Bio-ink, in-vitro humanized neural model and its application in constructing opioid drug classification prediction model and dose-effect relationship

Abstract

The application belongs to the technical field of in vitro drug evaluation model construction, and particularly relates to a biological ink, an in vitro humanized nerve model and application of the in vitro humanized nerve model in construction of an opioid drug classification prediction model and dose-effect relationship. The biological ink comprises a neuroblastoma cell line, methacrylated gelatin, a photoinitiator and a basement membrane matrix. The in vitro humanized nerve model prepared based on the biological ink provided by the application has microenvironment interaction characteristics of cell-cell and cell-matrix, can more truly restore the complexity of in vivo tissues, can highly bionically simulate the structure and function of human neurons at the cell level, the bionicity is significantly better than traditional two-dimensional culture and animal experiments, and the accuracy and reliability of drug toxicity and dependence evaluation are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of in vitro drug evaluation model construction, and particularly relates to a biological ink, an in vitro humanized nerve model and application of the in vitro humanized nerve model in construction of an opioid drug classification prediction model and dose-effect relationship. BACKGROUND

[0002] With the development of social economy, the demand and consumption of psychotropic drugs are on the rise, the types of abuse are updated and iterated at an accelerated pace, causing the legal management to lag behind, and the abuse of narcotic and psychotropic drugs is still severe and complex.

[0003] To solve the problem of lagging behind in the management of new narcotic and psychotropic drugs in judicial practice, prevent and combat the behavior of alternative abuse of psychotropic drugs, and protect public safety and people's health, judicial organs actively carry out related work, but face many difficulties and challenges. On the one hand, there are many types of psychotropic drugs, complex chemical structures, and unknown mechanisms of action, making it difficult to accurately identify and classify them. On the other hand, there is a lack of sufficient experimental data and scientific basis, making it difficult to determine the specific damage to human function, the strength of addiction, and the related criminal responsibility and penalty standards.

[0004] It is of great significance to clarify the psychotropic toxicity or physiological / psychological dependence of psychotropic drugs for judicial work and related research. At present, the methods for evaluating the psychotropic toxicity or physiological / psychological dependence of drugs mainly include animal experiments and in vitro experiments. Although animal experiments have high biological relevance and reliability, they have problems such as animal welfare, ethics, cost, time, scale, etc. Although in vitro experiments have the advantages of flexibility and high efficiency, the current in vitro models are mainly two-dimensional cell culture models, which cannot simulate the in vivo environment, resulting in inaccurate evaluation of drug absorption, metabolism and toxicity. SUMMARY

[0005] To achieve the above purpose, the application can adopt the following technical solutions:

[0006] In one aspect, the application provides a biological ink, which comprises a neuroblastoma cell line, methacrylated gelatin, a photoinitiator, and a basement membrane matrix.

[0007] Preferably, the above biological ink can satisfy one or more of the following conditions:

[0008] (i) the neuroblastoma cell line is selected from one or more of SH-SY5Y cell line, SK-N-SH cell line, or SK-N-BE cell line;

[0009] (ii) the photoinitiator is selected from one or more of LAP, Irgacure 2959, or Irgacure 819;

[0010] (iii) the neuroblastoma cell line is a single cell suspension;

[0011] (iv) in the bio-ink, the mass percentage of the methacrylated gelatin is 4-6%, and the mass percentage of the basement membrane matrix is 20-30%;

[0012] (v) in the bio-ink, the cell density of the neuroblastoma cell line is 1.2x10^6 / ml-1.2x10^7 / ml.

[0013] Preferably, the above bio-ink can be selected from any one of the following: 1) in the bio-ink, the mass percentage of the methacrylated gelatin can be 4.2%, and the mass percentage of the basement membrane matrix can be 30%; 2) in the bio-ink, the mass percentage of the methacrylated gelatin can be 5.6%, and the mass percentage of the basement membrane matrix can be 20%.

[0014] Another aspect of the present application provides an in vitro humanized neural model prepared by 3D printing using the bio-ink of the present application as a raw material.

[0015] Still another aspect of the present application provides a preparation method of the in vitro humanized neural model of the present application, which can include: printing the bio-ink into at least two droplets by micro-extrusion printing as an in vitro humanized neural model for high-throughput drug toxicology evaluation; or printing the bio-ink into a three-dimensional grid as an in vitro humanized neural model for drug dependence evaluation.

[0016] Preferably, in the above preparation method, the amount of cells in each droplet can be 1.0x10^ 5 cells-1.0x10^ 7 cells.

[0017] Still another aspect of the present application provides an application of the in vitro humanized neural model of the present application in evaluating the psychotropic toxicity or physiological / psychological dependence of a drug.

[0018] Still another aspect of the present application provides an opioid drug classification prediction model based on the in vitro humanized neural model of the present application, which includes:

[0019] Step 1, obtaining data features of neural toxicity target data, randomly selecting candidate features from the data features as root nodes of a classification tree, and selecting the best split point of the data features based on Gini importance to split the nodes to obtain child nodes of the classification tree, recursively selecting the best split point of the data features until the node splitting meets the stopping condition to obtain a classification tree model;

[0020] Step 2, repeating step 1 to obtain multiple classification tree models;

[0021] Step 3, sample the neurotoxicity target data to obtain a plurality of neurotoxicity target sub-data sets, wherein each neurotoxicity target sub-data set is used as a classification training sample, and the number of neurotoxicity target sub-data sets is the same as the number of classification tree models;

[0022] Step 4, training a classification tree model using a classification training sample to obtain a classification prediction label, calculating the accuracy of the classification prediction label and the test label, and completing the training process of a single classification tree model when the accuracy reaches a preset accuracy threshold;

[0023] Step 5, repeating step 4, training a classification tree model using a classification training sample in parallel, and integrating all trained classification tree models to obtain an opioid drug classification prediction model;

[0024] In step 1, the data characteristics of the neurotoxicity target data are obtained by the in-vitro humanized neural model.

[0025] Preferably, in the above opioid drug classification prediction model, after training a classification tree model using a classification training sample in parallel, the method further comprises: performing feature importance sorting on the data characteristics of the neurotoxicity target data to obtain data characteristic contribution degrees of the classification prediction, and constructing a dose-effect relationship of the opioid drug according to the data characteristic contribution degrees.

[0026] More preferably, in the above opioid drug classification prediction model, the dose-effect relationship of the opioid drug is constructed according to the data characteristic contribution degrees, comprising:

[0027] calculating a single toxicity ratio of the opioid drug on a single biological characteristic using a single-variable toxicity calculation formula;

[0028] using the data characteristic contribution degrees as toxicity weights to calculate a toxicity ratio of the opioid drug on each biological characteristic, and performing summation operation on the toxicity ratios of the opioid drug on each biological characteristic to obtain a weighted toxicity ratio total score;

[0029] calculating a total weighted toxicity ratio of the opioid drug at multiple concentrations using a total weighted toxicity ratio calculation formula;

[0030] performing sensitivity analysis on the single toxicity ratio, the weighted toxicity ratio total score, and the total weighted toxicity ratio, and obtaining a target weight scheme according to the sensitivity analysis result, and calculating a comprehensive toxicity ratio of the opioid drug.

[0031] The present application has at least the following advantages:

[0032] (1) The in-vitro humanized nerve model prepared based on the bio-ink provided by the application has cell-cell and cell-matrix microenvironment interaction characteristics, can more truly restore the complexity of in-vivo tissues, can highly bionically simulate the structure and function of human neurons at the cell level, exhibits physiological properties similar to natural nerve tissues, and the high bionics is significantly better than traditional two-dimensional culture, and the accuracy and reliability of drug toxicity and dependence evaluation are improved.

[0033] (2) The construction method of the in-vitro humanized nerve model provided by the application can complete self-assembly and printing of cell clusters in a short time, effectively saves time and cost compared with animal experiments (animal experiments need to go through three processes of modeling, sampling and analysis, and the drug dose used is much higher than that of the in-vitro model, and the modeling cycle is long), and the in-vitro humanized nerve model can replace animal experiments, reduce the species difference of animal experiments (the animal individual difference of animal experiments is large, and there is a species genetic difference with humans), and improve the experimental accuracy.

[0034] (3) Based on the bio-ink provided by the application, a large number of in-vitro humanized nerve models can be quickly prepared on a hole plate, toxicity and dependence screening under multiple experimental conditions can be completed on a hole plate at the same time, the simultaneous evaluation needs of different drugs are met, and the experimental efficiency is greatly improved. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 It is a flowchart for constructing a SH-SY5Y nerve model (SH-SY5Y cell cluster) based on a low-adhesion micropore array; wherein (a) is a schematic diagram of a PDMS film; (b) is a schematic diagram of a 24-well plate device with a selectively gas-permeable PDMS film at the bottom; (c) is a PDMS film with a micropore array that can be placed in a culture well for cell culture;

[0036] Figure 2 It is a cell viability statistics;

[0037] Figure 3 It is the statistical situation of the influence of different concentrations of fentanyl on dopamine release;

[0038] Figure 4 It is the statistical situation of the influence of different concentrations of fentanyl on the expression of four kinds of opioid receptors;

[0039] Figure 5(a) is the expression level of undifferentiated neuron marker Tuj1 in each system; (b) is the expression level of mature differentiated neuron marker MAP2 in each system; (c) is the expression level of neuron differentiation marker neurofilament light chain protein NEFL in each system; (d) is the expression level of nerve fiber formation and maintenance marker neurofilament medium protein NEFM in each system; one-way ANOVA followed by Holm-Šidak post-hoc test, *p<0.05, **p<0.01, ***p<0.001; data derived from 3 independent experiments;

[0040] Figure 6 (a) is the expression level of undifferentiated neuron marker Tuj1 in each system; (b) is the expression level of mature differentiated neuron marker MAP2 in each system; (c) is the expression level of neuron differentiation marker neurofilament light chain protein NEFL in each system; (d) is the expression level of nerve fiber formation and maintenance marker neurofilament medium protein NEFM in each system; one-way ANOVA followed by Holm-Šidak post-hoc test, *p<0.05, **p<0.01, ***p<0.001; data derived from 3 independent experiments;

[0041] Figure 7 is the confusion matrix of the random forest model, where F is fentanyl, H is diacetylmorphine M is morphine, F-1 is 1 uM fentanyl treatment condition, and so on.

[0042] Figure 8 is the amount of dopamine release, cell survival rate, and the contribution of different gene expression levels to classification prediction. DETAILED DESCRIPTION

[0043] The examples are given to better illustrate the present application, but are not intended to limit the present application to the examples only. Therefore, the skilled in the art can make non-essential improvements and adjustments to the embodiments according to the above disclosure, which still fall within the protection scope of the present application.

[0044] The terms used herein are only used to describe specific embodiments and are not intended to limit the present disclosure. Unless there is a clear different meaning in the context, the singular form includes the plural form. As used herein, it should be understood that terms such as "include", "have", "contain", etc. are intended to indicate the presence of features, numbers, operations, components, parts, elements, materials, or combinations. The terms of the present application are disclosed in the specification, and are not intended to exclude the possibility that one or more other features, numbers, operations, components, parts, elements, materials, or combinations thereof can exist or can be added. As used herein, " / " can be interpreted as "and" or "or" depending on the circumstances.

[0045] The embodiment of the present application provides a biological ink, which comprises a neuroblastoma cell line, methacrylated gelatin (hereinafter also referred to as GelMA), a photoinitiator and a basement membrane matrix (hereinafter also referred to as Matrigel).

[0046] It should be noted that GelMA is a kind of light crosslinking hydrogel based on the dual characteristics of temperature sensitivity and light crosslinking, and has good biocompatibility and plasticity. Using GelMA hydrogel as a biological ink to print a neuroblastoma cell line can manufacture a nerve tissue scaffold with complex shape and porous structure, and provide a suitable growth environment for neuroblastoma cells. In addition, the RGD peptide in GelMA can specifically enhance the adhesion of SH-SY5Y cells, promote the expression of immature neuron marker tuj1; the matrix can induce neuron differentiation and promote the expression of mature neuron marker map2; in addition, the combination of GelMA and basement membrane matrix can prepare a biological model that can simulate the in vivo environment and simulate human nerve function (such as the characteristics and functions of human brain neurons), and promote the expression of dopaminergic neuron marker TH; when the two kinds of matrix glue are combined, the expression of the three markers TUJ1\MAP2\TH is induced to the highest.

[0047] It should also be noted that the cell clusters of the biological model constructed based on GelMA and basement membrane matrix in the present application have better stability than the existing cell clusters based on low adhesion micropore array, and at the same time have the advantages of cell clusters (such as having a dense extracellular matrix to reduce the transport efficiency of exogenous drugs and having a biomimetic nature).

[0048] In some specific examples, the neuroblastoma cell line in the above-mentioned biological ink can be selected from one or more of SH-SY5Y cell line, SK-N-SH cell line or SK-N-BE cell line.

[0049] It should be noted that the neuroblastoma cell line in the present application refers to a cell line with neuron characteristics or expressing nerve markers, such as SH-SY5Y cell line, SK-N-SH cell line or SK-N-BE cell line; preferably, the SH-SY5Y cell line, which is a human neuroblastoma cell line, has various neuron characteristics and can be used to study the protective or damaging effect of drugs on human brain neurons. In addition, the biological model constructed by the SH-SY5Y cell line maintains high activity of the cell cluster and significantly down-regulates the proliferation rate within 0-6 days of culture, indicating that the cells appear differentiation phenotype, which is a cell line suitable for constructing a biological model.

[0050] In some specific examples, the above-mentioned biological ink can satisfy one or more of the following conditions:

[0051] (i) the neuroblastoma cell line is selected from one or more of SH-SY5Y cell line, SK-N-SH cell line or SK-N-BE cell line;

[0052] (ii) the photoinitiator is selected from one or more of LAP, Irgacure 2959 or Irgacure 819;

[0053] (iii) the neurospheres are in a single cell suspension;

[0054] (iv) in the bio-ink, the mass percentage of the methacrylated gelatin is 4-6%, and the mass percentage of the basement membrane matrix is 20-30%; specifically, when the mass percentage of the methacrylated gelatin is higher than 8%, the Young's modulus is too large, which is not suitable for cell culture, and in the present application, the mass percentage is preferably 4-6%, such as 4.2%, 4.5%, 4.7%, 4.9%, 5.2%, 5.5%, 5.7% or 5.9%, etc.; in addition, when the mass percentage of the basement membrane matrix is too large, the bio-ink cannot be in a semi-gel state during printing, causing printing difficulties, in addition, more than 30% will not have a positive guiding effect on the expression of neuronal markers, and will increase the economic cost, and in the present application, the mass percentage is preferably 20-30%, such as 23%, 25%, 27% or 29%, etc.

[0055] (v) in the bio-ink, the cell density of the neurospheres is 1.2x10^6 / ml-1.2x10^7 / ml.

[0056] In some specific examples, in the above bio-ink, any one of the following can be preferentially selected:

[0057] 1) in the bio-ink, the mass percentage of GelMA is 4.2%, and the mass percentage of the basement membrane matrix is 30%;

[0058] 2) in the bio-ink, the mass percentage of GelMA is 5.6%, and the mass percentage of the basement membrane matrix is 20%.

[0059] It should be noted that the bio-model prepared from the above two groups of bio-ink is more accurate in evaluating the effect of psychotropic drugs on human body function damage or addiction and evaluating the effect of physiological / psychological dependence caused by psychotropic drugs.

[0060] In some specific examples, the neurospheres in the above bio-ink can be in a single cell suspension form.

[0061] It should be noted that the neurospheres in the bio-ink in the present application can be in a single cell suspension, and the preparation of the single cell suspension is well known in the art, such as can be obtained by trypsin digestion.

[0062] The present application also provides an in-vitro humanized neural model, which is prepared by 3D printing using the bio-ink in the present application as a raw material.

[0063] It should be noted that the three-dimensional bioprinting technology based on 3D printing is a technology of precisely positioning and stacking biological materials, cells and the like according to a preset three-dimensional structure, so as to generate an artificial tissue or organ with specific morphology and function. The biological model constructed based on the technology can be used to evaluate the influence of a drug on different tissues or organs, has high customization, complexity, flexibility and accuracy, can realize spatial distribution and interaction of multiple cell types, multiple biological materials and multiple biological signals, and can simulate the microstructure and macro function of a natural tissue.

[0064] The embodiment of the present application also provides a preparation method of the in-vitro humanized neural model in the present application, which comprises: printing biological ink into at least two droplets through micro-extrusion printing as an in-vitro humanized neural model for high-throughput drug toxicology evaluation; or printing the biological ink into a three-dimensional grid as an in-vitro humanized neural model for drug dependence evaluation.

[0065] In some specific examples, in the preparation method, the amount of cells in each droplet is 1.0x10^ 5 cells-1.0x10^ 7 cells.

[0066] It should be noted that the amount of cells in each droplet in the present application is the limit carrying capacity of two-dimensional cell culture, but in three-dimensional culture, the cells can exceed the carrying capacity after proliferation.

[0067] The embodiment of the present application also provides an application of the in-vitro humanized neural model in the present application in evaluating the psychotropic toxicity or physiological / psychological dependence of a drug.

[0068] In some specific examples, in the application, the drug comprises a psychoactive drug, an antitumor drug or a central nervous system drug.

[0069] It should be noted that the psychoactive drug, the antitumor drug or the central nervous system drug are all drugs known in the art; for example, the psychoactive drug can be a caffeine, a hallucinogen (such as diethylamide of lysergic acid, mescaline, phencyclidine or ketamine), an opiate (such as diacetylmorphine, morphine, methadone, dihydroetorphine, meperidine, buprenorphine or fentanyl), and the like; for another example, the antitumor drug can be carboplatin, vincristine sulfate or cisplatin, and the like; for another example, the central nervous system drug can be an antipsychotic drug (such as risperidone, aripiprazole, chlorpromazine, haloperidol, and the like, which alleviates symptoms by blocking dopamine receptors in the brain or regulating other neurotransmitter systems), an antiepileptic drug (such as carbamazepine, phenytoin sodium or valproate sodium, and the like, which prevents abnormal discharge by stabilizing nerve cell membranes or regulating neurotransmitters), an antidepressant drug (such as a tricyclic antidepressant (such as amitriptyline, imipramine or chlorimipramine), a selective serotonin reuptake inhibitor (such as paroxetine, sertraline, fluoxetine), a 5-HT1A receptor partial agonist, and the like, which exerts an antidepressant effect by increasing the level of 5-hydroxytryptamine in the brain, agonizing 5-HT1A receptors or other mechanisms), an anxiolytic drug (such as diazepam, lorazepam, and the like, which exerts an effect by enhancing the action of gamma-aminobutyric acid in the brain, which is an inhibitory neurotransmitter), a ganglionic blocking drug (such as mecamylamine, pirenzepine or hexamethonium, and the like, which can block the transmission of nerve impulses), a nerve terminal blocking drug (such as oryzanol, cobalamin or methylcobalamin), an adrenergic receptor blocking drug (such as metoprolol tartrate, metoprolol succinate or bisoprolol, and the like), and the like.

[0070] The embodiment of the present application also provides an opiate drug classification prediction model based on the in-vitro humanized neural model in the present application, and the opiate drug classification prediction model comprises:

[0071] Step 1, obtaining data features of the neurotoxicity target data, randomly selecting candidate features as root nodes of a classification tree from the data features, and selecting the best split point of the data features based on Gini importance to split the nodes to obtain child nodes of the classification tree, recursively selecting the best split point of the data features until the node splitting meets a stop condition to obtain a classification tree model;

[0072] Step 2, repeating step 1 to obtain a plurality of classification tree models;

[0073] Step 3, performing sampling processing on the neurotoxicity target data to obtain a plurality of neurotoxicity target sub-data sets, wherein each neurotoxicity target sub-data set is taken as a classification training sample, and the number of the neurotoxicity target sub-data sets is the same as the number of the classification tree models;

[0074] Step 4, training a classification tree model using a classification training sample to obtain a classification prediction label, calculating the accuracy of the classification prediction label and the test label, and completing the training process of a single classification tree model when the accuracy reaches a preset accuracy threshold;

[0075] Step 5, repeating step 4, training a classification tree model using a classification training sample in parallel, and integrating all trained classification tree models to obtain an opioid drug classification prediction model.

[0076] In step 1, the data characteristics of the neurotoxicity target data are obtained by the in vitro humanized neural model in the application.

[0077] In some specific examples, in the above opioid drug classification prediction model, after parallel training of a classification tree model using a classification training sample, the method further comprises: performing feature importance sorting on the data characteristics of the neurotoxicity target data to obtain the data characteristic contribution degree of the classification prediction, and constructing the dose-effect relationship of the opioid drug according to the data characteristic contribution degree.

[0078] In some specific examples, in the above opioid drug classification prediction model, constructing the dose-effect relationship of the opioid drug according to the data characteristic contribution degree comprises:

[0079] calculating the single toxicity ratio of the opioid drug on a single biological characteristic using a single variable toxicity calculation formula;

[0080] using the data characteristic contribution degree as a toxicity weight to calculate the toxicity ratio of the opioid drug on each biological characteristic, and performing summation operation on the toxicity ratio of the opioid drug on each biological characteristic to obtain a weighted toxicity ratio total score;

[0081] calculating the total weighted toxicity ratio of the opioid drug at multiple concentrations using a total weighted toxicity ratio calculation formula;

[0082] performing sensitivity analysis on the single toxicity ratio, the weighted toxicity ratio total score, and the total weighted toxicity ratio, and obtaining a target weight scheme according to the sensitivity analysis result, and calculating the comprehensive toxicity ratio of the opioid drug.

[0083] In order to better understand the application, the content of the application will be further illustrated below in combination with specific examples, but the content of the application is not limited to the following examples.

[0084] I. Construction of humanized neural model

[0085] In the following examples, GelMA and photoinitiator LAP are from EFL (Engineering For Life); Matrigel is from Corning, item number 356237.

[0086] Example 1

[0087] (1) Cell seeding: SH-SY5Y cell line (human neuroblastoma cell line) (cell concentration 1 x 10^4 cells / mL) was seeded in DMEM / F12 medium containing 10% fetal bovine serum, 1% penicillin-streptomycin double antibody (medium type: 11320033, Thermo, 1:1 (V / V) mixture of DMEM and Ham's F-12, which contains glucose, amino acids and vitamins of DMEM and F-12) in a 37°C, 5% carbon dioxide incubator, and when the cells grew to 90% confluence, stable passage was carried out (i.e. 2D culture cells were obtained);

[0088] (2) Manufacture of neural microspheres (hereinafter also referred to as cell microspheres): when the cells proliferated to 90% confluence in the two-dimensional culture dish, the proliferated SH-SY5Y cells were digested into a single cell suspension with 0.25% trypsin; then the single cell suspension was seeded into an oxygen-permeable device with a regular hexagonal honeycomb micro-pore array (an oxygen-permeable device with a regular hexagonal honeycomb micro-pore array was prepared by pouring polydimethylsiloxane (PDMS) into an etched silicon plate mold, with a single hole inner diameter of 126 μm and a PDMS membrane thickness of about 0.5 mm) and cultured for 48 h (2 days) to aggregate into neural microspheres;

[0089] (3) Bio-ink mixing: after centrifugation of the neural microspheres at 1000 rpm for 3 min, the supernatant was discarded, and the cells were resuspended to prepare a bio-ink (the photoinitiator LAP powder was dissolved in D-PBS buffer to form a 0.25% mass fraction solution, then the solution was used to dissolve GelMA to form a GelMA solution with a GelMA mass concentration of 6%, the GelMA solution and Matrigel were mixed in a volume ratio of 7:3 to form a mixture, and the mixture and the single cell suspension were mixed to form a bio-ink; in the bio-ink, the final GelMA mass fraction was 4.2%, the Matrigel mass fraction was 30%, and the cell density was 1.0 x 10^ 7 / mL);

[0090] (4) Droplet-like nerve model manufacturing: The prepared bio-ink was sucked into a 3 mL syringe, and a point injection needle with an inner diameter of 0.34 mm was assembled on a 3D printer (SunP BioMaker 4) (photocured for 20 s after printing was completed); in the 96-well plate high-throughput toxicology characterization experiment, 10 μl of droplets were printed in each well, and the cell amount in each well was 1.0x10^5, which played a better effect in subsequent experimental measurement; wherein the printing parameters included: the printing used a point injection needle with a size of 23G (inner diameter of 0.34 mm), printing speed: 4 mm / s; extrusion speed: 1 mm³ / s, to ensure that the droplet extrusion amount was 10 uL. In addition, a three-dimensional grid-like model for drug dependence evaluation was printed in a 60 mm culture dish by micro-extrusion (printing parameters: 23G (inner diameter of 0.34 mm), printing speed: 4 mm / s; extrusion speed: 1 mm³ / s, extrusion form: 15x15x1.2 mm grid-like printing layer height 0.4, printing layer number 3), and the grid-like structure was 15 mmx15 mmx1.2 mm, and about 100 μL of bio-ink was required for each structure.

[0091] Example 2

[0092] The difference between Example 2 and Example 1 is that the bio-ink mixture in step (3) is different, and the others are the same as Example 1; in Example 2, the bio-ink mixture includes: centrifuging the neural microspheres at 1000 rpm for 3 min, then discarding the supernatant, resuspending the cells, and then preparing the bio-ink (dissolving the photoinitiator LAP powder in D-PBS buffer to form a 0.25% solution, then dissolving GelMA in the solution to form a GelMA solution with a GelMA mass concentration of 6%, mixing the GelMA solution and Matrigel to form a mixture, and mixing the mixture with the single cell suspension to form the bio-ink; in the bio-ink, the final GelMA mass fraction is 5.6%, the Matrigel mass fraction is 20%, and the cell density is 1.0x10^ 7 / mL).

[0093] Comparative Example 1

[0094] According to Figure 1The flowchart shown realizes the rapid construction of SH-SY5Y neural model (SH-SY5Y cell cluster) based on a low-adhesion micropore array, as follows: an oxygen-permeable device with a honeycomb-shaped micropore array of regular hexagons is prepared by pouring polydimethylsiloxane (PDMS) into an etched silicon plate mold, with a single-hole inner diameter of 126 μm and a PDMS membrane thickness of about 0.5 mm; the low-adhesion microenvironment effectively guides the self-assembly of SH-SY5Y cells, and after 48 hours of culture, microscopic imaging confirms that a large number of cell clusters with uniform morphology, high stability and good activity are formed in the device, providing a reliable in vitro model for efficient construction of a neural model.

[0095] However, SH-SY5Y is a semi-adherent cell, and previous experiments have shown that it will be in a suspended state under environmental stress, and the SH-SY5Y cell cluster planted in the PDMS oxygen-permeable micropore array device cannot meet the subsequent evaluation of neurotoxicity, and in the presence of drugs, the morphology and structure of the cell cluster are unstable and prone to deformation or rupture.

[0096] Comparative Example 2

[0097] Comparative Example 2 is the 2D cultured cells prepared in step (1) of Example 1.

[0098] Comparative Example 3

[0099] The difference between Comparative Example 3 and Example 2 is that the biological ink is mixed differently in step (3), and the others are the same as in Example 2; in Example 2, the biological ink is mixed as follows: after centrifuging the neural microspheres at 1000 rpm for 3 min, the supernatant is discarded, and the cells are resuspended to prepare the biological ink, as follows: the photoinitiator LAP powder is dissolved in D-PBS buffer to form a 0.25% solution, then the GelMA is dissolved in the solution to form a GelMA solution with a GelMA mass concentration of 6%, and the GelMA solution and single-cell suspension are mixed to obtain the biological ink; in the biological ink, the cell density is 1.0x10^ 7 / mL.

[0100] Comparative Example 4

[0101] The difference between Comparative Example 3 and Example 2 is that the biological ink is mixed differently in step (3), and the others are the same as in Example 2; in Example 2, the biological ink is mixed as follows: after centrifuging the neural microspheres at 1000 rpm for 3 min, the supernatant is discarded, and the cells are resuspended to prepare the biological ink, as follows: the Matrigel solution and single-cell suspension are mixed to obtain the biological ink; in the biological ink, the cell density is 1.0x10^ 7 / mL.

[0102] Comparative Example 5

[0103] The difference between Comparative Example 3 and Example 1 is that the bio-ink mixing in step (3) is different, and in addition, Comparative Example 3 is soaked in a 2% calcium chloride solution in normal saline for 1 min after printing to complete the solidification process, and the others are the same as Example 1; in Comparative Example 3, the bio-ink mixing includes: centrifuging the neural microspheres at 1000 rpm for 3 min, then discarding the supernatant, and then preparing the bio-ink after resuspending the cells, which specifically includes: dissolving sodium alginate (sodium alginate, sigma, 180947) in D-PBS buffer solution to form a sodium alginate solution with a sodium alginate mass concentration of 2%; dissolving gelatin (gelatin, sigma, V900863) in D-PBS buffer solution to form a 10% gelatin solution; mixing the sodium alginate solution and the gelatin solution in a volume ratio of 1:1 to obtain a mixture, and mixing the mixture and the single cell suspension to obtain the bio-ink; in the bio-ink, the final sodium alginate mass fraction is 1%, the gelatin mass fraction is 5%, and the cell density is 1.0x10^ 7 / mL.

[0104] Comparative Example 6

[0105] The difference between Comparative Example 4 and Example 1 is that the bio-ink mixing in step (3) is different, and in addition, Comparative Example 4 is soaked in a 2% calcium chloride solution in normal saline for 1 min after printing to complete the solidification process, and the others are the same as Example 1; in Comparative Example 3, the bio-ink mixing includes: centrifuging the neural microspheres at 1000 rpm for 3 min, then discarding the supernatant, and then preparing the bio-ink after resuspending the cells, which specifically includes: dissolving sodium alginate (sodium alginate, sigma, 180947) in D-PBS buffer solution to form a sodium alginate solution with a sodium alginate mass concentration of 4%; dissolving gelatin (gelatin, sigma, V900863) in D-PBS buffer solution to form a 10% gelatin solution; mixing the sodium alginate solution and the gelatin solution in a volume ratio of 1:1 to obtain a mixture, and mixing the mixture and the single cell suspension to obtain the bio-ink; in the bio-ink, the final sodium alginate mass fraction is 2%, the gelatin mass fraction is 5%, and the cell density is 1.0x10^ 7 / mL).

[0106] II. Application of humanized neural model

[0107] (I) Cell differentiation phenotype

[0108] The humanized neural model prepared in Example 1 maintains high activity (p<0.001) within 0-6 days of culture (37°C, CO2) as a high-throughput drug toxicity evaluation model and as a drug dependence evaluation model, and the proliferation rate is significantly down-regulated, indicating that the cells appear a differentiation phenotype. Figure 2

[0109] (II) Fentanyl toxicity characterization study​

[0110] (1) The drug dependence evaluation models prepared in Example 1 and Example 2 and Comparative Examples 2 to 6 were respectively cultured in 35 mm dishes, 2 mL of culture medium was added in each dish, and the culture was carried out for 24 h (37°C, CO2);

[0111] (2) After the culture was completed, 100 uL of fentanyl solution (fentanyl hydrochloride was dissolved in methanol to prepare a mother liquor with a concentration of 100 mM, and then diluted with cell culture medium to different concentrations) of different concentrations (1 mM, 10 mM, 100 mM) was added for incubation for 24 h;

[0112] (3) After the incubation was completed, the CCK8 kit (YEASEN, 40203ES60) was used to detect the cell proliferation activity under the action of different concentrations of fentanyl (the detection method was referred to the kit instruction), and thus the fentanyl neurotoxicity characterization data was obtained.

[0113] (Three) Fentanyl neurodependence characterization

[0114] The increase of the expression amount of opioid receptors indicates that the neurodependence of the cells to fentanyl increases, which can be used as a reference standard for quantifying fentanyl crime; the desensitization and expression reduction of opioid receptors is the main mechanism of opioid drug tolerance; and the increase and decrease of the expression amount of dopamine and dopamine receptors after tolerance is the main mechanism of opioid drug addiction. Therefore, the influence of psychotropic drugs on the release amount of dopamine, the expression of four kinds of opioid receptors (mu opioid receptor, kappa opioid receptor, delta opioid receptor and nociceptin receptor (NOP)), and the expression of dopamine receptors in the in-vitro 3D neural model can be evaluated.

[0115] The specific steps in this characterization are as follows:

[0116] (1) The drug dependence evaluation models prepared in Example 1 and Example 2 and Comparative Examples 2 to 6 were respectively cultured in 35 mm dishes, 2 mL of culture medium was added in each dish, and the culture was carried out for 24 h (37°C, CO2);

[0117] (2) After the culture was completed, 100 uL of fentanyl solution (fentanyl hydrochloride was dissolved in methanol to prepare a mother liquor with a concentration of 100 mM, and then diluted with cell culture medium to different concentrations) of different concentrations (1 mM, 10 mM, 100 mM) was added for incubation for 24 h;

[0118] (3) After the incubation was completed, the CCK8 kit (YEASEN, 40203ES60) was used to detect the cell proliferation activity under the action of different concentrations of fentanyl (the detection method was referred to the kit instruction), and thus the fentanyl neurotoxicity characterization data was obtained.

[0119] (4) After sufficient lysis (generally 20 min), centrifuge at 1000 rpm for 5 min, discard the supernatant, and add 5 mL of complete medium to repeat the washing and centrifugation once to obtain the cell sample;

[0120] (5) The expression levels of μ, δ and σ opioid receptors were determined by qPCR kit (SYBR GREEN I Master Mix (11184ES08, YEASEN, refer to the kit instructions for testing method), and the expression level of secreted dopamine was determined by enzyme-linked immunosorbent assay kit (Bioek: Human Dopamine ELISA Detection Kit).

[0121] The statistical results of dopamine release under different concentrations of fentanyl (0, 1 μM, 10 μM, 100 μM) are shown in Table 1, and the expression levels of the four opioid receptors under different concentrations of fentanyl (0, 1 μM, 10 μM, 100 μM) are shown in Table 2. Figure 3 Figure 4

[0122] In addition, the expression of neuronal differentiation biomarkers Tuj1, MAP2, NEfL, NEfM and TH of the cell samples obtained in Examples 1 to 2 and Comparative Examples 2 to 6 was detected by qPCR kit (SYBR GREEN I Master Mix (11184ES08, YEASEN, refer to the kit instructions for testing method), and the results are shown in Tables 3 and 4. Figure 5 Figure 6 The expression levels of Tuj1, MAP2, NEfL, NEfM and TH in Examples 1 and 2 were significantly increased (the expression levels of the neural markers TUJ1\MAP2\TH in the model in Example 2 were higher than those in Example 1), indicating that the 3D neural model in the application had good biological activity and neural related functionality, and could be used for subsequent high-throughput drug screening and neurotoxicity evaluation. Moreover, the expression levels of neuronal differentiation biomarkers in the mixed system were significantly increased compared with the pure GelMA and Matrigel system or other mixed systems.

[0123] (Four) Construction of the relationship between the drug toxicology evaluation model and the opioid drug classification prediction model as a drug dependence evaluation model

[0124] In the examples of the application, the data characteristics of the neurotoxicity target data refer to factors related to the characteristics of the biological body affected by the opioid drug, such as the amount of dopamine secretion, cell survival rate, apoptosis index (Bax / bcl2 ratio), expression levels of different representative genes (μ opioid receptor, κ opioid receptor, δ opioid receptor, nociceptive peptide receptor (NOP), dopamine receptor and mitochondrial dynamin), etc. ​​​

[0125] As an embodiment of the present application, the opioid drug classification prediction model comprises:

[0126] Step 1, obtaining data features of the neurotoxicity target data, randomly selecting candidate features as root nodes of the classification tree from the data features, and selecting the best split point of the data features based on the Gini importance to split the nodes to obtain child nodes of the classification tree, recursively selecting the best split point of the data features until the node splitting meets the stopping condition to obtain the classification tree model;

[0127] Step 2, repeating step 1 to obtain a plurality of classification tree models;

[0128] Step 3, sampling the neurotoxicity target data to obtain a plurality of neurotoxicity target sub-data sets, wherein each neurotoxicity target sub-data set is a classification training sample, and the number of neurotoxicity target sub-data sets is the same as the number of classification tree models;

[0129] Step 4, training a classification tree model using a classification training sample to obtain a classification prediction label, calculating the accuracy of the classification prediction label and the test label, and completing the training process of a single classification tree model when the accuracy reaches a preset accuracy threshold;

[0130] Step 5, repeating step 4, training a classification tree model using a classification training sample in parallel, and integrating all trained classification tree models to obtain an opioid drug classification prediction model.

[0131] Further, after training a classification tree model using a classification training sample in parallel, the method further comprises: performing feature importance sorting on the data features of the neurotoxicity target data to obtain the data feature contribution degree of the classification prediction, and constructing the dose-effect relationship of the opioid drug according to the data feature contribution degree.

[0132] In the embodiment of the present application, training a classification tree model using a classification training sample can improve the training efficiency.

[0133] In the embodiment of the present application, the stopping condition for node splitting can be that the classification tree node reaches the maximum depth.

[0134] In the embodiment of the present application, the opioid drug classification prediction model can use a random forest prediction model, wherein the random forest (Random Forest) is a machine learning model based on ensemble learning, which improves the prediction performance and generalization ability by combining multiple decision trees.

[0135] The confusion matrix of the random forest model refers to Figure 7Confusion matrices can be used to evaluate the predictive performance of opioid classification prediction models, including: analyzing the model's classification performance in different categories by utilizing the correspondence between the true and predicted labels of each category in the confusion matrix; obtaining the number of correct predictions for each category by analyzing the diagonal elements in the confusion matrix, further evaluating the model's classification accuracy; evaluating the model's potential misclassification in certain categories by combining the off-diagonal elements in the confusion matrix; and optimizing model parameter settings or performing data augmentation based on the confusion matrix results to improve the model's overall classification performance.

[0136] Reference Figure 8 The diagram shown illustrates the contribution of data features to classification prediction in this invention.

[0137] As an embodiment of the present invention, the step of constructing the dose-response relationship of opioid drugs based on the contribution of data features includes:

[0138] The single toxicity ratio of opioid drugs on a single biological characteristic was calculated using the univariate toxicity calculation formula.

[0139] The contribution of data features is used as the toxicity weight. The toxicity ratio of opioids on each biological feature is calculated. The toxicity ratios of opioids on each biological feature are summed to obtain the weighted toxicity ratio total score.

[0140] The total weighted toxicity ratio of opioid drugs at multiple concentrations was calculated using the total weighted toxicity ratio calculation formula.

[0141] Sensitivity analysis was performed on the single toxicity ratio, the total weighted toxicity ratio, and the total weighted toxicity ratio. Based on the results of the sensitivity analysis, the target weight scheme was obtained, and the overall toxicity ratio of opioids was calculated.

[0142] In this embodiment of the invention, the single toxicity ratio of an opioid drug on a single biological characteristic is calculated using a univariate toxicity calculation formula, which can be expressed by the following formula:

[0143]

[0144] in, For the first Univariate toxicity ratios for each biological characteristic For morphine concentration, This refers to the concentration of diacetylmorphine. For the first A measurement function for a biological characteristic.

[0145] In the embodiment of the present application, the data feature contribution degree is taken as a toxicity weight, the toxicity ratio of the opiate drug on each biological feature is calculated, the toxicity ratio of the opiate drug on each biological feature is summed to obtain a total weighted toxicity ratio score The following calculation formula can be used:

[0146]

[0147] wherein, is the toxicity score of the opiate drug on the i-th feature, is the weight of the i-th biological feature, is the total number of biological features.

[0148] In the embodiment of the present application, the total weighted toxicity ratio calculation formula can use the following formula:

[0149]

[0150] wherein, is the number of concentrations tested.

[0151] Through the calculation of the univariate toxicity ratio, the weighted toxicity ratio and the total weighted toxicity ratio, the step-by-step analysis of the toxicity mechanism and the comprehensive comparison of the drug toxicity can be realized.

[0152] Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the purpose and scope of the present application, and all of them should be covered in the scope of the claims of the present application.​​

Claims

1. An opioid drug classification prediction model based on an in vitro humanized neural model, characterized in that, The method for constructing the opioid drug classification prediction model comprises: Step 1, obtaining data features of neurotoxicity target data, randomly selecting candidate features as root nodes of a classification tree from the data features, and selecting the best split point of the data features based on Gini importance to split the nodes to obtain child nodes of the classification tree, recursively selecting the best split point of the data features until the node splitting meets the stopping condition to obtain a classification tree model; Step 2, repeating step 1 to obtain multiple classification tree models; Step 3, sampling the neurotoxicity target data to obtain multiple neurotoxicity target sub-data sets, wherein each neurotoxicity target sub-data set is a classification training sample, and the number of neurotoxicity target sub-data sets is the same as the number of classification tree models; Step 4, training a classification tree model using a classification training sample to obtain a classification prediction label, calculating the accuracy of the classification prediction label and the test label, and completing the training process of a single classification tree model when the accuracy reaches a preset accuracy threshold; Step 5, repeating step 4, training a classification tree model using a classification training sample in parallel, and integrating all trained classification tree models to obtain an opioid drug classification prediction model; In step 1, the data features of the neurotoxicity target data are obtained by an in vitro humanized neural model; The in vitro humanized neural model is prepared by 3D printing using biological ink as raw material; The biological ink comprises a neuroblastoma cell line, methacrylated gelatin, a photoinitiator and a basement membrane matrix; The biological ink meets the following conditions: (i) the neuroblastoma cell line is selected from one or more of SH-SY5Y cell line, SK-N-SH cell line or SK-N-BE cell line; (ii) the photoinitiator is selected from one or more of LAP, Irgacure 2959 or Irgacure 819; (iii) the neuroblastoma cell line is a single cell suspension; (iv) in the biological ink, the mass percentage of methacrylated gelatin is 4% to 6%, and the mass percentage of basement membrane matrix is 20% to 30%; (v) the cell density of the neuroblastoma cell line in the bio-ink is 1.2 x 10 6 / ml - 1.2 x 10 7 / ml.

2. The in vitro humanized neural model-based opioid drug classification prediction model according to claim 1, wherein, The biological ink can be selected from any one of the following: 1) in the biological ink, the mass percentage of methacrylated gelatin is 4.2%, and the mass percentage of basement membrane matrix is 30%; 2) in the biological ink, the mass percentage of methacrylated gelatin is 5.6%, and the mass percentage of basement membrane matrix is 20%.

3. The in vitro humanized neural model-based opioid drug classification prediction model according to claim 1 or 2, characterized in that, The preparation method of the in vitro humanized neural model comprises: printing the biological ink into at least two droplets by micro-extrusion printing to obtain an in vitro humanized neural model for high-throughput drug toxicology evaluation; or printing the biological ink into a three-dimensional grid to obtain an in vitro humanized neural model for drug dependence evaluation.

4. The opioid drug classification prediction model based on an in-vitro humanized neural model according to claim 3, wherein, The amount of cells in each droplet was 1.0 x 10 5 cells-1.0 x 10 7 cells.

5. The in vitro humanized neural model-based opioid drug classification prediction model of claim 1, 2, or 4, wherein, After training a classification tree model using a classification training sample in parallel, the method further comprises: performing feature importance sorting on the data features of the neurotoxicity target data to obtain data feature contribution of the classification prediction, and constructing a dose-effect relationship of the opioid drug according to the data feature contribution.

6. The in vitro humanized neural model-based opioid drug classification prediction model according to claim 5, wherein, The method for constructing a dose-effect relationship of an opioid drug according to data feature contribution comprises: The single toxicity ratio of the opioid drug on a single biological characteristic is calculated by using a single variable toxicity calculation formula; The toxicity ratio of the opioid drug on each biological characteristic is calculated by taking the data feature contribution degree as a toxicity weight, and the weighted toxicity ratio total score is obtained by performing summation operation on the toxicity ratio of the opioid drug on each biological characteristic; The total weighted toxicity ratio of the opioid drug under multiple concentrations is calculated by using a total weighted toxicity ratio calculation formula; The single toxicity ratio, the weighted toxicity ratio total score and the total weighted toxicity ratio are subjected to sensitivity analysis, the target weight scheme is obtained according to the sensitivity analysis result, and the comprehensive toxicity ratio of the opioid drug is calculated.

Images