A method, storage chip and device for efficiently evaluating the efficacy of in vitro combined drug administration

Abstract

The present invention belongs to the field of drug research and development, and specifically discloses a method, storage chip and device for efficiently evaluating the efficacy of in vitro combined drug therapy. The present invention directly solves the problems of complexity and high cost of experiments and data processing in the existing drug development process by adopting an AI model and a simplified experimental mode. The AI ​​model optimizes concentration selection based on historical drug data, thereby reducing redundant concentrations and unnecessary number of experiments in the experiment, and improving the sensitivity and reliability of the evaluation. By simplifying the experimental design to single drug concentration point testing and a 3x3 combined drug matrix, the complexity and cost of the experiment are significantly reduced, while the speed of data processing and analysis is accelerated. The method disclosed in the present invention avoids excessive combined drug pre-testing by directly conducting single drug and 3x3 matrix combined drug testing, which can not only more effectively identify potential synergies between drugs, but also provide more accurate and practical data to support future clinical research.

Description

Technical Field

[0001] The present invention belongs to the field of drug research and development, and specifically discloses a method, a storage chip and a device for efficiently evaluating the efficacy of in vitro combined drug use. Background Art

[0002] In the field of modern pharmaceutical research and development, in vitro drug evaluation is a key step in the drug development process. This process not only helps to screen active ingredients at an early stage, but also provides a theoretical basis and data support for subsequent clinical trials. In the development of anti-cancer drugs, the evaluation of in vitro combination drugs is particularly important, because cancer usually involves multiple biological pathways, and a single drug is often difficult to achieve the desired therapeutic effect. Therefore, exploring the combined use of multiple drugs in order to discover synergistic or synergistic effects has become a research hotspot.

[0003] The major background technologies mainly involve strategies for multi-drug combination therapy, such as fixed concentration combination, fixed ratio combination and matrix combination. These methods are all aimed at exploring the synergistic effects of different drug combinations in vitro to provide a scientific basis for clinical treatment. The minor background technologies focus more on specific matrix combination modes, such as 6x6 or 9x9 matrices, which are currently the most commonly used evaluation modes. By systematically changing the concentration combination of two or more drugs, their combined effects on inhibiting cancer cell growth are evaluated.

[0004] Defects and shortcomings of existing technology

[0005] Although the existing in vitro combination drug evaluation methods can provide effective data to a certain extent, they still have some significant defects:

[0006] 1. Complexity and high cost of experiments and data processing: Common matrix combination modes such as 6x6 or 9x9 require testing drug combinations at multiple concentrations, which not only makes the experimental design complicated, but also the experimental operation is cumbersome, the amount of data is huge, and processing and analyzing this data requires a lot of time and resources. This high complexity makes the entire evaluation process expensive and inefficient.

[0007] 2. Irrationality in drug concentration selection: In the matrix combination model, uniform or logarithmic concentration increases are often used. Although this method is systematic, it does not always accurately reflect the actual situation in clinical use. Many concentration combinations may be redundant, which not only increases the burden of the experiment, but may also mask more effective concentration combinations.

[0008] 3. Insufficient sensitivity and reliability of analytical methods: Existing efficacy evaluation methods often lack sensitivity when dealing with complex data, especially when analyzing drug synergy in the low concentration range, which may lead to misunderstanding or inaccurate evaluation of the true effect of the drug. Summary of the invention

[0009] In response to the above problems, the present invention directly solves the problems of complexity and high cost of experiments and data processing in the existing drug development process by adopting AI models and simplified experimental modes. The AI ​​model optimizes concentration selection based on historical drug data, thereby reducing redundant concentrations and unnecessary number of experiments in the experiment, and improving the sensitivity and reliability of the evaluation; and by reducing the experimental mode, the experimental design can be reduced from a complex matrix to a 3x3 matrix, which significantly reduces the complexity and cost of the experiment, while speeding up data processing and analysis; the method disclosed in the present invention can not only more effectively identify potential synergies between drugs, but also provide more accurate and practical data to support future clinical research.

[0010] The technical solution of the present invention is as follows:

[0011] A method for efficiently evaluating the efficacy of in vitro combined drug therapy, characterized in that it comprises the following steps:

[0012] Step 1. AI model development: An AI model is constructed using a historical medication dataset. The AI ​​model combines the drug's SMILES sequence, concentration information, and cell type, and is trained and validated using machine learning techniques to predict cell survival rates at different drug concentrations and cell types.

[0013] Step 2. Prediction of combined drug concentration: Use the AI ​​model obtained in step 1 to predict the combined effects of different drug combinations at specific concentrations.

[0014] Step 3. Comprehensive test of single drug concentration point test and 3x3 combination test: Directly use the AI ​​model prediction results of step 2 to conduct experiments on single drug and 3x3 matrix combination. Test the cell survival rate of a single drug at different concentration points, and conduct a 3x3 matrix combination test at the same time. The three recommended concentrations of each drug form a matrix intersection to test the cell response under different combinations.

[0015] Step 4. Dose-response curve analysis: After obtaining data from the single-drug and combination drug tests in step 3, perform dose-response curve analysis to predict and verify the synergistic effect of combination drug use; determine whether there is synergistic, additive or antagonistic effect between drugs, and provide a basis for subsequent drug use strategies.

[0016] Step 5. Experimental optimization and adjustment: Based on the results of steps 1-4, optimize and adjust the experimental design, including but not limited to adjusting the drug concentration range, reselecting drug combinations, or modifying other experimental conditions.

[0017] Furthermore, in the above-mentioned method for efficiently evaluating the efficacy of in vitro combined drug therapy, step 1 comprises the following specific methods:

[0018] (11) Collect and organize the following data:

[0019] S: SMILES sequence of the drug represented by an ASCII string:

[0020] C: drug concentration;

[0021] T: cell type;

[0022] R: CTG readings, cell viability:

[0023] (12) Using the above data, a model f(S,C,T)->R is constructed. The goal of this model is to predict the cell survival rate R under a specific SMILES sequence S, drug concentration C, and cell type T;

[0024] The specific construction process of the above model is as follows:

[0025] Feature Engineering:

[0026] SMILES sequence processing: convert SMILES sequences into numerical molecular descriptors or directly process sequence data through neural networks;

[0027] Drug concentration processing: Standardize the drug concentration C value to fit the input range of the model;

[0028] Cell type encoding: converting cell type T into a processable form;

[0029] Model selection and training:

[0030] Choose a machine learning model that is suitable for processing such data: including but not limited to random forests, support vector machines, or deep learning models;

[0031] Use historical data sets for model training: During the training process, optimize the model parameters to minimize the prediction error, and use the mean square error (MSE) as the loss function: E = (1 / N) * Σ (R_i-R'_i)^2,

[0032] Where R_i is the actual cell survival rate, R'_i is the cell survival rate predicted by the model, and N is the number of samples;

[0033] Model Evaluation:

[0034] The predictive performance of the model was evaluated by cross-validation techniques;

[0035] Model Application:

[0036] The trained AI model is used to predict the effect of new drug combinations, that is, the model f is applied to practical problems and R is predicted based on given S, C, and T.

[0037] Furthermore, in the above-mentioned method for efficiently evaluating the efficacy of in vitro combined medication, step 2 comprises the following specific methods:

[0038] (13) Model application

[0039] Use the developed model f to handle the combined effects of two drugs; drugs S1 and S2, at concentrations C1 and C2 respectively, predict the cell survival rate R_comb when they are combined in cell type T;

[0040] R_comb formula description:

[0041] The prediction model of the effect of combined medication is expressed as: R_comb = f(S1, C1, T) + f(S2, C2, T) + Interaction(S1, S2, C1, C2, T)

[0042] in:

[0043] f(S,C,T) represents the predicted cell survival rate of a single drug at a specific concentration and cell type;

[0044] Interaction (S1, S2, C1, C2, T) is a supplementary term used to estimate the possible interaction between two drugs and is a linear or nonlinear function;

[0045] Variable explanation:

[0046] S1, S2: SMILES sequences of the two drugs;

[0047] C1, C2: concentrations of the two drugs, respectively;

[0048] T: cell types used in the experiments;

[0049] R_comb: cell survival rate when two drugs are used in combination in a specific cell type;

[0050] When the AI ​​model is used to predict the effect of combined drug therapy, if the AI ​​model has been developed in step 1, then when predicting the recommended concentration of 3x3 combined drug therapy, there is no need to enter the specific concentration values ​​C1 and C2. The model will recommend the best concentration combination based on historical data and its internal logic.

[0051] Specifically, the AI ​​model uses existing drug SMILES sequences, possible cell types and other information to predict the effects of combined medications at different concentration combinations. The goal of the model is to find possible optimal concentration points through machine learning techniques. These points are learned from previous data sets and represent the points with the best cell survival rate or the most research value at different concentration combinations.

[0052] Therefore, in practical applications, the AI ​​model in step 2 will output a 3x3 concentration matrix, in which each point represents a concentration combination that the model predicts to have the most likely synergistic effect. This process reduces experimental redundancy and improves experimental efficiency and accuracy.

[0053] In summary, once the AI ​​model is correctly trained and validated in step 1, it can automatically predict drug concentration combinations suitable for further experimental validation without the need for manual input of these concentration values. This approach greatly simplifies the experimental design and implementation process and makes research more efficient.

[0054] (21) Screening mechanism

[0055] In order to screen out the combination showing the strongest synergistic effect from possible drug combinations and concentrations, a standard for quantifying synergy is defined. The above standard is the synergy index Synergy Score:

[0056] Synergy Score formula description:

[0057] Synergy Score=R_comb-(R_pred(S1,C1,T)+R_pred(S2,C2,T))

[0058] Among them, R_pred(S1,C1,T) and R_pred(S2,C2,T) are the predicted effects of single drugs.

[0059] R_comb is the actual predicted effect of combined medication;

[0060] Synergy Score calculation steps:

[0061] For each pair of drugs (S1, S2) and each possible concentration combination (C1, C2), R_comb was calculated using the above model;

[0062] Calculate the Synergy Score of each pair of combinations;

[0063] The three concentration combinations with the highest Synergy Scores were selected.

[0064] Furthermore, in the above-mentioned method for efficiently evaluating the efficacy of in vitro combined medication, step 3 comprises the following specific methods:

[0065] Single drug concentration point test and 3x3 combined comprehensive test:

[0066] Based on the AI ​​model prediction in step 2, the single drug test is integrated with the 3x3 combined drug concentration test. In this step, the selected drug is not only tested at a single concentration point (C1, C2, ..., C9), but also tested for its combined effect with other drugs;

[0067] (31) Experimental settings:

[0068] The selected drugs were tested at 9 different concentration points, which were also used for combination testing with other drugs. Each concentration point was selected based on the expected range of drug activity and previous experimental data to ensure coverage from ineffective doses to concentrations exceeding the expected therapeutic window.

[0069] (32) Formula description and variable explanation:

[0070] The dose-response curves for single-drug and combination therapy were described using the following model:

[0071] Formula: R_i=100 / (1+(EC50 / C_i)^n)

[0072] in:

[0073] R_i: cell survival rate at concentration C_i, expressed as a percentage;

[0074] C_i: the i-th concentration point, i=1,2,...,9;

[0075] EC50: the concentration of a drug required to produce 50% of the maximal effect;

[0076] n: hill shape parameter, describing the slope of the curve;

[0077] (33) Determination method:

[0078] Cell viability was determined using the CTG Cell Viability / Cytotoxicity Assay Kit:

[0079] Cell inoculation: Inoculate a certain number of target cells in a 96-well plate;

[0080] Drug treatment: different concentrations of single drug and its combination solutions with other drugs were added to the corresponding wells;

[0081] Cultivation: The treated cell plates are cultured under appropriate conditions for a certain period of time to allow the drug to act;

[0082] Add CTG reagent: After the incubation period, add CTG reagent to each well. This reagent is reduced in living cells to produce a fluorescent or colorimetric signal.

[0083] Signal detection: Use a microplate reader to read the signal in each well. The signal intensity is proportional to the cell survival rate.

[0084] Data processing: Collect readings at all concentration points, including data for single drug and combination drug, and calculate cell viability relative to untreated controls

[0085] This integrated testing approach allows the evaluation of single-drug and combination drug effects in the same experimental setting, which not only improves experimental efficiency but also enables drug effect analysis based on more comprehensive data.

[0086] At the same time as the test in step (33), the 3x3 matrix test of combined drug use is performed in combination with the prediction of the AI ​​model, and the data of the single drug is also obtained in this process, rather than performing a single drug concentration point test alone;

[0087] (34) Experimental matrix setting:

[0088] In this step, three concentration points predicted by the AI ​​model are set for each drug to form a 3x3 matrix. These concentration points are C1a, C1b, C1c for drug 1, and C2a, C2b, C2c for drug 2. Each point is used for both single-drug and combination drug testing;

[0089] (35)Matrix description:

[0090] Each matrix element R_ij represents the cell survival rate under the concentration Ci of drug 1 and the concentration Cj of drug 2, where i and j are indexes from 1 to 3, representing different concentration levels. This design allows the simultaneous evaluation of single drug effects and drug interactions;

[0091] (36) Determination method:

[0092] Cell seeding: Plant target cells in a 96-well plate to ensure that the number of cells in each well is consistent;

[0093] Drug treatment: According to the setting of 3x3 matrix, add corresponding drug concentration combination to each well, including single drug and combination drug;

[0094] Incubation: Incubate the treated cell plates under appropriate conditions, allowing the drug to act for an appropriate amount of time;

[0095] Add CTG reagent: After the incubation, add CTG reagent to each well. The reagent is reduced in living cells to produce a quantitative signal.

[0096] Signal reading: Use a microplate reader to read the signal intensity of each well, which is proportional to the cell viability;

[0097] Data processing: Calculate the cell survival rate under each drug combination and compare it with the untreated control group, and record the effect of a single drug;

[0098] (37)Formula and data analysis:

[0099] Formula: R_ij=(Signal_ij / Signal_control)*100

[0100] R_ij: percentage of cell survival under drug combination (i, j);

[0101] Signal_ij: signal intensity measured under drug combination (i, j);

[0102] Signal_control: Signal intensity of untreated cells, used to normalize the results.

[0103] Through this comprehensive testing method, not only can the effect of a single drug be evaluated, but the synergistic, additive or antagonistic effects of different drug combinations can also be analyzed in detail, thereby providing more comprehensive data support for future medication strategies; this method improves the efficiency of the experiment while reducing the number of experiments required, effectively reducing costs and complexity.

[0104] Furthermore, in the above-mentioned method for efficiently evaluating the efficacy of in vitro combined drug therapy, step 4 comprises the following specific methods:

[0105] (41) Single drug dose response curve fitting

[0106] formula:

[0107] R_single=R_max / (1+(EC50 / C)^HillSlope);

[0108] R_single: cell survival rate at concentration C;

[0109] R_max: maximum possible cell survival rate, usually close to 100%;

[0110] EC50: drug concentration required to produce 50% of the maximal effect;

[0111] C: drug concentration;

[0112] HillSlope: Hill coefficient, which describes the slope of the curve and reflects the sensitivity of the drug effect to concentration changes;

[0113] Fitting steps:

[0114] Data preparation: Collect data from 9 concentration points from step (4): C1 to C9, and the corresponding cell viability: R1 to R9;

[0115] Select Model: Select Hill equation as the fitting model;

[0116] Parameter estimation: Use nonlinear least squares method to estimate the values ​​of R_max, EC50 and HillSlope and fit the experimental data;

[0117] Model validation: Verify the quality of model fit by calculating the residuals between the predicted and actual values;

[0118] (42) Analysis of the effect of combined medication

[0119] After obtaining the dose-response curves of the single drugs, use these curves to predict the theoretical effects of the single drugs in combination testing to assess whether the interaction between the two drugs is synergistic, additive, or antagonistic;

[0120] Calculation steps:

[0121] Theoretical single drug effect prediction: For each concentration combination of the combined drug, the expected individual effects of the two drugs are calculated using the fitted dose-response curves;

[0122] Comparison of actual combined effects: Comparison of theoretical single-drug effects with the combined drug effects actually measured by the CTG method;

[0123] Synergy was assessed using the following formula:

[0124] SynergyScore=ObservedEffect-(PredictedEffect1+PredictedEffect2)

[0125] ObservedEffect: Observed effect of combined medication;

[0126] PredictedEffect1, PredictedEffect2: the effects of two drugs predicted based on the single-drug dose-response curves;

[0127] The above formula helps determine whether the drugs, when used in combination, have more than the simple addition of their individual effects, thus determining whether a synergistic effect exists.

[0128] Furthermore, in the above-mentioned method for efficiently evaluating the efficacy of in vitro combined medication, step 5 comprises the following specific methods:

[0129] (51) Data evaluation

[0130] Evaluation Metrics:

[0131] EC50 value: a measure of drug potency, referring to the concentration required to produce 50% of the maximum biological effect;

[0132] Therapeutic index (TI): Therapeutic index is an indicator of drug safety and is calculated as follows: TI = TD50 / ED50;

[0133] TD50: the dose that causes toxic reactions in 50% of the subjects;

[0134] ED50: effective dose that produces 50% of the desired effect;

[0135] Synergy Index (CI): It is used to evaluate the synergistic effect of two drugs when used in combination. The calculation formula is:

[0136] CI = (D1 / Dx1) + (D2 / Dx2);

[0137] D1 and D2: actual doses of the two drugs when used in combination;

[0138] Dx1 and Dx2: doses of the two drugs that produce equivalent effects when used alone;

[0139] A CI value less than 1 indicates synergism, equal to 1 indicates additive effect, and greater than 1 indicates antagonism;

[0140] (52) Optimize experimental design

[0141] Based on the results of the data evaluation in (52), the experimental design was adjusted to optimize the drug concentration range and combination of drugs;

[0142] Adjustment steps:

[0143] Concentration range adjustment: Redefine the concentration range of the drug based on EC50 values ​​and safety data, including but not limited to lowering the maximum concentration to avoid toxicity, or adjusting the minimum concentration to ensure efficacy;

[0144] Optimization of drug combinations: Based on the results of the synergy index analysis, select those drug combinations that show synergistic effects and redesign the experiments to verify these findings;

[0145] Repeat the experiment: Perform a new round of testing with adjusted drug concentrations and combinations to confirm the improved effects of these adjustments.

[0146] Furthermore, the present invention also discloses the use of the above method in drug preparation or research and development.

[0147] Furthermore, the present invention also discloses a storage chip storing a system or program capable of running the above method.

[0148] Furthermore, the present invention also discloses a device comprising the above-mentioned storage chip.

[0149] Compared with the prior art, the present invention has the following beneficial effects:

[0150] The present invention significantly improves the efficiency and accuracy of in vitro combined drug efficacy evaluation by introducing AI technology and optimizing experimental design. The theoretical basis and experimental data support of the specific advantages are as follows:

[0151] 1. Improve efficiency and reduce costs:

[0152] Simplified experimental mode: The traditional 6x6 or 9x9 matrix mode requires a large number of drug concentration combination tests, which not only consumes a lot of reagents and time, but also increases the complexity of data processing. The present invention adopts a 3x3 matrix mode, which significantly reduces the number of drug concentration combinations required in the experiment. This simplification can directly reduce reagent consumption, reduce the required experimental time and human resource investment.

[0153] AI optimizes concentration selection: Using AI technology to predict and optimize concentration combinations based on historical data can effectively avoid unnecessary concentration tests. This not only reduces the number of experiments, but also ensures the quality of the experiments and avoids the waste of resources caused by redundant concentrations.

[0154] 2. Improve sensitivity and reliability:

[0155] Data-driven concentration optimization: By analyzing a large amount of historical data, the AI ​​model can accurately predict the drug concentration combination that is most likely to show synergistic effects. This precise prediction not only improves the sensitivity of the experiment, but also more accurately evaluates the actual effect of the drug combination.

[0156] Improved data analysis methods: Traditional evaluation methods usually rely on intuitive comparisons or simple statistical analysis, which may not accurately identify small differences in effect. The present invention uses AI technology to perform data analysis, which can process and interpret experimental data more carefully, thereby improving the reliability and accuracy of the results.

[0157] 3. Easy operation and optimized control:

[0158] Standardized experimental process: The standardized process of 3x3 matrix experimental design not only simplifies the operation, but also makes the experimental results more reproducible. The simplified operation process reduces the technical requirements of the operator and makes the experiment easier to popularize and execute.

[0159] AI-assisted experimental design and data analysis: The introduction of AI technology not only provides decision support at the front end of the experiment, but also provides intelligent assistance in the data analysis stage, reducing the possibility of human errors and improving the accuracy of the entire experimental design and execution.

[0160] Through these improvements, the present invention not only significantly improves the economic benefits and scientificity of drug evaluation, but also indirectly reduces environmental pollution by reducing the use of chemical reagents and optimizing experimental processes. In addition, more accurate drug effect evaluation can also help to exclude drug combinations with poor effects or high toxicity in the preclinical stage, thereby reducing the potential risk of side effects in subsequent clinical trials. These comprehensive advantages make the present invention have important application value in the modern drug research and development process. BRIEF DESCRIPTION OF THE DRAWINGS

[0161] Figure 1 The overall method steps of the present invention;

[0162] Figure 2 Dose-response curve of drug A;

[0163] Figure 3 Dose-response curve of drug B;

[0164] Figure 4 Combination drug inhibition rate matrix table and heat map;

[0165] Figure 5 Bliss score matrix for combined medication;

[0166] Figure 6 Heat map and 3D display of Bliss score for combined medication;

[0167] Figure 7 Loewe score matrix for combined medication;

[0168] Figure 8 Heat map and 3D display of Loewe scores for combined medication. DETAILED DESCRIPTION

[0169] The present invention relates to a method optimized by artificial intelligence (AI) to evaluate the efficacy of in vitro combined drug therapy. The method is particularly suitable for the development of anticancer drugs, which can significantly improve experimental efficiency and reduce experimental costs, while enhancing the accuracy and reliability of drug evaluation. First, by developing an AI model based on historical data, the model can process a variety of input data including the SMILES sequence, concentration and specific cell type of the drug, and predict the cell survival rate of a single drug or a drug combination at different concentrations. The development of this model is the basis of the entire evaluation process and provides scientific prediction support for subsequent steps. Using this model, combined drug concentration prediction is performed to screen out the drug combination that is most likely to show synergy. Next, through single drug concentration point testing, the prediction of the AI ​​model is experimentally verified and the dose-response curve of each drug is established. Based on these data, a 3x3 matrix test of combined drug therapy is further implemented. In this step, the cell survival rate of different drug combinations is tested to evaluate the potential synergistic effect. Dose-response curve analysis is a stage for in-depth analysis of the aforementioned test data, through which the synergistic, additive or antagonistic effects of drug combinations can be accurately evaluated and verified. Finally, the experiment is optimized and adjusted based on the analysis results to optimize the drug dosage and combination strategy and improve the efficacy and safety of the drug. The entire method combines modern AI technology with traditional pharmacological testing, which not only improves the efficiency and effectiveness of the research, but also provides a new and efficient tool for the development of anti-cancer drugs. This method is expected to have an important impact on the field of drug research and development, especially in accelerating drug launch and reducing research and development costs.

[0170] The overall method steps of the present invention are as follows Figure 1 As shown, the following steps are included:

[0171] Step 1: AI model development

[0172] This step involves building an AI model using historical data sets, which is specifically designed to predict cell survival rates of drugs at different concentrations and specific cell types. This model combines the SMILES sequence of the drug, concentration information, and cell type, and is trained and validated through advanced machine learning techniques such as deep learning to ensure the accuracy of the prediction. The development of this model is fundamental and critical, and it directly affects the scientificity and effectiveness of drug combination and concentration selection in subsequent steps.

[0173] Step 2: Prediction of combined drug concentration

[0174] After the AI ​​model is successfully developed, the next step is to use the model to predict the combined effects of different drug combinations at specific concentrations. In this process, the model will evaluate possible drug combinations and predict which combinations may show synergistic effects biologically. This step is the key to experimental design. It pre-screens the most promising drug combinations through scientific methods, laying the foundation for the efficiency and cost-effectiveness of the experimental stage.

[0175] Step 3: Comprehensive test of single drug concentration point test and 3x3 combination test: Directly use the AI ​​model prediction results of step 2 to conduct experiments on single drug and 3x3 matrix combination. Test the cell survival rate of a single drug at different concentration points, and conduct a 3x3 matrix combination test at the same time. The three recommended concentrations of each drug form a matrix intersection to test the cell response under different combinations.

[0176] Step 4: Dose-response curve analysis

[0177] After obtaining data from single-drug and combination drug tests, dose-response curve analysis is performed. This includes the use of advanced data analysis techniques, such as nonlinear regression analysis, to fit dose-response curves and use these curves to predict and verify the synergistic effects of combination drugs. Through the detailed analysis of this step, it can be determined whether there is a synergistic, additive or antagonistic effect between drugs, providing a scientific basis for subsequent drug strategies.

[0178] Step 5: Experimental optimization and adjustment

[0179] The final step is to optimize and adjust the experimental design based on the results of the previous steps. This may include adjusting the drug concentration range, reselecting drug combinations, or modifying experimental conditions to improve experimental effects and precision. This step ensures flexibility and adaptability throughout the research process, making the final drug strategy both scientifically effective and cost-effective. Through continuous data evaluation and adjustment, the research team can refine drug dosages and combinations to maximize therapeutic effects while minimizing potential side effects.

[0180] These five steps form a complete drug evaluation and optimization process. Each step provides support for the subsequent steps, and together they build a closed-loop system from theoretical prediction to experimental verification to experimental optimization. The relationship between these steps is explained in detail below:

[0181] The first step, AI model development, is the foundation of the entire drug evaluation process. In this step, the model trained with historical data is able to predict the effects of a given drug and cell type at a specific concentration. This model is directly used in the second step, combination drug concentration prediction, where the AI ​​model evaluates the potential synergistic effects of different drug combinations at specific concentrations. Therefore, the accuracy of the AI ​​model directly affects the effectiveness of combination drug predictions and the direction of experimental design. After the possible effective drug combinations and concentrations are screened by the AI ​​model, single drug concentration point tests are used to verify and refine these predictions. This step is necessary because it provides baseline data for combination drug testing, that is, the biological effects of each drug when acting independently. These data are the basis for analyzing drug interactions and synergistic effects. Single drug concentration point tests provide preliminary data on drug dose responses, which are extremely important in the 3x3 matrix test of combination drugs. This matrix test combines the data of these single drugs with other drug combinations to evaluate drug synergy, additiveness or antagonism under different combinations. This step verifies the application and effect of single drug data in actual multi-drug combinations. The data obtained through the 3x3 matrix test need to be analyzed in detail to understand the specific effects of different drug combinations. The dose-response curve analysis step uses this data to fit curves, predict and verify drug interactions. This analysis helps determine the optimal concentrations of drug combinations and potential biological synergies. Dose-response curve analysis provides in-depth insights into how different drug combinations and concentrations affect therapeutic effects. These insights are the basis for experimental optimization and adjustment, allowing researchers to adjust drug doses or modify combination strategies based on scientific data, optimize experimental design, and improve drug efficacy and safety.

[0182] Here are detailed instructions for each step:

[0183] Step 1: AI model development

[0184] In drug development, it is important to be able to predict the effects of different drug combinations and concentrations on specific cell types, especially when evaluating the effects of in vitro drug combinations. This step details how to use artificial intelligence technology to build a predictive model using historical data that can accurately predict the effects of drugs on cell survival.

[0185] First, we collected and organized the following data:

[0186] S (SMILES sequence): This is a method of representing the structure of a chemical substance using an ASCII string. SMILES sequences contain rich information about the molecular structure and can be converted into molecular descriptors or directly used as input to deep learning models.

[0187] C (drug concentration): This refers to the concentration of the drug used in the experiment, usually measured in micromoles (μM). Drug concentration is one of the key variables that determine drug efficacy.

[0188] T (cell type): Different cell types can respond to drugs in significantly different ways, so cell type is an important component of model input.

[0189] R (CTG readings, cell viability): This is the result obtained by the cell viability test and serves as the prediction target of the model.

[0190] Using these data, we construct a function f(S,C,T)->R, which aims to predict the cell survival rate R under a specific SMILES sequence S, drug concentration C, and cell type T. The specific construction process of the model is as follows:

[0191] Feature Engineering:

[0192] SMILES sequence processing: Convert SMILES sequences into numerical molecular descriptors or directly process sequence data through neural networks (such as convolutional neural networks).

[0193] Drug concentration processing: Normalize the concentration values ​​to fit the input range of the model.

[0194] Cell type encoding: Convert the cell type into a processable form, such as using One-Hot Encoding.

[0195] Model selection and training:

[0196] Choose a machine learning model that is suitable for processing this type of data, such as random forest, support vector machine, or deep learning model such as multi-layer perceptron (MLP) or convolutional neural network (CNN).

[0197] Use historical data sets for model training. During the training process, optimize the model parameters to minimize the prediction error. The commonly used loss function is the mean square error (MSE): E = (1 / N) * Σ (R_i-R'_i)^2, where R_i is the actual cell survival rate, R'_i is the cell survival rate predicted by the model, and N is the number of samples.

[0198] Model Evaluation:

[0199] The predictive performance of the model is evaluated through techniques such as cross-validation to ensure that the model has good generalization ability.

[0200] Model Application:

[0201] The trained model is used to predict the effect of new drug combinations. This step is to apply the model f to the actual problem and predict R based on the given S, C, and T.

[0202] We use the model f that we have developed to predict cell survival for a single drug at a specific concentration and cell type. Now, we extend this model to handle the combined effects of two drugs. Suppose we have two drugs S1 and S2 at concentrations C1 and C2, respectively, and we need to predict the cell survival rate R_comb when they are combined in cell type T.

[0203] Formula description:

[0204] The prediction model of the effect of combined drug therapy can be expressed as: R_comb = f(S1, C1, T) + f(S2, C2, T) + Interaction(S1, S2, C1, C2, T).

[0205] in:

[0206] f(S,C,T) represents the predicted cell survival rate of a single drug at a specific concentration and cell type.

[0207] Interaction(S1,S2,C1,C2,T) is a complementary term used to estimate the possible interaction between two drugs, which can be a linear or nonlinear function.

[0208] Variable explanation:

[0209] S1, S2: SMILES sequences of two drugs.

[0210] C1, C2: are the concentrations of the two drugs respectively.

[0211] T: cell types used in the experiments.

[0212] R_comb: Cell survival rate when two drugs are used in combination in a specific cell type

[0213] Step 2: Prediction of combined drug concentration

[0214] Understanding the combined effects of two or more drugs at specific concentrations is critical during drug development and efficacy evaluation. This step describes in detail how to use an artificial intelligence model to predict cell survival when two drugs are used in combination and select the drug concentration combination that is most likely to show synergistic effects.

[0215] Screening mechanism

[0216] In order to screen out the combinations that show the strongest synergy from possible drug combinations and concentrations, we define a criterion for quantifying synergy, usually using a synergy index (Synergy Score). This score can be calculated based on Bliss independence prediction or other relevant drug interaction models.

[0217] Formula description:

[0218] Synergy Score=R_comb-(R_pred(S1,C1,T)+R_pred(S2,C2,T))

[0219] Among them, R_pred(S1,C1,T) and R_pred(S2,C2,T) are the predicted effects of single drugs, and R_comb is the actual predicted effect of combined drug use.

[0220] Calculation steps:

[0221] For each pair of drugs and each possible concentration combination (C1, C2), R_comb was calculated using the above model.

[0222] Calculate the Synergy Score for each pair of combinations.

[0223] Three concentration combinations with the highest Synergy Scores were selected, and these combinations were expected to show the strongest synergistic effects.

[0224] With this approach, we can not only predict the combined effects of two drugs, but also determine the combination of drug concentrations that may be most effective in experimental and clinical applications. This step is crucial because it can significantly improve R&D efficiency and reduce the number of trials for ineffective or suboptimal drug combinations;

[0225] Step 3: Single drug concentration point test and 3x3 combined comprehensive test:

[0226] Based on the AI ​​model prediction in step 2, the single drug test and the combined drug concentration test are integrated; in this step, the selected drug is not only tested at a single concentration point (C1, C2, ..., C9), but also tested for its combined effect with other drugs;

[0227] Experimental setup:

[0228] The selected drugs were tested at 9 different concentration points. At the same time, these concentration points were also used for combination testing with other drugs. The selection of each concentration point was determined based on the expected drug activity range and previous experimental data to ensure coverage from ineffective doses to concentrations exceeding the expected therapeutic window.

[0229] Formula description and variable explanation:

[0230] The dose-response curves for single-drug and combination therapy were described using the following model:

[0231] Formula: R_i=100 / (1+(EC50 / C_i)^n)

[0232] in:

[0233] R_i: cell survival rate at concentration C_i, expressed as a percentage;

[0234] C_i: the i-th concentration point, i=1,2,...,9;

[0235] EC50: the concentration of a drug required to produce 50% of the maximal effect;

[0236] n: hill shape parameter, describing the slope of the curve;

[0237] Determination method:

[0238] Cell viability was determined using the CTG Cell Viability / Cytotoxicity Assay Kit:

[0239] Cell inoculation: Inoculate a certain number of target cells in a 96-well plate;

[0240] Drug treatment: different concentrations of single drug and its combination solutions with other drugs were added to the corresponding wells;

[0241] Cultivation: The treated cell plates are cultured under appropriate conditions for a certain period of time to allow the drug to act;

[0242] Add CTG reagent: After the incubation period, add CTG reagent to each well. This reagent is reduced in living cells to produce a fluorescent or colorimetric signal.

[0243] Signal detection: Use a microplate reader to read the signal in each well. The signal intensity is proportional to the cell survival rate.

[0244] Data processing: Collect readings at all concentration points, including data for single-drug and combination-drug treatments, and calculate cell viability relative to untreated controls.

[0245] In addition to the above tests, the 3x3 matrix test of combined medication was conducted in combination with the prediction of the AI ​​model. In the process, the data of single drugs were also obtained, rather than testing single drug concentration points alone.

[0246] Experimental matrix settings:

[0247] In this step, three concentration points predicted by the AI ​​model are set for each drug to form a 3x3 matrix; these concentration points are C1a, C1b, C1c for drug 1, and C2a, C2b, C2c for drug 2. Each point is used for both single-drug and combination drug testing.

[0248] Matrix description:

[0249] Each matrix element R_ij represents the cell survival rate under the concentration Ci of drug 1 and the concentration Cj of drug 2, where i and j are indexes from 1 to 3, representing different concentration levels. This design allows the simultaneous evaluation of single drug effects and drug interactions;

[0250] Determination method:

[0251] Cell seeding: Plant target cells in a 96-well plate to ensure that the number of cells in each well is consistent;

[0252] Drug treatment: According to the setting of 3x3 matrix, add corresponding drug concentration combination to each well, including single drug and combination drug;

[0253] Incubation: Incubate the treated cell plates under appropriate conditions, allowing the drug to act for an appropriate amount of time;

[0254] Add CTG reagent: After the incubation, add CTG reagent to each well. The reagent is reduced in living cells to produce a quantitative signal.

[0255] Signal reading: Use a microplate reader to read the signal intensity of each well, which is proportional to the cell viability;

[0256] Data processing: Calculate the cell survival rate under each drug combination and compare it with the untreated control group, and record the effect of a single drug;

[0257] Formula and data analysis:

[0258] Formula: R_ij=(Signal_ij / Signal_control)*100;

[0259] R_ij: percentage of cell survival under drug combination (i, j);

[0260] Signal_ij: signal intensity measured under drug combination (i, j);

[0261] Signal_control: Signal intensity of untreated cells, used to normalize the results.

[0262] Through this comprehensive testing method, not only can the effect of a single drug be evaluated, but the synergistic, additive or antagonistic effects of different drug combinations can also be analyzed in detail, thereby providing more comprehensive data support for future medication strategies; this method improves the efficiency of the experiment while reducing the number of experiments required, effectively reducing costs and complexity.

[0263] Step 4: Dose-response curve analysis

[0264] Dose-response curve analysis is a key step in drug evaluation. It provides a basis for understanding the drug's mechanism of action by quantitatively describing the relationship between drug concentration and biological effect. This step details how to use artificial intelligence technology to fit dose-response curves and apply these curves to evaluate the effects of combined drug use.

[0265] Single drug dose response curve fitting

[0266] The single drug dose response curve describes the relationship between drug concentration and cell survival rate. This relationship can usually be modeled by the Hill equation, which is a common model in pharmacology to describe the change of drug effect with concentration.

[0267] formula:

[0268] R_single=R_max / (1+(EC50 / C)^HillSlope)

[0269] R_single: Cell survival rate at concentration C.

[0270] R_max: Maximum possible cell viability, usually close to 100%.

[0271] EC50: The concentration of drug required to produce 50% of the maximal effect.

[0272] C: drug concentration.

[0273] HillSlope: Hill coefficient, which describes the slope of the curve and reflects the sensitivity of drug effects to changes in concentration.

[0274] Fitting steps:

[0275] Data preparation: The 9 concentration point data (C1 to C9) and the corresponding cell viability (R1 to R9) collected from step 2.1.

[0276] Select Model: Select the Hill equation as the fitting model.

[0277] Parameter estimation: The values ​​of R_max, EC50 and HillSlope were estimated using nonlinear least squares method to best fit the experimental data.

[0278] Model validation: Verify the quality of the model fit by calculating the residuals between the predicted and actual values.

[0279] Analysis of the effect of combined medication:

[0280] After obtaining the dose-response curves for a single drug, we can use these curves to predict the theoretical effects of a single drug in a combination test to assess whether the interaction between the two drugs is synergistic, additive, or antagonistic.

[0281] Calculation steps:

[0282] Prediction of theoretical single-drug effects: For each concentration combination of the combined drug (such as C1a, C2a in step 2.2), the expected individual effects of the two drugs were calculated using the fitted dose-response curves.

[0283] Comparison of actual combined effects: Compare the theoretical single-drug effects with the combined drug effects actually measured by the CTG method.

[0284] Synergy Assessment:

[0285] SynergyScore=ObservedEffect-(PredictedEffect1+PredictedEffect2)

[0286] ObservedEffect: The observed effect of combined medication.

[0287] PredictedEffect1, PredictedEffect2: The effects of two drugs predicted based on the single-drug dose-response curves.

[0288] This formula helps determine if the drugs, when used in combination, have more than the simple addition of their individual effects, thus determining if there is a synergistic effect.

[0289] Through this method, we can not only understand the efficacy of a single drug in detail, but also evaluate the potential advantages of drug combinations, providing a scientific basis for clinical application and further drug development. This analysis method has important guiding significance for optimizing drug formulation and dosage design.

[0290] Step 5: Experimental optimization and adjustment

[0291] Experimental optimization and adjustment is an important part of the drug evaluation process. The purpose is to adjust and optimize the experimental design based on preliminary experimental data to ensure more accurate and scientific results. This step involves in-depth analysis of the data, evaluating the effects of single drugs and combination drugs, and adjusting drug concentrations and test parameters accordingly to optimize the combination drug strategy.

[0292] Data evaluation

[0293] Data evaluation is the first step in this process and includes analysis of all preliminary experimental results, including single-drug dose-response curves and the effects of combination therapy. The purpose of this analysis is to determine the drug's effectiveness, safety margin, and possible synergistic effects.

[0294] Evaluation Metrics:

[0295] EC50 value: A commonly used indicator to measure drug efficacy, which refers to the concentration required to produce 50% of the maximum biological effect.

[0296] Therapeutic index (TI): Therapeutic index is an important indicator of drug safety, and the calculation formula is:

[0297] TI=TD50 / ED50

[0298] TD50: The dose that causes toxic reactions in 50% of the subjects.

[0299] ED50: The effective dose that produces 50% of the desired effect.

[0300] Synergy Index (CI): It is used to evaluate the synergistic effect of two drugs when used in combination. The calculation formula is:

[0301] CI=(D1 / Dx1)+(D2 / Dx2)

[0302] D1 and D2: The actual doses of the two drugs when used in combination.

[0303] Dx1 and Dx2: Doses of two drugs that produce equivalent effects when used alone.

[0304] A CI value less than 1 indicates synergism, a CI value equal to 1 indicates additive effect, and a CI value greater than 1 indicates antagonism.

[0305] Optimizing experimental design

[0306] Based on the results of the above data evaluation, the experimental design was adjusted, mainly involving the optimization of the drug concentration range and the combination of combined drugs.

[0307] Adjustment steps:

[0308] Concentration range adjustment: Redefine the concentration range of the drug based on the EC50 value and safety data. This may include lowering the highest concentration to avoid toxicity or adjusting the lowest concentration to ensure efficacy.

[0309] Optimization of drug combination therapy: Based on the results of the synergy index analysis, select those drug combinations that show synergistic effects and redesign the experiments to verify these findings.

[0310] Repeat the experiment: Perform a new round of testing with adjusted drug concentrations and combinations to confirm the improved effects of these adjustments.

[0311] Through the systematic evaluation of preliminary experimental data and the scientific adjustments made accordingly, the drug dosage design and combination drug strategy can be effectively optimized, and the success rate of drug development and the safety of clinical application can be improved. This process not only reduces the risk of ineffective or harmful experiments, but also helps to accurately reveal the mechanism of action and interaction of drugs by finely adjusting the experimental conditions, providing a solid foundation for subsequent clinical trials and drug marketing.

[0312] It should be noted that the following embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention 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 invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0313] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0314] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0315] Example

[0316] To evaluate the efficacy of a certain ataxia telangiectasia mutation inhibitor (hereinafter referred to as "drug A") and a certain ataxia telangiectasia and Rad3-related inhibitor (hereinafter referred to as "drug B") in combination in RD-ES cells.

[0317] The research implementation process is as follows:

[0318] 1. AI model construction method

[0319] Data preparation: Collect historical experimental data, including the response data of various cell lines to single drugs and combination drugs, drug concentration information, SMILES sequences, and cell survival rate measured by CTG method.

[0320] Feature extraction: Chemical descriptors were extracted from SMILES sequences using RDKit, combined with drug concentration and cell type as input features for the model.

[0321] 2. Model structure: Construct a multi-layer perceptron (MLP) network, which includes an input layer, several hidden layers and an output layer. The hidden layer uses the ReLU activation function and the output layer uses the Sigmoid activation function to predict cell survival rate.

[0322] 3. Model training process

[0323] Split the data: Divide the dataset into training set (80%) and testing set (20%).

[0324] Training parameters: Set the learning rate to 0.001, the batch size to 32, and the training epochs to 100.

[0325] Training Execution: The training is performed using the Adam optimizer, with the goal of minimizing the mean squared error between the actual survival rate and the predicted survival rate.

[0326] 4. Analysis of the combined use of drug "A" and drug "B"

[0327] 1) Prediction of combined drug concentration

[0328] The SMILE sequences of drug "A" and drug "B" were input into the AI ​​model to predict the optimal concentration of the 3X3 combination drug in Table 1 below:

[0329] Table 1 Concentration of combined medication

[0330] Concentration 1 (micromolar) Concentration 2 (micromolar) Concentration 3 (micromolar) Drug A 0.5 0.75 1 Drug B 0.04 0.1 0.25

[0331] 2) 3x3 matrix experiment

[0332] The RD-ES cell stock was taken out of liquid nitrogen, revived and cultured in RPMI1640 medium containing 10% fetal bovine serum. When the cells grew to a certain number, the cells were digested with trypsin, diluted and mixed with culture medium, and the number of cells was recorded by Count-star. The concentration of the cell suspension was adjusted.

[0333] RD-ES cell suspension was inoculated into 96-well plates at a ratio of 6300 cells / well. 150 μl of cell suspension was added to each well, and an equal volume of PBS was added to the edge to avoid edge effect. After the plate was inoculated, it was placed back to 37°C and 5% CO. 2 Incubate in the incubator for 24 hours for later use. After 24 hours, dilute each drug according to Table 2, add different concentrations of drug solution to each well, and make the concentration in the well reach the working concentration. After adding the drug, put the well plate back into the incubator to allow the drug to act for 120 hours.

[0334] Table 2 Combination medication plan of drug A and drug B

[0335]

[0336] After 5 days of drug treatment, the drug-doped plate was removed and equilibrated at room temperature for 30 minutes. 50 μl of CellTiter-Glo reagent was added to each well and the contents were mixed on an orbital shaker for 5 minutes to cause cell lysis. The operation was protected from light. The plate was then incubated at room temperature for 20 minutes to stabilize the luminescent signal. The readings were recorded using an EnVision multi-label reader. The inhibition rate of each well of the drug-doped plate was calculated according to the formula: Inhibition rate (%) = 1-cell viability (%) [(drug-doped reading-blank control reading) / (DMSO control reading-blank control reading)], and the dose-response curve of a single drug and the drug concentration-inhibition rate matrix were plotted. The results are shown in Figure 2. Figure 2-4 The results were analyzed by the AI ​​model using the Bliss and Loewe models, and the results were as follows: Figure 5-8 The Bliss index matrix and 3D graph and the Loewe index matrix and 3D graph are shown.

[0337] Synergy scores were calculated using the Bliss independence model and the Loewe additivity model. Scores above 5 indicate synergy, and scores below -5 indicate antagonism.

[0338] Note: The Bliss independence model is appropriate for non-interacting drugs that produce responses independently through different pathways. For example, by targeting different pathways. In contrast, the Loewe additivity model is more applicable when two drugs have similar modes of action on the same target or pathway. As pointed out in the Saariselka protocol (Greco et al., 1992), and by many others, neither Loewe additivity nor Bliss independence necessarily reflect the expected mode of action of a drug combination. Instead, the Loewe and Bliss models should be used as data exploration methods with the primary goal of identifying potential synergistic drug combinations for further mechanistic studies, rather than the other way around, i.e., using mechanistic evidence to determine which reference model is more appropriate.

[0339] It can be seen from the above embodiments that the present invention significantly improves the efficiency and accuracy of in vitro combined drug efficacy evaluation by introducing AI technology and optimizing experimental design.

[0340] The above are only a few preferred embodiments of the present invention, and the description is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A method for efficiently evaluating the efficacy of in vitro combined drug therapy, characterized in that: The following steps are involved: Step 1. AI model development: Use historical drug use datasets to build an AI model. The AI ​​model combines the drug's SMILES sequence, concentration information, and cell type, and is trained and validated through machine learning techniques to predict cell survival rates at different drug concentrations and cell types. Step 2. Prediction of combined drug concentration: Use the AI ​​model obtained in step 1 to predict the combined effects of different drug combinations at specific concentrations and give the recommended 3x3 combined drug concentrations; Step 3. Comprehensive test of single drug concentration point test and 3x3 combination test: directly use the AI ​​model prediction results of step 2 to conduct experiments on single drug and 3x3 matrix combination, test the cell survival rate of a single drug at different concentration points, and simultaneously conduct 3x3 matrix combination test. The three recommended concentrations of each drug form matrix intersections to test the cell response under different combinations; Step 4. Dose-response curve analysis: After obtaining data from the single-drug and combination-drug tests in step 3, perform dose-response curve analysis to predict and verify the synergistic effect of combination therapy; determine whether there is synergistic, additive or antagonistic effect between drugs, and provide a basis for subsequent medication strategies; Step 5. Experimental optimization and adjustment: Based on the results of steps 1-4, optimize and adjust the experimental design, including adjusting the drug concentration range, reselecting drug combinations, or modifying other experimental conditions; The step 1 includes the following specific methods: (11) Collect and organize the following data: S: SMILES sequence of the drug represented by an ASCII string: C: drug concentration; T: cell type; R: CTG readings, cell viability: (12) Using the above data, a model f(S,C,T)->R was constructed to predict the cell survival rate R under a specific SMILES sequence S, drug concentration C, and cell type T; The specific construction process of the above model f(S,C,T)->R is as follows: Feature Engineering: SMILES sequence processing: convert SMILES sequences into numerical molecular descriptors or directly process sequence data through neural networks; Drug concentration processing: Standardize the drug concentration C value to fit the input range of the model; Cell type encoding: converting cell type T into a processable form; Model selection and training: Choose a machine learning model that is suitable for processing this type of data: including random forests, support vector machines, or deep learning models; Use historical data sets for model training: During the training process, optimize the model parameters to minimize the prediction error, and use the mean square error (MSE) as the loss function: E = (1 / N) * Σ (R_i-R'_i)^2, Where R_i is the actual cell survival rate, R'_i is the cell survival rate predicted by the model, and N is the number of samples; Model Evaluation: The predictive performance of the model was evaluated by cross-validation techniques; Model Application: Use the trained AI model to predict the effect of new drug combinations, that is, apply the model f to the actual problem and predict R based on the given S, C, and T; (13) Model application Use the developed model f to handle the combined effects of two drugs; drugs S1 and S2, at concentrations C1 and C2 respectively, predict the cell survival rate R_comb when they are combined in cell type T; R_comb formula description: The prediction model of the effect of combined medication is expressed as: R_comb = f(S1, C1, T) + f(S2, C2, T) + Interaction(S1, S2, C1, C2, T) in: f(S,C,T) represents the predicted cell survival rate of a single drug at a specific concentration and cell type; Interaction (S1, S2, C1, C2, T) is a supplementary term used to estimate the possible interaction between two drugs and is a linear or nonlinear function; Variable explanation: S1, S2: SMILES sequences of the two drugs; C1, C2: are the concentrations of the two drugs respectively T: cell types used in the experiments; R_comb: Cell survival rate when two drugs are used in combination in a specific cell type.

2. A method for efficiently evaluating the efficacy of in vitro combined drug therapy according to claim 1, characterized in that: The step 2 includes the following specific methods: (21) Screening mechanism In order to screen out the combination showing the strongest synergistic effect from possible drug combinations and concentrations, a standard for quantifying synergy is defined. The above standard is the synergy index Synergy Score: Synergy Score formula description: Synergy Score=R_comb-(R_pred(S1,C1,T)+R_pred(S2,C2,T)) Among them, R_pred(S1,C1,T) and R_pred(S2,C2,T) are the predicted effects of single drugs. R_comb is the actual predicted effect of combined medication; Synergy Score calculation steps: For each pair of drugs (S1, S2) and each possible concentration combination (C1, C2), use R_comb=f(S1,C1,T)+f(S2,C2,T)+Interaction(S1,S2,C1,C2,T) Calculate R_comb; Calculate the Synergy Score of each pair of combinations; The three concentration combinations with the highest Synergy Scores were selected.

3. A method for efficiently evaluating the efficacy of in vitro combined drug therapy according to claim 1, characterized in that: The step 3 includes the following specific methods: Single drug concentration point test and 3x3 combined comprehensive test: Based on the AI ​​model prediction in step 2, the single drug test is integrated with the 3x3 combined drug concentration test; in this step, the selected drug is not only tested at a single concentration point (C1, C2, ..., C9), but also tested for its 3x3 combined effect with other drugs; (31) Experimental settings: The selected drugs were tested at 9 different concentration points. At the same time, these concentration points were also used for combination testing with other drugs. The selection of each concentration point was determined based on the expected drug activity range and previous experimental data to ensure coverage from ineffective doses to concentrations exceeding the expected therapeutic window. (32) Formula description and variable explanation: The dose-response curves for single-drug and combination therapy were described using the following model: Formula: R_i=100 / (1+(EC50 / C_i)^n) in: R_i: cell survival rate at concentration C_i, expressed as a percentage; C_i: the i-th concentration point, i=1,2,...,9; EC50: the concentration of a drug required to produce 50% of the maximal effect; n: hill shape parameter, describing the slope of the curve; (33) Determination method: Cell viability was determined using the CTG cell viability assay kit: Cell inoculation: Inoculate a certain number of target cells in a 96-well plate; Drug treatment: different concentrations of single drug and its combination solutions with other drugs were added to the corresponding wells; Cultivation: The treated cell plates are cultured under appropriate conditions for a certain period of time to allow the drug to act; Add CTG reagent: After the incubation period, add CTG reagent to each well. This reagent is reduced in living cells to produce a fluorescent or colorimetric signal. Signal detection: Use a microplate reader to read the signal in each well. The signal intensity is proportional to the cell survival rate. Data processing: Collect readings at all concentration points, including data of single-drug and combination-drug treatments, and calculate cell viability relative to untreated controls; At the same time as the test in step (33), the 3x3 matrix test of combined drug use is performed in combination with the prediction of the AI ​​model, and the data of the single drug is also obtained in this process, rather than performing a single drug concentration point test alone; (34) Experimental matrix setting: In this step, three concentration points predicted by the AI ​​model are set for each drug to form a 3x3 matrix; these concentration points are C1a, C1b, C1c for drug 1, and C2a, C2b, C2c for drug 2. Each point is used for both single-drug and combination drug testing. (35)Matrix description: Each matrix element R_ij represents the cell survival rate under the concentration Ci of drug 1 and the concentration Cj of drug 2, where i and j are indexes from 1 to 3, representing different concentration levels. This design allows the simultaneous evaluation of single drug effects and drug interactions; (36) Determination method: Cell seeding: Plant target cells in a 96-well plate to ensure that the number of cells in each well is consistent; Drug treatment: According to the setting of 3x3 matrix, add corresponding drug concentration combination to each well, including single drug and combination drug; Incubation: Incubate the treated cell plates under appropriate conditions, allowing the drug to act for an appropriate amount of time; Add CTG reagent: After the incubation, add CTG reagent to each well. The reagent is reduced in living cells to produce a quantitative signal. Signal reading: Use a microplate reader to read the signal intensity of each well, which is proportional to the cell viability; Data processing: Calculate the cell survival rate under each drug combination and compare it with the untreated control group, and record the effect of a single drug; (37)Formula and data analysis: Formula: R_ij=(Signal_ij / Signal_control)*100; R_ij: percentage of cell survival under drug combination (i, j); Signal_ij: signal intensity measured under drug combination (i, j); Signal_control: Signal intensity of untreated cells, used to normalize the results.

4. The method for efficiently evaluating the efficacy of in vitro combined drug therapy according to claim 1, characterized in that: The step 4 includes the following specific methods: (41) Single drug dose response curve fitting formula: R_single=R_max / (1+(EC50 / C)^HillSlope); R_single: cell survival rate at concentration C; R_max: maximum possible cell survival rate; EC50: drug concentration required to produce 50% of the maximal effect; C: drug concentration; HillSlope: Hill coefficient, which describes the slope of the curve and reflects the sensitivity of the drug effect to concentration changes; Fitting steps: Data preparation: Collect data from 9 concentration points from step (4): C1 to C9, and the corresponding cell viability: R1 to R9; Select Model: Select Hill equation as the fitting model; Parameter estimation: Use nonlinear least squares method to estimate the values ​​of R_max, EC50 and HillSlope and fit the experimental data; Model validation: Verify the quality of model fit by calculating the residuals between the predicted and actual values; (42) Analysis of the effect of combined medication After obtaining the dose-response curves of the single drugs, use these curves to predict the theoretical effects of the single drugs in combination testing to assess whether the interaction between the two drugs is synergistic, additive, or antagonistic; Calculation steps: Theoretical single drug effect prediction: For each concentration combination of the combined drug, the expected individual effects of the two drugs are calculated using the fitted dose-response curves; Comparison of actual combined effects: Comparison of theoretical single-drug effects with the combined drug effects actually measured by the CTG method; Synergy was assessed using the following formula: SynergyScore=ObservedEffect-(PredictedEffect1+PredictedEffect2) ObservedEffect: Observed effect of combined medication; PredictedEffect1, PredictedEffect2: the effects of two drugs predicted based on the single-drug dose-response curves; The above formula SynergyScore helps determine whether the combined use of drugs exceeds the simple addition of their individual effects, thereby determining whether there is a synergistic effect.

5. The method for efficiently evaluating the efficacy of in vitro combined drug therapy according to claim 1, characterized in that: The step 5 includes the following specific methods: (51) Data evaluation Evaluation Metrics: EC50 value: a measure of drug potency, referring to the concentration required to produce 50% of the maximum biological effect; Therapeutic index (TI): Therapeutic index is an indicator of drug safety and is calculated as follows: TI = TD50 / ED50; TD50: the dose that causes toxic reactions in 50% of the subjects; ED50: effective dose that produces 50% of the desired effect; Synergy Index (CI): It is used to evaluate the synergistic effect of two drugs when used in combination. The calculation formula is: CI = (D1 / Dx1) + (D2 / Dx2); D1 and D2: actual doses of the two drugs when used in combination; Dx1 and Dx2: doses of the two drugs that produce equivalent effects when used alone; A CI value less than 1 indicates synergism, equal to 1 indicates additive effect, and greater than 1 indicates antagonism; (52) Optimize experimental design Based on the results of the data evaluation in (51), the experimental design was adjusted to optimize the drug concentration range and combination of drugs; Adjustment steps: Concentration range adjustment: Redefine the concentration range of the drug based on the EC50 value and safety data, including lowering the maximum concentration to avoid toxicity or adjusting the minimum concentration to ensure efficacy; Optimization of drug combinations: Based on the results of the synergy index analysis, select those drug combinations that show synergistic effects and redesign the experiments to verify these findings; Repeat the experiment: Perform a new round of testing with adjusted drug concentrations and combinations to confirm the improved effects of these adjustments.

6. A memory chip, characterized in that: A system or program capable of executing the method according to any one of claims 1 to 5 is stored.

7. A device for efficiently evaluating the efficacy of combined medication in vitro, characterized in that: Contains the memory chip as claimed in claim 6.

Images