Method and system for establishing a prediction model of liver toxicity of a chemical

Abstract

The application relates to a method and system for establishing a prediction model of chemical liver toxicity, which comprises the following steps: (1) differentiating human embryonic stem cells or human induced pluripotent stem cells into hepatocyte-like cells and constructing a three-dimensional hepatocyte model; (2) taking miR-122, LDH and Cyto C in the three-dimensional hepatocyte as combined test indexes to judge liver toxicity of a liver toxicity mode compound; (3) classifying the liver toxicity mode compound into three categories through two canonical discriminant functions by a linear discriminant analysis modeling method; and (4) converting the canonical discriminant function into a Fishers discriminant function to obtain a liver toxicity prediction model. The application has the general characteristics of a conventional cell model for evaluating compounds, and can also evaluate the toxicity effect and health risk of the compounds in a high-throughput manner with relatively less manpower.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of compound risk assessment and management, and particularly relates to a method and system for establishing a prediction model of chemical liver toxicity. BACKGROUND

[0002] The liver is an important metabolic organ of compounds in the human body, and the toxicity of compounds can change after being metabolized by the liver. Therefore, the liver is one of the most important target organs of compound toxicity. At present, the number of compounds contacted in our daily life has exceeded 30,000, including drugs, food additives, health products, etc. Among them, more than 1,000 compounds can definitely induce liver damage. Therefore, liver damage induced by exogenous compounds is undoubtedly an important indicator in the risk assessment and management of compounds. However, with the rapid development of industrialization, the number of new compounds increases dramatically every year, while the traditional toxicity test of compounds is restricted by objective factors such as long time consumption, high cost, and human time and manpower consumption. The efficiency and rate of the test are obviously lagging behind the growth of the number of compounds, thus resulting in a huge gap in toxicity information and highlighting the importance of rapid and accurate evaluation of compound toxicity in ensuring human health. Therefore, it is urgent to develop a precise, rapid, and high-throughput toxicity test system and method.

[0003] The liver is one of the organs with significant species differences. There are obvious differences in metabolic characteristics between species, so the liver toxicity research using experimental animals cannot accurately reflect the response mode of human genotype, nor can it accurately represent the disease or harmful effect phenotype of liver damage caused by compounds. In addition, the current in vitro liver toxicity experiment mostly uses two-dimensional cultured liver cells as experimental materials. However, many studies have reported that two-dimensional cultured liver cells are difficult to truly reflect the liver toxicity induced by chemicals acting on the body due to the lack of reasonable intercellular or extracellular matrix connection.

[0004] Therefore, it is of great significance to develop an in vitro liver model that can accurately reflect human biological characteristics for compound toxicity testing and prediction by referring to the idea of alternative methods. SUMMARY

[0005] The present application solves the technical problem of providing a method and system for establishing a prediction model of chemical liver toxicity. For the first time, a human in vitro three-dimensional cell model is used to characterize the information of whether a chemical can induce liver toxicity through mathematical modeling, and to prompt the health risk of the chemical to humans.

[0006] The present application provides a method for establishing a prediction model of chemical liver toxicity, comprising the following steps:

[0007] (1) Directly differentiate human embryonic stem cells or human induced pluripotent stem cells into hepatocyte-like cells and construct a three-dimensional hepatocyte model;

[0008] (2) Take miR-122, LDH and Cyto C in the three-dimensional hepatocyte as the combined test index to judge the hepatotoxicity of the hepatotoxicity pattern compound;

[0009] (3) Through the linear discriminant analysis modeling method, the hepatotoxicity pattern compound is classified into three categories through two canonical discriminant functions:

[0010] LD1 = 1.183 x log 10(LDH 测试组 / LDH 对照组 ) + 4.119 x log 10(Cyto C 测试组 / Cyto C 对照组 ) + 1.889 x log 10(miR-122 测试组 / miR-122 对照组 );

[0011] LD2 = 1.429 x log 10(LDH 测试组 / LDH 对照组 ) + 0.753 x log 10(Cyto C 测试组 / Cyto C 对照组 ) - 1.157 x log 10(miR-122 测试组 / miR-122 对照组 );

[0012] (4) Convert the canonical discriminant function above into Fishers discriminant function to obtain a hepatotoxicity prediction model; wherein the Fishers discriminant function is:

[0013] I: -1.727 + 2.440 x log 10(miR-122 测试组 / miR-122 对照组 ) + 3.763 x log 10(Cyto C 测试组 / Cyto C 对照组 ) - 0.208 x log 10(LDH 测试组 / LDH 对照组 );

[0014] II: -13.052 + 9.818 x log 10(miR-122 测试组 / miR-122 对照组 ) + 19.950 x log 10(Cyto C 测试组 / Cyto C 对照组 ) + 4.477 x log 10(LDH测试组 / LDH 对照组 );

[0015] III:-7.406+6.667xlog 10(miR-122 测试组 / miR-122 对照组 )+15.028xlog 10(CytoC 测试组 / Cyto C 对照组 )+3.777xlog 10(LDH 测试组 / LDH 对照组 )。

[0016] The three-dimensional hepatocyte model in the step (1) comprises a liver organoid, a hepatocyte cluster or a hepatocyte spheroid.

[0017] The miR-122, LDH and Cyto C in the step (2) are selected by a coefficient of variation.

[0018] The three liver toxicity pattern compounds in the step (3) are specifically no liver toxicity, weak liver toxicity and strong liver toxicity.

[0019] The application method of the liver toxicity prediction model in the step (5) is as follows: after the detection data of the to-be-detected compound are substituted into the equations I, II and III, the calculation results of the three equations are compared, and the number of the equation with the maximum value is the probability of the prediction of the type.

[0020] The application further provides a liver toxicity prediction system of chemicals, comprising: a culture module for directionally differentiating human embryonic stem cells or human induced pluripotent stem cells into hepatocyte-like cells and constructing a three-dimensional hepatocyte model; a grouping module for grouping according to miR-122, LDH and Cyto C as combined test indexes into a test group and a control group; a detection module for IC20 values of liver toxicity pattern compounds and data of the combined test indexes after the liver toxicity pattern compounds act on the liver toxicity test model; an analysis modeling module for modeling by a linear discriminant analysis modeling method; and a prediction module for predicting a to-be-detected compound by using the established prediction model.

[0021] The three-dimensional hepatocyte model comprises a liver organoid, a hepatocyte cluster or a hepatocyte spheroid.

[0022] The compound concentration for predicting liver toxicity is an inhibition concentration (IC20) of 20% of cell viability, and when the IC20 is greater than 1000 μM, 1000 μM is used for the detection of the three indexes of miR-122, LDH and Cyto C.

[0023]

[0024]

[0025] Beneficial effects

[0026] (1) The present application first uses a liver organ derived from a human embryonic stem cell line to establish a liver toxicity test model, and liver toxicity information of a compound can be obtained through an in vitro test in a short time;

[0027] (2) The present application first combines cytochrome C, lactate dehydrogenase and miR-122 as liver toxicity test indicators at the same time, and further verifies the effectiveness and reliability of the combination of the three indicators for identifying liver toxicity of chemicals;

[0028] (3) The present application first uses a human in vitro three-dimensional cell model to characterize information about whether a chemical induces liver toxicity through a mathematical modeling method, and to indicate the health risk of the chemical to humans;

[0029] (4) The present application has general characteristics of evaluating a compound by a conventional cell model, such as only requiring a very small amount of test compound, saving time and test cost compared to animal experiments, and being able to evaluate the toxicity effect and health risk of a compound with high throughput under the condition of using relatively less manpower. BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 is a decision boundary for liver toxicity identification of a compound by a linear discriminant function; and each letter in the figure is an abbreviation of a pattern compound. DETAILED DESCRIPTION

[0031] The present application will be further described below in combination with specific embodiments. It should be understood that these embodiments are only used to illustrate the present application and not used to limit the scope of the present application. In addition, it should be understood that after reading the content taught by the present application, those skilled in the art can make various modifications or changes to the present application, and these equivalent forms also fall within the scope defined by the appended claims of the present application.

[0032] The cell material, reagents and experimental instruments used in the present application are described as follows:

[0033] 1. Human embryonic stem cell line (human embryonic stem cell, hES)

[0034] The human embryonic stem cell line H1 used in the present application is obtained from a commercial cell line.

[0035] 2. Culture solution and reagents

[0036] The culture solution used in the present application is DMEM / F12 (Dubecco's Modified Eagle's Medium), KO DMEM (KnockOut Dubecco's Modified Eagle's Medium), the formula of which is well known in the art, not only described in detail in general textbooks and experimental manuals, but also directly available in the form of finished products from companies (such as ThermoFisher Company in the United States, etc.).

[0037] The components as supplements are any components that maintain or promote cell growth, for example, they can include but are not limited to: amino acids, vitamins, proteins, trace elements, sugars, lipids, etc.

[0038] The detection indexes cytochrome C and lactate dehydrogenase can be obtained from commercial channels. miR-122 can be detected by using a real-time quantitative PCR detection kit and primers. An enzyme marker and a real-time quantitative PCR instrument are required.

[0039] 3. Experimental instruments

[0040] In the present application, an inverted microscope and an enzyme marker are used, both of which are common laboratory instruments. The bioreactor is obtained from commercial channels, and has a technical parameter that can freely adjust the rotation speed between 5 to 60 revolutions per minute.

[0041] Example 1

[0042] Culture of liver organoids of different diameters:

[0043] After resuscitation, the normal growth of human embryonic stem cells (hPSC) is induced to differentiate into hepatocyte-like cells according to the method in the literature "Polystyrene microplastics induce hepatotoxicity and disrupt lipid metabolism in the liver organoids" or the method in the application patent 202011528400.1. After 18 days of induction and differentiation culture, the cells are digested into single cells, and then centrifuged at 1000 revolutions per minute for 5 minutes, and the density of hepatocyte-like cells is adjusted to 1×10 6 6 cells / ml in the culture solution and placed in a vertical rotating bioreactor with a rotation speed of 30 revolutions per minute for 48 or 60 hours, to obtain spherical liver organoids with a diameter of 200 to 300 μm.

[0044] Example 2

[0045] Liver toxicity model compounds:

[0046] The 30 compounds listed in this example are liver toxicity model compounds, and the liver toxicity prediction mathematical model is based on the test data of the 30 compounds.

[0047] According to the definition of "drug-induced liver injury", and taking into account the frequency and types of compound exposure in daily life, the compounds included in the model compound library must meet the premise of "liver injury caused by the compound itself or / and its metabolites, or hypersensitivity or reduced tolerance to the compound due to special constitution", covering multiple fields such as drugs, food additives, food and environmental chemical pollutants, nanomaterials, vitamins and health products.

[0048] Table 1. Liver toxicity information of 30 model compounds

[0049]

[0050]

[0051] Example 3

[0052] Selection of liver injury effect test indicators for model compounds:

[0053] Due to the long time required for tissue / cellular level liver injury effect markers, the detection method has many steps, which is not conducive to the realization of high-throughput screening of compound liver toxicity. The new type of molecular marker microRNA has the characteristics of stability, easy detection, and "one-to-many" regulation, so it can correspond to multiple tissue / cellular level liver injury effects through the changes in the expression of a certain microRNA. Therefore, it can be considered as a new type of liver injury effect detection index with strong universality.

[0054] Among the several microRNAs detected by 5 independent experiments in the liver organoids obtained in Example 1, the coefficient of variation of miR-122 was the smallest, which was 3.90%, indicating good stability, so it was selected as a candidate toxicity effect detection index.

[0055] At the same time, the typical cytotoxicity indicator lactate dehydrogenase (LDH) and the liver injury and mitochondrial injury indicator cytochrome C (Cyto C) were also detected from the supernatant. In the liver organoids obtained in Example 1, the coefficient of variation of Cyto C was 6.89% and the coefficient of variation of LDH was 11.24% after 5 independent experiments, both of which were less than 20%, indicating good stability, so these two indicators were also selected as candidate toxicity effect detection indicators.

[0056] Coefficient of variation (CV) is a statistical measure of the variation (dispersion) in a distribution. A low coefficient of variation indicates that the data points are close to the mean; a high coefficient of variation indicates that the data points are spread out over a wider range of values. CV is a measure of the variability of a data set, and is used to evaluate the stability of a test index or method. The smaller the CV, the smaller the variability between test data. Generally, the acceptable range of CV should be ≤20% (reference: Analytical Quality Requirements of CLIA Proficiency Testing Program in the United States). In this study, the CV of the test index and the test method was used to indicate the variability and stability of the test index and the test method.

[0057] In summary, the CV of the three candidate test indices (miR-122, LDH and Cyto C) were all within the acceptable range, and therefore, the three indices were selected as the combined test indices for the establishment of a liver toxicity prediction model.

[0058] Example 4

[0059] IC20 reflects the cytotoxicity (general toxicity) effect of the test compound:

[0060] When performing a cytotoxicity test of a compound in two-dimensional cultured cells, the cytotoxicity induced by the compound is mainly apoptosis and necrosis. These apoptotic and necrotic cells are often removed when the culture medium is changed, and finally, the number of surviving cells in the culture environment at the end of the test is reduced. In a three-dimensional cell model, intercellular communication and tight junction are important ways of interaction between cells, and are the biological basis for the formation of three-dimensional structure. The apoptosis and necrosis of some cells often lead to the disintegration of the three-dimensional cell clusters in the form of spheres, ellipses or irregular spheres, and the apoptotic and necrotic cells are not removed or not completely removed by the operation of changing the culture medium, and further trigger the apoptosis and necrosis of adjacent cells. Therefore, when a test compound with liver toxicity acts on a three-dimensional liver cell model, the 50% inhibition concentration (IC50) of the test compound on cell viability may be lower than that in two-dimensional cultured cells. Therefore, in a three-dimensional cell model, the 20% inhibition concentration (IC20) value of the test compound on cell viability is calculated according to the dose-response curve of the test compound, and the liver toxicity related effects of the test compound at IC20 concentration are further observed. If no dose-response relationship of cell viability is obtained (i.e. little cytotoxicity), the recommended test concentration is 1000 μM according to the solubility of the test compound, the final concentration of the solvent should not exceed 1‰, and the change of pH value of the culture medium, etc. The IC20 values of the 30 model compounds after LOs test are shown in Table 2.

[0061] Table 2. IC20 values of 30 model compounds after LOs test

[0062]

[0063]

[0064] Example 5

[0065] Detection method and data of test index of liver injury effect of model compounds:

[0066] After the IC20 value of the selected model compounds is determined, the values of the candidate test indexes miR-122, LDH and Cyto C of the toxic effect of each compound after acting on the three-dimensional hepatocyte model for 24 hours are detected respectively.

[0067] LDH can be quantitatively detected by using a commercial lactate dehydrogenase (LDH) release level detection kit; Cyto C can be quantitatively detected by using a commercial cytochrome C (Cyto C) ELISA detection kit.

[0068] miR-122 can be quantitatively detected by using a commercial kit to extract RNA in each group, and using a commercial reverse transcription kit and a real-time quantitative PCR detection kit.

[0069] After the measured data is compared with the control group and converted into the relative change value of each index, log transformation is performed, i.e. the value = log (test group / control group). Among them, the three-dimensional hepatocyte group without compound treatment is the control group.

[0070] Table 3. Test data of 3 candidate indexes of 30 model compounds

[0071]

[0072]

[0073] * The values in Table 3 are the values of the log transformation of the relative change values of each index, i.e. the value = log (test group / control group).

[0074] Example 6

[0075] Construction and verification of liver toxicity prediction model:

[0076] Linear discriminant analysis is based on the relative values and ID20 values obtained when each model compound acts on LOs, i.e. the test values of the test compound in 7 indexes; the test compound is divided into three categories (no liver toxicity, weak liver toxicity, and strong liver toxicity) by two canonical discriminant functions through the linear discriminant analysis modeling method.

[0077] The judgment standard for the success of pre-modeling is that after modeling of all model compounds, the prediction accuracy is >80%, and the cross-validation accuracy is >70%.

[0078] The model compounds were randomly divided into two groups in the ratio of 8:2 (considering the 3 degrees of severity of hepatotoxicity based on Table 3) Training set and Test set That is, 24 compounds in the table are divided into a training set for establishing a prediction model; 6 compounds DOX (strong hepatotoxicity), FLT (strong hepatotoxicity), PVS (weak hepatotoxicity), DX (weak hepatotoxicity), SA (no hepatotoxicity), and VB (no hepatotoxicity) are divided into a test set for testing the prediction model established based on the training set.

[0079] A toxicity prediction model capable of distinguishing 3 classes of hepatotoxicity was established using linear discriminant analysis. This model consists of two linear discriminant functions (LD1 and LD2), which can classify 3 classes of toxicity test data features (see attached Figure 1 ).

[0080] After the model is successfully constructed, the discriminant functions are as follows:

[0081] LD1 = 1.183 x log10(LDH 测试组 / LDH 对照组 ) + 4.119 x log10(Cyto C 测试组 / Cyto C 对照组 ) + 1.889 x log10(miR-122 测试组 / miR-122 对照组 )

[0082] LD2 = 1.429 x log10(LDH 测试组 / LDH 对照组 ) + 0.753 x log10(Cyto C 测试组 / Cyto C 对照组 ) - 1.157

[0083] x log10(miR-122 测试组 / miR-122 对照组 )

[0084] Example 7

[0085] Application of the prediction model and prediction evaluation criteria:

[0086] In order to achieve the purpose of application and rapid judgment, the discriminant function established for distinguishing hepatotoxicity is shown as Fisher's discriminant function. The transformation of Model 1 is

[0087] I: -1.727 + 2.440 x log10(miR-122 测试组 / miR-122 对照组)+ 3.763 x log10(CytoC 测试组 / Cyto C 对照组 )- 0.208 x log10(LDH 测试组 / LDH 对照组 )

[0088] II: -13.052 + 9.818 x log10(miR-122 测试组 / miR-122 对照组 )+ 19.950 x log10(CytoC 测试组 / Cyto C 对照组 )+ 4.477 x log10(LDH 测试组 / LDH 对照组 )

[0089] III: -7.406 + 6.667 x log10(miR-122 测试组 / miR-122 对照组 )+ 15.028 x log10(CytoC 测试组 / Cyto C 对照组 )+ 3.777 x log10(LDH 测试组 / LDH 对照组 )

[0090] After the detection data is substituted into the equations I, II and III respectively, the calculation results of the three equations are compared, and the probability of the first equation is the highest if the value of the first equation is the largest.

[0091] According to the calculation results, the liver toxicity is judged according to the following standards:

[0092] If I > II and I > III, the compound is judged to be non-liver toxicity;

[0093] If II > I and II > III, the compound is judged to be weak liver toxicity;

[0094] If III > I and III > II, the compound is judged to be strong liver toxicity.

[0095] Based on the prediction model established by 24 training set compounds, the liver toxicity severity of 20 of the 24 compounds can be correctly identified, with an accuracy rate of 83.33%, which is more than 80%; the accuracy rate of leave-one-out cross-validation is 80%. The linear discriminant function (LD) eigenvalue is 2.400, and the regular correlation is 0.840, indicating that the discriminant function has strong discriminant ability; among the two discriminant functions, LD1 mainly contributes to the distinction of non-liver toxicity, weak liver toxicity and strong liver toxicity compounds, with high eigenvalue and strong distinguishing power; LD2 mainly contributes to the distinction of non-liver toxicity and liver toxicity compounds.

[0096] Using the established model, the liver toxicity of the test compounds in the training set and the test set is predicted, and the prediction results are shown in Table 4, and the evaluation of the prediction ability is shown in Table 5.

[0097] Substituting the test set compounds into the above prediction model, the accuracy rate of predicting the liver toxicity of 6 compounds is 100%, indicating that the established model has good application adaptability.

[0098] Table 4. Prediction results of liver toxicity prediction model

[0099]

[0100]

[0101] * The predicted liver toxicity is different from the actual liver toxicity severity

[0102] Table 5. Evaluation of the prediction ability of the model

[0103]

[0104] (Supplementary note: The Youden index is the correct index, which indicates the total ability of the prediction to distinguish between compounds with liver toxicity and compounds without liver toxicity. The closer the value is to 1, the better the prediction distinction effect)

[0105] Based on the above data, the prediction total accuracy, sensitivity and specificity are high, the prediction result is good in authenticity, and has good distinguishing power.

[0106] Example 8

[0107] Application demonstration based on the development of the kit

[0108] In actual kit assembly, toxicity prediction application, without repeating the model establishment prediction of Example 6, without repeating the model transformation in Example 7. Only need to detect the IC20 of the compound (Example 4) and the values of the three indicators Cyto C, LDH and miR-122 under the action of IC20 (Example 5), substitute into the equations I, II, III in Example 7 and according to the judgment criteria, the liver toxicity of the compound can be quickly and accurately predicted. The necessary options of the chemical liver toxicity prediction kit assembly include three-dimensional hepatocyte model (such as liver organoids, hepatocyte spheres or hepatocyte clusters) and judgment criteria (such as Fisher discriminant functions I, II, III or linear discriminant functions LD1, LD2), and optional options include matching detection reagents for the three indicators Cyto C, LDH and miR-122, such as ELISA method matching reagents for detecting Cyto C, biochemical method matching reagents for detecting LDH, or real-time quantitative PCR method matching reagents for detecting miR-122.

Claims

1. A method for establishing a prediction model of chemical liver toxicity, comprising the following steps: (1) differentiating human embryonic stem cells or human induced pluripotent stem cells into hepatocyte-like cells and constructing a three-dimensional hepatocyte model; (2) taking miR-122, LDH and Cyto C in the three-dimensional hepatocyte as a combination of test indicators to judge the liver toxicity of a liver toxicity model compound; (3) classifying the liver toxicity model compound into three categories by two canonical discriminant functions through a linear discriminant analysis modeling method; LD1= 1.183 x log 10 (LDH 测试组 / LDH 对照组 ) + 4.119 x log 10 (Cyto C 测试组 / Cyto C 对照组 ) + 1.889 x log 10 (miR-122 测试组 / miR-122 对照组 ); LD2= 1.429 x log 10 (LDH 测试组 / LDH 对照组 ) + 0.753 x log 10 (Cyto C 测试组 / Cyto C 对照组 ) - 1.157 x log 10 (miR-122 测试组 / miR-122 对照组 ); wherein LDH 测试组 and LDH 对照组 refers to the numerical value of LDH in the test group and the control group, Cyto C 测试组 and Cyto C 对照组 refers to the numerical value of Cyto C in the test group and the control group, miR-122 测试组 and miR-122 对照组 refers to the numerical value of miR-122 in the test group and the control group; (4) converting the canonical discriminant functions into Fishers discriminant functions to obtain a liver toxicity prediction model; wherein the Fishers discriminant functions are: I: -1.727 + 2.440 x log 10 (miR-122 测试组 / miR-122 对照组 ) + 3.763 x log 10 (CytoC 测试组 / Cyto C 对照组 ) - 0.208 x log 10 (LDH 测试组 / LDH 对照组 ); II: -13.052 + 9.818 x log 10 (miR-122 测试组 / miR-122 对照组 ) + 19.950 x log 10 (CytoC 测试组 / Cyto C 对照组 ) + 4.477 x log 10 (LDH 测试组 / LDH 对照组 ); III: -7.406 + 6.667 x log 10 (miR-122 测试组 / miR-122 对照组 ) + 15.028 x log 10 (Cyto C 测试组 / Cyto C 对照组 ) + 3.777 x log 10 (LDH 测试组 / LDH 对照组 ); according to the calculation results, the liver toxicity is judged according to the following standards: if I > II and I > III, the compound is judged to be non-liver toxicity; if II > I and II > III, the compound is judged to be weak liver toxicity; if III > I and III > II, the compound is judged to be strong liver toxicity.

2. The establishment method according to claim 1, characterized in that: The three-dimensional hepatocyte model in the step (1) comprises a liver organoid, a hepatocyte cluster or a hepatocyte spheroid.

3. The establishment method of claim 1, wherein: The miR-122, LDH and Cyto C in the step (2) are selected by a coefficient of variation.

4. A system for predicting liver toxicity of a chemical, characterized by, It comprises: a culture module for differentiating human embryonic stem cells or human induced pluripotent stem cells into hepatocyte-like cells and constructing a three-dimensional hepatocyte model; a grouping module for grouping into a test group and a control group according to miR-122, LDH and Cyto C as a combination of test indicators as claimed in claim 1; a detection module for the IC20 value of a liver toxicity model compound and the data of the combination of test indicators after the liver toxicity model compound acts on the liver toxicity test model; an analysis modeling module for modeling through a linear discriminant analysis modeling method; a prediction module for predicting a test compound using the established prediction model. 2.The method according to claim 1, wherein the three-dimensional hepatocyte model in the step (1) is a liver organoid. 3.The method according to claim 1, wherein the three-dimensional hepatocyte model in the step (1) is a hepatocyte cluster. 4.The method according to claim 1, wherein the three-dimensional hepatocyte model in the step (1) is a hepatocyte spheroid. 5.The method according to claim 1, wherein the miR-122, LDH and Cyto C in the step (2) are selected by a coefficient of variation. 6.The method according to claim 1, wherein the linear discriminant analysis modeling method is a stepwise discriminant analysis method. 7.The method according to claim 1, wherein the Fishers discriminant functions are:

Images