0.900–1.113, p=0.994).

[0066] Example 5: Analysis of the total population The inventors analyzed the entire population of 180 patients. Considering the results obtained in the validation set, the expression levels of CLU and DPT, measured by ΔCT values, were included in this analysis for the expression of housekeeping genes in tumor tissue.

[0067] Factors associated with recurrence in this population were tested using multiple logistic regression (Table 3). The inventors observed that tumor number, microvascular invasion, and poor differentiation were independently associated with recurrence, as were histological total tumor volume (TTV). Using DPT expression reduced the predictive risk of recurrence by more than 5-fold, while CLU expression predicted a reduction of up to 61% in recurrence risk. However, when both were introduced into the model, only DPT expression still independently predicted recurrence (OR 0.178, CI 0.063~0.507, P=0.001), while CLU (OR 0.729, CI 0.063~0.507, P=0.554) did not. No statistical interaction was detected between the variables “DPT expression” and “CLU expression” when recurrence was used as the outcome variable (p=0.402). Expression of both CLU (p=0.94) and DPT (p=0.960) was not associated with poorly differentiated tumors.

[0068] [Table 3]

[0069] High DPT expression was associated with reduced microvascular invasion (OR 0.370, CI 0.140~0.979, p=0.045), whereas CLU expression was not (p=0.173). However, when microvascular invasion was used as the outcome variable, an interaction was detected between DPT and CLU (p=0.033), suggesting that the development of microvascular invasion in tumors with low or high DPT expression is strongly correlated with CLU expression. In this study, 21.2% (14 / 66) of patients with low DPT expression had microvascular invasion. In this subgroup, the incidence of microvascular invasion was 4.8% (1 / 21) when CLU was strongly expressed, and simultaneously, when CLU was low, microvascular invasion was observed in 28.9% (13 / 45) of patients (p=0.027). This interaction has not been detected in the scientific literature and should be explained in future research, suggesting a relationship between these genes.

[0070] [Table 4]

[0071] Table 4 shows the results of multivariate Cox regression analysis of DFS in the total population. Microvascular infiltration was the only factor correlated with a decrease in DFS. On the other hand, high expression levels of DPT and CLU acted as positive predictors and correlated with increased survival. When both CLU and DPT were included in the model simultaneously, CLU ceased to be an independent variable (HR 0.600, CI 0.351~1.026, p=0.062), but no statistical interaction was detected between CLU and DPT (p=0.822).

[0072] Dermatopontin (DPT) The inventors observed disease-free survival (DFS) based on the strength or weakness of DPT expression (Figure 3). Patients exhibiting strong DPT expression, defined as a ΔCT level greater than 7, had DFS rates of 70% and 52.2% at 5 and 10 years, respectively. Strong DPT expression also allowed for the identification of a subgroup of patients with a better prognosis, even among those exceeding MC (Figure 4). This subgroup was characterized by TTV > 115 cm. 3 The study was also applied to patients exhibiting other poor prognostic criteria, such as the presence of microvascular infiltration or poor differentiation (Figures 5A, B, and C). Finally, in the subgroup of HCV patients showing strong DPT expression, the 5-year and 10-year DFS were 79.2% and 58.8%, respectively (Figure 6).

[0073] Clasterin (CLU) CLU expression was analyzed in 198 HCC specimens using tissue microarrays (Wang et al., Oncotarget 2015;6(5);2903-16). The CLU protein was mainly detected in the cytoplasm of tumor cells. Multivariate Cox regression analysis showed that CLU overexpression was an independent prognostic factor for tumor recurrence after resection (HR 1.628). In the same study, it was found that CLU overexpression significantly promoted HCC cell invasion in vitro and distant lung metastasis in vivo, while silencing CLU reduced the invasiveness of HCC cells both in vitro and in vivo. CLU overexpression may enhance metastatic potential in prostate cancer, renal cell carcinoma, gallbladder cancer, and breast cancer.

[0074] In this cohort, CLU overexpression was associated with relapse and survival, and patients with a good prognosis after LT were identified from the poor prognosis group. CLU also had a significant impact on disease-free survival (DFS) (Figure 7). Patients with high CLU expression had a 68.9% 5-year survival rate and a 56.1% 10-year survival rate. Although not as pronounced as DPT, CLU expression alone was associated with patients exceeding the Milan criteria (A) or having a TTV of 115 cm². 3When used within groups with poor prognosis, such as patients exceeding (B), it allows for the selection of patients with better outcomes. Patients with microvascular infiltration (C) or poor differentiation (D) showed a clear trend, but there was no statistically significant association with survival due to CLU expression (Figure 8). In the subgroup of HCV patients with high CLU expression, the 5-year and 10-year DFS were 74% and 67%, respectively (Figure 9).

[0075] [Table 5]

[0076] Multivariate Cox regression analysis of overall survival (OS) was also performed. Ethanol intake and microvascular infiltration were independently associated with decreased OS. While microvascular infiltration is a known risk factor, the decrease in OS due to ethanol intake was not accompanied by a corresponding decrease in disease-free survival (DFS) and may be related to comorbidities commonly associated with ethanol intoxication, such as ethanol cardiomyopathy. Strong expression of both DPT and CLU was independently associated with increased OS (Table 5).

[0077] A combination of multivariate analyses with predictive power To enhance the prognostic power of genes associated with HCC outcomes, we developed a multi-gene model combining the predictive power of all genes and investigated the synergistic effect between DPT and CLU. A simple score based on the combined expression of these two genes was used. Since their DFS hazard ratios were similar, each gene was assigned the same points: 0 points for weak expression of both genes, 1 point for strong expression of one gene, and 2 points for strong expression of both genes. Strong expression of both genes was associated with 78% and 60% survival at 5 and 10 years, respectively, and the gene combination was superior to DPT alone or CLU alone (Figure 10). This score was calculated for patients excluded by the Milan criteria (Figure 11A) with a TTV of 115 cm. 3Identify favorable prognoses after LT for patients with tumor thrombus extension beyond the main portal vein (Figure 11B) and for patients with poorly differentiated tumors (Figure 11C). Again, although statistical significance was not reached, a trend towards increased survival was observed in patients with microvascular invasion (Figure 11D). The results are superior to those obtained individually for each gene. Figure 12 shows predicted survival using this simple composite gene score in combination with a TTV of 115 cm 3 or less. This combination identifies groups of patients with very favorable prognoses after LT, groups with moderately acceptable but still acceptable prognoses after LT, and groups with poor prognoses after LT.

[0078] According to the previously shown results for each individual gene, DPT may have superior predictive power when compared to CLU. As a result, the combination of these genes was evaluated without using a score in order to better assay the prognostic power of patients expressing only one of these two genes strongly (Figure 13).

[0079] In a second approach incorporating multiple genes into the HCC prediction strategy, the expression level values of two genes, CLU and DPT, were used together with TTV in a linear kernel support vector machine classification algorithm. A jackknife strategy was used to evaluate the performance of the algorithm. That is, the algorithm was trained using all but one data point, this one was used for evaluation, and then this process was repeated for all data points and the results were collected. After removing the test samples, the remaining dataset was subjected to data augmentation (SMOTE) and normalization (unit variance and zero mean) procedures. This was done to obtain a balanced training dataset, to prevent overfitting to the most dominant class, and to ensure that all features are equal in the classification feature space.

[0080] A cohort of 154 patients was evaluated using a multivariate algorithm with three variables. The variables were binarized using a thresholding method, where patients below the threshold were assigned a value of 0, and patients above the threshold were assigned a value of 1. The threshold units were the same as those for each feature (ΔCt for gene expression, and volume for tumor volume). The threshold values ​​used were -0.54 ΔCt for CLU, 7 ΔCt for DPT, and 115 cm for TTV. 3 The following results, obtained after the evaluation procedure, demonstrate that an algorithm incorporating binary data from multiple indicator genes and tumor volume accurately classifies patients who are likely to show disease recurrence after LT for treatment of HCC (accuracy 67%, false positives 22%, precision 91%, recall 64%, Figures 14 and 15). The accuracy of the classification was also good when performed on a subset of patients classified either within the Milan criteria used in current clinical practice (accuracy 69%, false positives 36%, precision 92%, recall 71%, Figure 16) or outside the Milan criteria (accuracy 62%, false positives 11%, precision 87%, recall 45%, Figure 17), suggesting that patient classification based on binary scores based on thresholds of indicator gene expression (and tumor volume) can identify a subset of patients who are likely to have good outcomes after LT and, otherwise, can be excluded from this treatment.

[0081] For example, Patient 1 has the following values: DPT=10.48, CLU=5.34, and TTV=18.0, while Patient 2 has DPT=4.39, CLU=0.84, and TTV=5.0. Next, these values ​​are binarized using their respective thresholds to obtain binary values ​​for each variable. These variables are then subjected to a linear kernel SVM algorithm to generate binary values ​​indicating the prediction of recurrence. This model can generate estimated probabilities, and in this example, to simplify the output, the results themselves are binarized to separate values ​​above and below a 50% probability. The interpretation of the results is straightforward, and the output is simply a binary value indicating whether recurrence is predicted (R,0) or not (NR,1) (Table 7).

[0082] [Table 6]

[0083] Finally, the issue of heterogeneity was also considered. Samples from seven patients, taken from different locations within the same tumor, were available. Variations in expression levels were measured for housekeeping genes for CLU and DPT, and whether different values ​​triggered group migration ("group migration") based on previously calculated cutoff values. In these samples, there was little variation in DPT expression levels, and no patients needed to be moved to a different group based on these values. On the other hand, CLU expression appeared to be associated with tumor heterogeneity. These results suggest that intratumoral gene expression of DPT is likely to be homogeneous.

[0084] CAPNS1 In a study involving 192 patients undergoing LT for HCC, CAPNS1 overexpression was significantly associated with tumor number and size, tumor encapsulation, venous invasion, and pTNM stage. Multivariate analysis revealed that CAPN4 expression was a strong independent prognostic factor for survival in HCC patients (HR 4.068, CI 2.524-6.555; p<0.001). We demonstrated an association between CAPNS1 expression and relapse (p<0.001). However, no correlation with OS or DFS was demonstrated, and the ΔCt cutoff value was not proven useful in distinguishing between patients who relapse and those who do not. CAPNS1 is a promising marker for combination with other discriminatory markers.

[0085] Example 6: Materials and Methods Molecular prognostic biomarkers After identifying clinical factors related to prognosis, the inventors aimed to identify molecular biomarkers that may have greater discriminative power regarding the prognosis of HCC. To this end, a collaborative study was established between the CHBPT at Curry Cabral Hospital and Ophiomics-Precision Medicine, with Professor Jose Pereira Leal as principal investigator and Joana Cardoso Vaz as co-investigator. The molecular biomarker study included data mining and bioinformatics analysis of public repositories of HCC patients and published literature on HCC biomarkers.

[0086] Data mining and bioinformatics analysis We performed molecular data mining and bioinformatics analysis on public repositories of HCC patients who underwent liver resection and transplantation. Available public data on HCC patients were re-analyzed using our proprietary analysis pipeline, which includes miRNA and mRNA expression profiling data.

[0087] Molecular markers for prognosis in hepatocellular carcinoma: A systematic review The inventors conducted a literature search on previously published biomarkers in HCC patients who underwent LT (Long-Term Transplantation). Because data in LT patients is limited, and important information can be obtained from cohorts of patients who underwent hepatectomy (Long-Term Transplantation), studies using biomarkers in LR (Long-Term Transplantation) were also included in this analysis. A systematic literature review was conducted on the available evidence regarding the role of molecular biomarkers in the prognosis of patients undergoing resection or transplantation for HCC. This review followed the general guidelines of the National Academy of Medicine's systematic review criteria. The purpose of this review was to evaluate the role of molecular biomarkers in the prognosis of patients undergoing LR or LT for HCC.

[0088] Population: The study was limited to English-language academic papers and was conducted from January 2008 to October 2016. Studies published only in abstract form, unpublished studies, and articles published in non-peer-reviewed journals were excluded. Animal studies and in vitro studies were also excluded. Studies on molecular markers not related to the prognosis of HCC were also excluded. Therapeutic intervention: Hepatectomy and transplantation for HCC. The reason for choosing resection was the lack of research on LT. Outcomes: The primary outcomes investigated were disease-free survival, recurrence, and overall survival. Study design: Because randomized controlled trials (RCTs) and controlled trials are rare in this setting, cohort studies were deemed suitable for inclusion, even if retrospective in nature, due to the lack of data. Case series were acceptable in the review, but case reports were excluded. Economic evaluations were also excluded. Searches of the PubMed, ClinicalKey, and Cochrane databases were performed using the following keywords in various combinations: hepatocellular carcinoma, surgery, resection, transplantation, prognosis, molecular, and biomarkers. The cited references of identified papers were used to find further relevant publications. The data obtained included data type (mRNA, miRNA, protein), prognostic information, specific relevant genes, genes associated with good or poor prognosis, types of changes (overexpression, downregulation, hypermethylation, and mutation), patient samples, statistical data, and author observations. Biomarkers were selected according to their predictive power, number of citations across various centers, ability to replicate techniques, and available reagents. After selecting top biomarker candidates, pilot studies were conducted.

[0089] research group The inclusion and exclusion criteria were the same as those previously used in studies on clinical biomarkers. Inclusion criteria: Patients who underwent liver transplantation for hepatocellular carcinoma. Exclusion criteria: Age under 18 years; absence of cirrhosis; histological type of fibrous laminae or hepatobiliary carcinoma; absence of histological confirmation of HCC; and further, since the inventors' primary outcome measure was recurrence / disease-free survival, they excluded cases with perioperative death, extrahepatic invasion, and residual disease. The complete population of 301 patients who underwent liver transplantation for HCC between September 1992 and February 2014 was considered. From the 231 patients obtained after applying the exclusion criteria, the inventors included only those who underwent follow-up for more than 5 years. The final set of samples used in this study (experimental set and validation set) included 180 patients. Experimental set: The experimental set included patients with relapse exceeding the Milan criteria (n=6) and patients without relapse (n=7), as well as patients with early relapse within the Milan criteria (n=6) and patients without relapse (n=7). Subsets of biomarkers that met the inclusion criteria in the experimental study underwent two rounds of analysis using the validation set. Validation set: From the initial population of 275 patients, patients previously included in the experimental set, those with perioperative mortality, extrahepatic invasion, residual disease, and those with less than 5 years of follow-up were excluded, resulting in 154 patients initially included in the validation set.

[0090] Sample collection For 180 patients who underwent liver transplants at the Hepatobiliary and Pancreatic Transplant Center of Cale Cabral Hospital, tumor specimens were fixed in formalin and preserved in paraffin blocks. For all selected blocks, histopathological characterization and regional selection were performed on hematoxylin-eosin (HE) stained sections under the supervision of an experienced pathologist. Both the Institutional Review Board of Health and Medical Ethics at Cale Cabral Hospital and the Institutional Review Board of Health and Medical College NOVA approved this study.

[0091] RNA extraction and cDNA synthesis Preserved formalin-fixed paraffin-embedded (FFPE) tissue sections (5 μm) were deparaffinized and counterstained with Mayer's hematoxylin-eosin. All samples were macro-dissected under the guidance of a pathologist. Total RNA was extracted using the RNeasy FFPE kit (Qiagen) with some modifications, following the manufacturer's instructions: Proteinase K cell lysis at 56°C was performed overnight. This included the "on-column" DNA digestion procedure using the RNase-free DNase set (Qiagen). Each extracted RNA sample was reverse transcribed using the SuperScript® VILO® cDNA synthesis kit (Thermo Fisher Scientific).

[0092] [Table 7]

[0093] Quantitative real-time PCR Due to known FFPE degradation issues and the small sample size, standard methods could not assess RNA concentration and integrity. Therefore, the inventors obtained the quantity and quality of isolated RNA samples using a highly sensitive RNA ScreenTape (Agilent) on an Agilent 2200 TapeStation system (Agilent). Primer sets were designed to operate at 60°C using the NCBI Primer BLAST tool (Ye et al., BMC Bioinformatics 2012;13:134) with amplification product lengths of 70–100 bp (Table 6), and were purchased from Invitrogen (Thermo Fisher Scientific). Replication of each sample was analyzed by RT-qPCR using SsoFast® EvaGreen® Supermix reagent (BioRad, Hercules, California, USA) in a 10 μL reaction mixture containing a template (1 μL, 0.5–1 ng / μL) and primers (0.5 μM each). Samples were processed using a CFX96 Touch® real-time PCR detection system (BioRad, Hercules, California, USA) according to the following cycle program: 50 cycles of 120 seconds at 98°C, 5 seconds at 98°C, and 15 seconds at 60°C. Fluorescence data were collected at 60°C. For data and statistical analysis, relative expression difference analysis of target genes by RT-qPCR was performed, calculating the multiplicative change in expression based on the 2ΔΔCt or ΔCt method. Here, the average quantitative cycle of replication was used as the cycle threshold (Ct), compared to the Ct of the calibrator gene ribosomal protein L13a (RPL13A) according to Livak et al., Methods 2001;25(4):402-8. Values ​​greater than 1 indicate upregulation, and values ​​less than 1 indicate downregulation. Differences in target gene expression using RT-qPCR data between sets of disease-recurrent (R) and non-recurrent (non-R) samples were performed using the R language for statistical calculations (Team RDC, Austria, 2009), and statistical significance was calculated using the Wilcoxon rank-sum test (confidence level = 0.95).

[0094] Furthermore, the obtained RT-qPCR data were also correlated individually with patient disease-free survival for each candidate target gene using the same method described above for clinical markers. Continuous variables were expressed as the median or mean and standard deviation (SD) of the interquartile range (IQR) and compared using independent sample t-tests. The demographic variables of interest for transplant patients were compared using Student's t-test, Pearson's chi-squared test, or Fisher's exact test, as needed. The outcome variables were relapse (disease-free survival) and death (overall survival). Time to outcome was calculated using the date from transplantation to the event date, or to the date of the last follow-up period for patients who did not experience an event. Kaplan-Meier survival curves were created for post-transplant outcome analysis. The effects of demographic variables on disease-free survival and overall survival were examined using log-rank tests and Cox regression models. Cutoff values ​​were determined by receiver operating characteristic (ROC) analysis. In multivariate analysis, all variables significant for outcome P<0.20 were included in either a Cox proportional hazards model (Therneau et al., Springer-Verlag, 2000) or a multiple logistic regression model, depending on the type of outcome. Backward selection was performed to preserve significant variables. Individual stratified survival analyses were performed as needed. A P-value less than 0.05 was considered statistically significant. Statistical analysis was performed using SPSS versions 22.0 and 24.0 (SPSS Inc., Chicago, Illinois). The linear kernel SVM algorithm was developed in a Python 3.6.7 environment using the following packages for data manipulation and classification: scikit-learn (0.23.2); numpy (1.19.1); pandas (0.23.4).

Claims

1. 1. A method for predicting the outcome of liver transplantation for the treatment of hepatocellular carcinoma (HCC), comprising: - in a determining step, in a liver sample obtained from a patient suffering from HCC, a. Clusterin (CLU, gene ID GC08M027596), and b. Dermatopontin (DPT, gene ID GC01M168664) determining the expression level of each of a group of indicator genes comprising: - in a classification step, assigning patients to a favorable prognosis for liver transplant outcome based on the expression levels of said indicator genes; The method comprising:

2. The group of indicator genes is a. Calpain small subunit 1 (CAPNS1, gene ID GC19P036434) b. F-box and WD repeat-containing protein 7 (FBXW7, gene ID GC04M152321) c. Sprouty RTK signaling antagonist 2 (SPRY2, gene ID GC13M080335) The method for predicting the outcome of liver transplantation for the treatment of HCC according to claim 1, further comprising at least one of:

3. The method for predicting the outcome of liver transplantation for the treatment of HCC according to claim 1 or 2, wherein the indicator gene is overexpressed relative to a threshold value.

4. A method for predicting the outcome of liver transplantation for the treatment of HCC according to any one of claims 1 to 3, wherein overexpression of the indicator genes, in particular overexpression of DPT and / or CLU, indicates a good prognosis.

5. a. Overexpression of DPT and CLU, and b. Total tumor volume 115 cm ３ Is less than or equal to indicates a good prognosis, A method for predicting the outcome of liver transplantation for the treatment of HCC according to any one of claims 1 to 4.

6. 6. The method for predicting the outcome of liver transplantation for the treatment of HCC according to any one of claims 1 to 5, wherein the expression level of the indicator gene is determined by polymerase chain reaction, and the expression value of the indicator gene relative to the threshold is determined as the difference between the threshold cycle number of the indicator gene and the threshold cycle number of an internal control gene, wherein the threshold cycle number is the PCR cycle number at which products (the indicator gene and the internal control gene) are detected.

7. The difference in threshold cycle number of the indicator gene is a. For DPT, DPT is greater than 7, and b. If CLU is greater than -0.54, The method for predicting the outcome of liver transplantation for the treatment of HCC according to any one of claims 1 to 6, wherein the indicator gene can be said to be overexpressed.

8. A method for predicting the outcome of liver transplantation for the treatment of HCC as described in claim 1 or 7, wherein the expression levels of one or more of the indicator genes in the patient sample and / or measurements of tumor volume are incorporated into an algorithm to provide a value reflecting the likelihood of disease recurrence, particularly wherein the algorithm is a support vector machine algorithm, more particularly wherein the algorithm is a linear kernel support vector machine algorithm.

9. A system for detecting high expression of liver transplant biomarkers, comprising means for determining the expression of DPT and the expression of genes selected from CLU, CAPNS1, FBXW7 and SPRY2, in particular the expression of DPT and CLU.

10. Use of primers for amplifying and detecting the expression of DPT and CLU, and optionally the expression of an additional biomarker selected from CLU, CAPNS1, FBXW7 and SPRY2, in a kit for analyzing biomarkers for predicting the outcome of liver transplantation for the treatment of HCC.

11. 10. A method of treating a patient previously diagnosed with HCC with liver transplantation treatment, wherein the patient has been classified as likely to have a good prognosis according to a method as defined in any one of claims 1 to 9.

Images