Virtual drug evaluation methods and applications in drug design, screening

Abstract

The present specification relates to the technical field of drug analysis, and particularly relates to a virtual drug evaluation method and application in drug design and screening. The virtual drug evaluation method comprises: obtaining characteristic data of a single cell; simulating a drug to obtain characteristic data of the drug and attribute data of the drug; combining the characteristic data of the drug with the characteristic data of the single cell to calculate interaction intensity between the drug and the single cell; analyzing the interaction intensity in combination with the attribute data of the drug to obtain differences after the drug is added; and evaluating the drug according to the differences. The embodiment of the present specification can have the advantages of human origin, the ability to reflect the characteristics of complex microenvironment in vivo, and low technical threshold and high-throughput screening.

Description

Technical Field

[0001] This specification relates to the field of pharmaceutical analysis technology, and in particular to a virtual drug evaluation method and its application in drug design and screening. Background Technology

[0002] In current technologies, drug development suffers from drawbacks such as long development cycles, high investment, and high risks, failing to meet the needs of public health. How to utilize cutting-edge technologies to shorten development cycles, reduce investment, and increase success rates is a crucial issue in drug development. A significant reason for the long development cycles, high investment, and high risks is the lack of models that can realistically simulate human disease states and drug responses. Current screening models, primarily based on cells and animal models, face challenges in translating from models to humans. For example, in vitro cultured cell models cannot reflect in vivo conditions, while animal models exhibit differences between humans and animals, and also fail to reflect differences between individuals. Due to these limitations, drugs often fail to achieve the expected effects in humans as in the models, leading to development failures. With the continuous development of new technologies, alternative models to animal models are emerging, such as organoid models. However, these technologies are still immature and face high technical barriers. Recently, the US FDA issued regulations reducing the use of animal models, thus the search for alternative models that more closely resemble human characteristics is receiving increasing attention.

[0003] Therefore, there is an urgent need for a more accurate and efficient method and technology for drug evaluation, design, and screening. Summary of the Invention

[0004] To address the problems in the prior art, this specification provides a virtual drug evaluation method and its application in drug design and screening, which solves the problem that the prior art lacks human-derived drugs that can reflect the characteristics of complex microenvironments in the body, while also having the advantages of low technical threshold and high throughput screening.

[0005] This specification provides a virtual drug evaluation method, including,

[0006] Obtain feature data of single cells;

[0007] Drug simulation is used to obtain the characteristic data and attribute data of the drug.

[0008] The interaction strength between the drug and the single cell is calculated by combining the characteristic data of the drug with the characteristic data of the single cell.

[0009] The interaction strength is analyzed by combining the property data of the drug to obtain the difference after the drug is added;

[0010] The drug is evaluated based on the differences described.

[0011] The embodiments in this specification also provide applications of drug design using the virtual drug evaluation method described above.

[0012] The embodiments in this specification also provide applications of using the virtual drug evaluation method described above for drug screening.

[0013] The embodiments in this specification also provide applications of using the virtual drug evaluation method described above for drug target discovery.

[0014] This specification also provides a virtual drug evaluation device, including a single-cell feature data acquisition unit, a drug simulation unit, an interaction strength calculation unit, a difference calculation unit, and an evaluation unit.

[0015] The single-cell feature data acquisition unit is used to acquire feature data of a single cell;

[0016] The drug simulation unit is used for drug simulation to obtain the characteristic data and attribute data of the drug.

[0017] The interaction strength calculation unit is used to combine the characteristic data of the drug with the characteristic data of the single cell to calculate the interaction strength between the drug and the single cell;

[0018] The difference calculation unit is used to analyze the interaction strength by combining the property data of the drug to obtain the difference after the drug is added;

[0019] The evaluation unit is used to evaluate the drug based on the differences.

[0020] This specification also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method.

[0021] This specification also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the above-described method.

[0022] Using the embodiments in this specification, current public databases store massive amounts of single-cell public data. This single-cell data realistically preserves typical cell types and characteristic phenotypes in the human body's internal environment, possessing advantages such as large data volume, high complexity, and strong accessibility. Theoretically, by obtaining typical cell types and phenotypes (such as gene expression levels) related to a patient's disease, a model simulating the patient's disease-related internal environment can be constructed in a computer. Furthermore, the process of exogenous substances disturbing cells can be simulated, and virtual "drug administration" can be performed on this model to observe the biological effects induced by the drug. Simultaneously, with the aid of high-performance computers, the above process can be carried out in parallel on hundreds or thousands of patients or drugs, sufficient to meet the needs of large-scale, high-throughput drug design, evaluation, and screening. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 The diagram shown is a flowchart of a virtual drug evaluation method according to an embodiment of this specification;

[0025] Figure 2A The diagram shown is a cell-cell interaction strength matrix A from an embodiment of this specification. cc Schematic diagram;

[0026] Figure 2B The diagram shown is a composite matrix of the interaction strengths of the components in the drug group in the embodiments of this specification.

[0027] Figure 3 The diagram shown is another flowchart of a drug analysis method according to an embodiment of this specification;

[0028] Figures 4A-4D The diagram shown illustrates the simulation of different types of drugs and pathogenic factors in the method for constructing an evaluation model based on single-cell transcriptomics, as described in the embodiments of this specification.

[0029] Figures 5A-5D The diagram shown is a simulation of the dose-response relationship of PD1 antibody in the embodiments of this specification;

[0030] Figures 6A-6D The figure shown is a simulation diagram of the affinity-pharmacodynamic relationship of the PD1 antibody in the embodiments of this specification;

[0031] Figure 7The diagram shows a comparison of the consistency between the predicted and actual results of single-cell data from patients with renal cell carcinoma (Kevin et al., Cancer Cell, 2021) as shown in the embodiments of this specification.

[0032] Figure 8A and Figure 8B The diagram shows a comparison of the consistency between the predicted and actual results of single-cell data from clinical breast cancer patients (Ayse et al., Nature Medicine, 2021) as used in the embodiments of this specification.

[0033] Figure 9 The diagram illustrates the evaluation of potential sensitive new immunotherapy targets for an ICB-resistant breast cancer patient (BIOKEY_8) from the aforementioned breast cancer dataset (Ayse et al., Nature Medicine, 2021) in an embodiment of this specification.

[0034] Figure 10A and Figure 10B The diagram illustrates an embodiment of this specification evaluating potential combination therapies that could reduce resistance to PD1 antibodies in all ICB-resistant breast cancer patients in the aforementioned breast cancer dataset (Ayse et al., Nature Medicine, 2021).

[0035] Figure 11 The diagram shows the simulation of PD-L1 antibody monotherapy and PD-L1 combined with VEGF antibody therapy on a single-cell dataset of 160 liver cancer patients (Xue et al., Nature, 2022).

[0036] Figures 12A-12C The diagram shows a CAR-T drug delivery simulation performed on a single-cell dataset of a glioma patient (Patient ID: ndGBM-09) according to an embodiment of this specification (Abdelfattah et al., Nature Communications, 2022).

[0037] Figure 13 The diagram shown is a structural schematic of a drug evaluation device according to an embodiment of this specification.

[0038] Figure 14 The diagram shown is a structural schematic of a computing device according to an embodiment of this specification.

[0039] [Explanation of Labels in the Attached Image]

[0040] 1301. Single-cell feature data acquisition unit;

[0041] 1302. Drug simulation unit;

[0042] 1303, Interaction Strength Calculation Unit;

[0043] 1304, Difference Calculation Unit;

[0044] 1305. Evaluation Unit;

[0045] 1402. Computing equipment;

[0046] 1404. Processing equipment;

[0047] 1406. Storage resources;

[0048] 1408. Drive mechanism;

[0049] 1410. Input / Output Module;

[0050] 1412. Input devices;

[0051] 1414. Output devices;

[0052] 1416. Presentation equipment;

[0053] 1418. Graphical User Interface;

[0054] 1420. Network interface;

[0055] 1422. Communication link;

[0056] 1424. Communication bus. Detailed Implementation

[0057] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.

[0058] This specification presents a virtual drug evaluation method based on single-cell data. Starting with human single-cell data from clinical samples and other sources, this method simulates the effects of drug addition to a complex internal environment on cell spatial relationships, interactions, and biological phenotypes, based on the drug's targeting and affinity for cells, and the self-organizing reaction mediated by affinity and steric repulsion. The drug evaluation model constructed using this method is human-derived, reflecting the complex in vivo microenvironment and individual differences between individuals, while also possessing high-throughput screening advantages. Based on this model, virtual pharmacodynamic evaluation, toxicology evaluation, drug screening, delivery efficiency evaluation, pathological mechanism research, and other work can be carried out, significantly saving unnecessary investment, reducing R&D risks, and shortening the R&D cycle, demonstrating broad application prospects.

[0059] like Figure 1 The diagram shows a flowchart of a virtual drug evaluation method according to an embodiment of this specification. The diagram illustrates the process of merging and analyzing simulated drug data and single-cell data based on drug simulation to obtain the interaction between the drug and single cells. The drug is evaluated based on the difference between the simulated drug and single cells after combination and the original single cells. The drugs mentioned in this embodiment may include antibodies, bispecific antibodies, triple antibodies, viral vector drugs, gene therapy, cell therapy, etc. In this embodiment, the properties of the drug are abstracted into several aspects such as targeting, affinity, concentration, size, and intrinsic activity. These concepts are then mathematically modeled to allow the drug to interact with human single-cell data from clinical samples, etc., based on targeting and affinity, to undergo an affinity-mediated self-organizing reaction. The efficacy and activity are evaluated by statistically analyzing changes in cell-cell and cell-drug interactions before and after drug addition. The method specifically includes:

[0060] Step 101: Obtain feature data for single cells;

[0061] Step 102: Drug simulation to obtain the characteristic data and attribute data of the drug;

[0062] Step 103: Combine the characteristic data of the drug with the characteristic data of the single cell to calculate the interaction strength between the drug and the single cell;

[0063] Step 104: Analyze the interaction strength by combining the property data of the drug to obtain the difference after the drug is added;

[0064] Step 105: Evaluate the drug based on the differences.

[0065] The feature data includes gene expression data, protein data, chromatin accessibility data, methylation modification data, and chromatin immunoprecipitation followed by sequencing, etc. In the subsequent embodiments of this specification, the feature data can be exemplified by gene expression data.

[0066] The drug evaluation methods described in the embodiments of this specification simulate drug delivery efficiency, targeting, and biological activity through affinity-mediated cell-drug self-organized reactions. These methods offer advantages such as direct application from clinical samples, low cost, and rapid computation. Therefore, they have broad application prospects in clinical diagnosis and treatment, including drug recommendation, drug target recommendation, virtual drug design, and efficacy evaluation, and are expected to significantly shorten drug development cycles, save development costs, and reduce development risks.

[0067] As one aspect of the embodiments of this specification, the property data of the drug includes at least one or a combination of the following:

[0068] Targeting, affinity, concentration, size, saturation, and intrinsic activity.

[0069] The drugs described in the embodiments of this specification include, but are not limited to, biological drugs such as antibodies, cytokines, feature-edited products, drug-conjugated formulations, and CAR-T cells, as well as small molecule drugs whose efficacy is mainly mediated by affinity. Furthermore, pathogenic agents such as bacteria and viruses can also be evaluated using this method. Based on different drug types and characteristics, parameters such as drug targeting, affinity, concentration, size, saturation, and intrinsic activity are simulated.

[0070] As one aspect of the embodiments of this specification, drug simulation, obtaining characteristic data and attribute data of the drug, further includes,

[0071] Perform non-cellular drug simulation and cellular drug simulation to obtain the characteristic data and attribute data of the non-cellular drug, and obtain the characteristic data and attribute data of the cellular drug.

[0072] In this embodiment, the drug includes non-cellular drugs and cellular drugs. Non-cellular drugs may include, but are not limited to, biological drugs such as antibody drugs and vector-conjugated drugs, or small molecule drugs whose efficacy is mainly mediated by affinity, or pathogenic factors such as viruses and allergens. Cellular drugs may include, but are not limited to, biological drugs such as CAR-T cells and blood products, or pathogenic factors such as bacteria.

[0073] As one aspect of the embodiments of this specification, the non-cellular drug simulation, which obtains the characteristic data and attribute data of the non-cellular drug, further includes,

[0074] The virtual features of the non-cellular drug are set to form a set of virtual features that serve as the feature data of the non-cellular drug;

[0075] Based on the virtual characteristics of the non-cellular drug, the attribute data of the non-cellular drug are obtained.

[0076] In this embodiment, the virtual features include one or more features. A non-cellular drug can be considered as a "cell-like" expression of a virtual feature set X. The virtual feature set X is a set of virtual features that do not exist in the single-cell feature data. The abundance of the virtual feature set X can be set to a fixed value (e.g., 1) or a unit value related to concentration or dose, reflected in the abundance of the virtual features. The abundance is positively correlated with the drug concentration, and this unit value can reflect the real drug concentration (but is not necessarily equivalent). For example, the abundance of the virtual feature set X can be the number of drugs (or logarithm), etc. The virtual feature set X can be reflected in the amount of drug data (the number of columns in the matrix). For example, 1 column of data represents 1 unit of drug, 10 columns represent 10 units of drug, and so on, simulating concentration. In addition, random noise or gradients can be introduced into the abundance of the virtual feature set X to simulate the intermolecular differences of drugs (such as the order of entry into lesion tissue, drug purity, viral structural variations, etc.). For example, the abundance of the virtual feature set X can be randomly sampled from a Gaussian distribution with a mean of the number of drugs (or logarithmic form, etc.). Based on the virtual characteristics of the non-cellular drug, various properties of the non-cellular drug can be obtained.

[0077] As one aspect of the embodiments of this specification, the attribute data of the non-cellular drug includes at least one or a combination of the following: targeting, affinity, concentration, size of the non-cellular drug, saturation, and intrinsic activity.

[0078] In this embodiment, the targeting refers to the correspondence between the virtual features of the non-cellular drug and its target; the affinity refers to the weighting coefficient of the correspondence; the concentration refers to the number of non-cellular drug data points and the abundance of the virtual feature set; the size of the non-cellular drug refers to its volume in three-dimensional space; the saturation refers to the upper limit of non-cellular drug binding to the target cell; and the intrinsic activity refers to the relative potential of the non-cellular drug to induce a maximum functional response after binding to the target, and its potential impact on downstream phenotypic and characteristic changes. The number of data points refers to the unit of drug quantity, equivalent to the number of columns in the virtual drug matrix.

[0079] As one aspect of an embodiment of this specification, obtaining the attribute data of the non-cellular drug based on its virtual characteristics further includes,

[0080] In the background network of receptor-ligand interactions, the virtual feature set of the non-cellular drug and the binding relationship of the drug's target receptor-ligand are added to obtain the targeting of the non-cellular drug.

[0081] Add affinity weighting coefficients to the receptor-ligand binding relationships between the virtual feature set and non-cellular drug targets;

[0082] The concentration of the non-cellular drug is obtained by setting the number of non-cellular drug data and the abundance of the virtual feature set;

[0083] The non-cellular drug is defined as a molecule with radius r that can rotate freely in space, and the size of the non-cellular drug is obtained.

[0084] The upper limit of the amount of non-cellular drug that binds to the target cells of the non-cellular drug is set to obtain the saturation.

[0085] The intrinsic activity is obtained based on the changes in the downstream phenotype and characteristics of the target cells corresponding to the non-cellular drug.

[0086] In this embodiment, in the background network of receptor-ligand interaction, a receptor-ligand binding relationship is added between the virtual feature set of the non-cellular drug and the drug target. For example, if the target of feature X is EGFR (Epidermal Growth Factor Receptor), a connection between the two nodes of feature X and EGFR is added in the background network of receptor-ligand interaction to reflect the targeting effect of the non-cellular drug on the target on the cell.

[0087] To differentiate the differences in affinity intensity of different non-cellular drugs for targets, it is necessary to further add an affinity weighting coefficient to the receptor-ligand binding relationship between the virtual feature set X and the drug target. The affinity weighting coefficient W is a real number greater than 0.

[0088] To simulate different non-cellular drug concentrations, it is necessary to set the amount of non-cellular drug data and the abundance of virtual feature X. The larger the amount of non-cellular drug data and the higher the abundance of virtual feature X, the greater its concentration is reflected.

[0089] Since non-cellular drugs are entities existing in three-dimensional space, their spatial properties need to be defined. In the embodiments of this specification, a non-cellular drug is abstracted as a molecule that can rotate freely in a space of radius r. The radius r of the drug is usually smaller than the radius R of the cell. Because each entity occupies a certain spatial position, the interaction between different entities will be affected by steric hindrance. Each non-cellular drug will only come into contact with a limited number of other drugs or cells.

[0090] To obtain drug-cell interaction relationships, the attribute data of non-cellular drugs in the embodiments of this specification also include saturation constraint conditions. Since the number of drug targets expressed by each cell is limited, there is an upper limit to the number of drugs that can bind to the target cell. Saturation constraint can be implemented through hard constraints or flexible constraints. Hard constraints are constraints when constructing the interaction strength matrix, setting an upper limit of m drugs that can bind to the target cell, where m is a positive integer greater than 0, such as equal to 10, 50, or the abundance of cell target features, etc., and the interaction potential of drugs binding to target cells with other cells will also disappear or weaken. Flexible constraints can be implemented by adjusting parameters during dimensionality reduction. For example, for drugs beyond m, the minimum cell-drug distance threshold will be adjusted, resulting in cells having a greater spatial repulsion to other drugs. In addition, not all conditions need to consider saturation constraints. For example, some polymeric compounds can embed a large number of cells through their special spatial structure, so it is assumed that the non-cellular drug can bind to an infinite number of cells.

[0091] Intrinsic activity refers to the relative potential of non-cellular drugs to induce a maximum functional response after binding to a target. Based on the magnitude of this relative potential, they can be classified as agonists, partial agonists, inhibitors, etc. For example, PD-1 receptor agonists can induce increased expression of immune checkpoint-related features (IRLs) in target cells, while PD-1 receptor inhibitors may downregulate the expression of IRLs in target cells. The effects of different receptor agonists / inhibitors on downstream phenotypes can be experimentally determined or referenced from databases such as cMAP and KEGG.

[0092] As one aspect of the embodiments of this specification, the cell-based drug simulation, which obtains the characteristic data and attribute data of the cell-based drug, further includes,

[0093] Obtain characteristic data for each cell type of the cell-based drug;

[0094] Based on the characteristics of the cell-based drug, the attribute data of the cell-based drug are obtained.

[0095] In this embodiment, the feature data includes cell features and corresponding abundance. The feature data of each cell type of the cell-based drug can be derived from real single-cell data. By splicing the matrix of these features with the single-cell data of the control group, the perturbation of the control group environment by exogenous cells can be simulated. For example, the single-cell data of tumors can be spliced ​​with the single-cell data of peripheral blood mononuclear cells to simulate the effect of blood immune cells entering tumor tissue.

[0096] As one aspect of the embodiments of this specification, obtaining the characteristic data of each cell type of the cell-based drug also includes,

[0097] Set the abundance of the feature data.

[0098] In this embodiment, the different effects of exogenous cellular drugs with different phenotypes / states on the target environment are simulated by changing the abundance of real features (e.g., a single feature or a set of pathway features).

[0099] As one aspect of the embodiments of this specification, the property data of the cell-based drug includes at least one or a combination of the following: targeting, affinity, concentration, size of the cell-based drug, and saturation.

[0100] In this embodiment, the targeting refers to the correspondence between the features of the cell-based drug and the target of the cell-based drug; the affinity refers to the weighting coefficient of the correspondence; the size of the cell-based drug refers to the volume of the cell-based drug in three-dimensional space; and the saturation refers to the upper limit of the cell-based drug that binds to the target cell of the cell-based drug.

[0101] As one aspect of an embodiment of this specification, obtaining the attribute data of the cell-based drug based on its characteristics further includes,

[0102] The targeting of the cell-based drug is obtained by matching the feature set in the real cell-based drug with the target of the cell-based drug.

[0103] Based on the receptor-ligand binding relationship between the feature set and the cell drug target, the affinity weighting coefficient is obtained;

[0104] The number of cell-based drug data points is set to obtain the concentration of the cell-based drug;

[0105] The size of the cell-based drug is obtained based on its radius r.

[0106] The upper limit of the cell-based drug binding to the target cells of the cell-based drug is set to obtain the saturation.

[0107] The intrinsic activity is obtained based on the effects of the cell-based drug on the drug target, including inhibition, activation, and regulation, and the resulting changes in the downstream phenotypes and characteristics of the corresponding cell-based drug target.

[0108] In this embodiment, the targeting, affinity, and size of the cell-based drug can be obtained through experimental means or existing databases, wherein the radius r of the cell-based drug is usually equal to the radius R of other cells.

[0109] As one aspect of the embodiments of this specification, obtaining the characteristic data of each cell type of the cell-based drug also includes,

[0110] A virtual feature set is added to the feature data of each cell in the cell-based drug to form the feature data of the cell-based drug.

[0111] In this embodiment, virtual features include one or more features. For simulation of a specific type of cell-based drug, a set of virtual features X needs to be added. For example, when simulating chimeric antigen receptor T cells (CAR-T), the virtual feature of chimeric antigen receptor (CAR) can be added to the single-cell matrix of T cells. The CAR abundance of each cell can be set according to the abundance of other features of that cell or the expression efficiency of the feature expression vector (e.g., CAR abundance equals the maximum value of the original feature abundance of that cell).

[0112] As one aspect of the embodiments of this specification, when the feature data of the cell-based drug includes a virtual feature set, the step of obtaining the attribute data of the cell-based drug based on its features further includes,

[0113] By adding the feature set of the cell-based drug to the background network of receptor-ligand interactions and the binding relationship between the drug's target receptor and ligand, the targeting of the cell-based drug can be obtained.

[0114] Add an affinity weighting coefficient to the receptor-ligand binding relationship between the aforementioned feature data and the cell-based drug target;

[0115] The number of cell-based drug data points is set to obtain the concentration of the cell-based drug;

[0116] The size of the cell-based drug is obtained based on its radius r.

[0117] The upper limit of the cell-based drug binding to the target cells of the cell-based drug is set to obtain the saturation.

[0118] The intrinsic activity is obtained based on the changes in the downstream phenotype and characteristics of the target of the cell-based drug. The effect on the drug target may include, for example, inhibition, activation, and regulation.

[0119] As one aspect of an embodiment of this specification, combining the characteristic data of the drug with the characteristic data of the single cell to calculate the interaction strength between the drug and the single cell further includes,

[0120] The characteristic data of the single cells were used as a control group.

[0121] The drug feature matrix formed by the feature data of the drug is concatenated with the single cell feature matrix formed by the feature data of the single cell to form drug group data.

[0122] Calculate the cell-cell interaction strength matrix A in the control group. cc ; Calculate the cell-cell interaction strength matrix (A) in the drug group. cc ), cell-drug interaction strength matrix (A cb ), drug-drug interaction strength matrix (A bb ).

[0123] In this embodiment, the interaction strengths of the aforementioned components are calculated based on a publicly available or custom-designed receptor-ligand database. Each element in the interaction strength matrix A represents the interaction strength between the corresponding component i and component j, where component i represents cell i or drug i, and component j represents cell j or drug j. The relationship of the interaction strengths is obtained based on single-cell data features, for example, by using the law of mass action formula to obtain the cell-cell (A) interaction strength. cc The interaction strength matrix is ​​used to simulate the original state without drugs; for the drug group, the obtained cell-cell (A) interaction strength matrix needs to be used. cc ), Cell-Drug (A) cb ), drug-drug (A) bb Interaction strength matrices are spliced ​​together to simulate the perturbation state when drugs are present.

[0124] As one aspect of the embodiments of this specification, the cell-cell interaction strength matrix A in the control group is calculated. cc ; Calculate the cell-cell interaction strength matrix A in the drug group. cc Cell-drug interaction strength matrix A cb Drug-drug interaction strength matrix A bb Including the following formulas:

[0125] or

[0126] or

[0127]

[0128] Among them, A i,j W represents the interaction strength between component i and component j, where component i represents cell i or drug i, and component j represents cell j or drug j. A,B W represents the weight of the interaction between ligand A and receptor B. A,B The default value is 1, A i and A j Represent the expression levels of ligand A in components i and j, respectively, and B i and B jrepresents the expression level of receptor B in component i and component j, respectively, and k represents the total number of ligand-receptor pairs.

[0129] In this embodiment, other methods for calculating interaction strength can also be used, which will not be elaborated here.

[0130] As one aspect of this specification, the method further includes, before analyzing the interaction strength in conjunction with the property data of the drug, the following steps are taken:

[0131] The interaction strength is sparsified based on the drug's property data.

[0132] In this embodiment, the interaction strength is represented as an interaction strength matrix, which records the interaction potential of any two cells / drugs. However, due to factors such as spatial hindrance and receptor saturation, a cell / drug can only interact with a limited number of cells / drugs. Therefore, each cell / drug only needs to retain a limited number of objects with the strongest interaction potential. That is, the interaction strength matrix is ​​sparsified to simulate spatial obstruction, interaction saturation, etc.

[0133] As one aspect of an embodiment of this specification, sparsifying the interaction strength based on the property data of the drug further includes,

[0134] For the control group, only the top N cells with the highest interaction strength with each cell were retained, and the interaction strength of other cells was set to 0.

[0135] For the drug group, in the cell-cell interaction strength matrix (A cc In the cell-cell interaction strength matrix (A), only the cell-cell interaction strength matrix is ​​retained. cc The interaction strength of the top N cells with the highest interaction strength with each cell is set to 0, and the interaction strength of other cells-cell interactions is set to 0. Cell-cell interaction strengths that are less than the maximum cell-drug interaction strength are also set to 0.

[0136] In the cell-drug interaction strength matrix (A cb In this process, based on the saturation of the drug, only the n cells with the highest interaction strength with each drug are retained, while the interaction strength of other cells-drugs is set to 0. When these n cells bind to m drugs, their interaction strength with other drugs will be set to 0.

[0137] In the drug-drug interaction strength matrix (A bb In the example, set all elements to values ​​greater than or equal to 0.

[0138] In this embodiment, for the control group, the cell-cell interaction strength matrix (A) cc Each element in the equation represents the interaction strength A between cell i and cell j. i,j ,like Figure 2A The diagram shown is a cell-cell interaction strength matrix A from an embodiment of this specification. cc This diagram illustrates that for each cell in the matrix, only the top N cells with the highest interaction strength are retained (the interaction strength of the Nth cell is A). i,(N) ), and less than A i,(N) The cell-cell interaction strength is set to 0, and N is a positive integer greater than 0, such as 50, 100, etc.

[0139] For the drug group, the cell-cell interaction strength matrix (A) cc Each element in the equation represents the interaction strength A between cell i and cell j. i,j ,like Figure 2B The diagram shown is a composite matrix of the interaction strengths of the components in the drug group of this specification. First, for each cell in the matrix, only the top N cells with the highest interaction strength are retained (the interaction strength of the Nth cell is A). i,(N) ), and less than A i,(N) The cell-cell interaction strength is set to 0, and N is a positive integer greater than 0, such as 50, 100, etc. In addition, considering the competitive binding between the drug and cells, values ​​less than the maximum cell-drug interaction strength max(A) are further set. i,t The cell-cell interaction strength of A was also set to 0. i,j ≤max(A i,t (A) refers to the strength of cell-cell interactions. i,j Will be affected by cell-drug interaction strength A i,t Replace, less than A i,t A i,j Set to 0.

[0140] For the drug group, the cell-drug interaction strength matrix (A) cb Each element in the equation represents the interaction strength A between the corresponding cell i and drug t. i,t ,like Figure 2B As shown in the embodiments of this specification, saturation constraints can be incorporated into the consideration of drug-cell interactions. If saturation constraints are not applied, each drug can retain only the n cells with the highest interaction strength (the interaction strength of the nth cell is A). (n),t ), and less than A (n),tCell-drug interactions are set to 0, where n is a positive integer greater than 0, such as 1, 10, etc. If saturation constraints are applied, hard constraints can be applied during this sparsification process. Hard constraints mean that when each drug is sparsified, it is necessary to determine whether the cell has reached saturation. If the cell has been bound by m drugs (the interaction strength of the m-th drug is A...), then... i,(m) ), that is, A i,(m) If the value is greater than 0, the cell is considered to have reached saturation and no longer has the potential to bind other drugs, and the remaining drug A... i,t All values ​​are set to 0, and the remaining drugs can only find binding targets in unsaturated cells, meaning only the drug-unsaturated cell interaction strength matrix A is retained. cb The interaction strength between the top n drugs with the highest interaction strength with each drug and unsaturated cells, where m is a positive integer greater than 0, such as equal to 10, 50, or the abundance of the target feature per cell.

[0141] For drug groups, the drug-drug interaction strength matrix (A) bb Each element in the equation represents the interaction strength A between the corresponding drug t and drug s. t,s ,like Figure 2B As shown, when the research objective is primarily related to cell-cell and cell-drug interactions, A t,s It can be set to 0.

[0142] As one aspect of an embodiment of this specification, sparsifying the interaction strength based on the property data of the drug further includes,

[0143] In the drug group, in the drug-drug interaction strength matrix (A bb In the drug-drug interaction strength matrix, only the interaction strengths of the top N drugs with the highest interaction strength are retained, and the interaction strengths of other drug-drug interactions are set to 0. Cell-cell interaction strengths that are less than the maximum drug-drug interaction strength are also set to 0.

[0144] Based on the saturation of the drugs, each drug is judged by retaining only the n drugs with which it has the strongest interaction. If a drug has already interacted with m drugs, it is considered to have reached saturation and no longer has the potential to interact with more drugs.

[0145] In this embodiment, when simulating specific scenarios (such as drug interactions) or multiple drug interactions (such as drug targeting pathogens), A t,s It can be set to a value greater than 0, and its sparsity rule can be the same as A. cc Or A cb same

[0146] As one aspect of the embodiments of this specification, the analysis of the interaction strength in conjunction with the property data of the drug to obtain the difference before and after the addition of the drug further includes,

[0147] By comparing the interaction strength matrices of the drug group and the control group, the differences before and after the addition of the drug can be obtained.

[0148] In this embodiment, by analyzing the difference in the interaction strength matrix before and after drug addition, important information related to the drug can be extracted. For example, a simple method can be used to statistically analyze whether the interaction between the drug and the target cell is stronger or weaker than the interaction between other cells and the target cell, and how they are ranked.

[0149] As one aspect of the embodiments of this specification, the analysis of the interaction strength in conjunction with the property data of the drug, and the resulting differences before and after the addition of the drug, also include:

[0150] Based on the interaction strength, the interaction network of the control group and the interaction network of the drug group are obtained;

[0151] By comparing the interaction network of the control group and the interaction network of the drug group, the differences before and after the addition of the drug were obtained.

[0152] In this embodiment, the interaction networks of the control group and the drug group can be obtained based on the interaction strength matrix. Cell-cell, cell-drug, and drug-drug interaction networks can be constructed by setting the same or different interaction strength thresholds. For example, when constructing the cell-cell interaction network, an intercellular affinity threshold can be selected, such that each cell interacts with an average of h cells, resulting in an intercellular interaction network. Here, h is a positive integer greater than 0, such as 3 or 5. For each cell, the affinity with the cell closest to it in the h-th order is calculated, and the median of the calculated affinity for all cells is taken as the threshold. After obtaining the threshold, for each pair of cells, if their affinity is greater than the specified threshold, they are considered to interact, thus obtaining the cell interaction network.

[0153] As another embodiment, after obtaining the drug interaction network, the effects of drugs with specific intrinsic activities on cell phenotypes and characteristics can also be considered, which requires iterative processing. First, based on the current drug interaction network, the target cell set of the drug is obtained, and the original single-cell data of the target cells is perturbed, for example, by upregulating the levels of target features and target pathway-related features, to simulate the effect of the drug. Then, all steps in steps 103 to 104 above are repeated on the perturbed expression matrix.

[0154] As one aspect of the embodiments of this specification, the analysis of the interaction strength in conjunction with the property data of the drug, and the resulting differences before and after the addition of the drug, also include:

[0155] Calculate the probability matrix of the interaction strength based on the interaction strength;

[0156] Construct a dimensionality reduction loss function;

[0157] The virtual spatial coordinates of cells or drugs in the control group and drug group are obtained using the probability matrix and the dimensionality reduction loss function;

[0158] Based on the virtual space coordinates, the interaction network of the control group and the interaction network of the drug group are obtained;

[0159] By comparing the interaction network of the control group and the interaction network of the drug group, the differences before and after the addition of the drug were obtained.

[0160] In this embodiment, since the interaction between the drug and cells actually occurs in a specific space and is inherently low-dimensional, while the resulting interaction intensity matrix is ​​high-dimensional, to achieve a more realistic simulation, the cells / drugs of the two groups of samples can be projected into a virtual space based on the interaction intensity matrices of the two groups (i.e., the interaction intensity matrix of the control group and the interaction intensity matrix of the drug). The virtual space can be a one-dimensional linear, two-dimensional planar, three-dimensional spherical, or other more complex manifold structure, or a real tissue structure obtained by spatial grouping technology. This method of reconstructing the spatial relationships of the components can refer to existing tools such as CSOmap. For example, after obtaining the interaction intensity matrices of the control group and the drug group, each element of the interaction intensity matrix is ​​divided by the sum of all elements, Z. p To obtain the probability matrix P of the interaction strength matrix:

[0161]

[0162]

[0163] Where I represents the total number of cells and drugs. p ij This represents the interaction probabilities between cell / drug i and cell / drug j in the probability matrix P of the interaction strength. Based on this matrix, a low-dimensional spatial structure of cell-drug interactions will be further constructed. Dimensionality reduction involves constructing a cell / drug distribution similar to the high-dimensional space in a low-dimensional virtual space. Therefore, a loss function for dimensionality reduction must first be constructed. Kullback-Leibler divergence, Jensen-Shannon divergence, etc., can be used. For example, when the loss function is Kullback-Leibler...

[0164] The dimension reduction loss function for divergence is: At that time, it made:

[0165]

[0166]

[0167]

[0168] Where I is the total number of cells and drugs, q ij Z is the probability that cell / drug j is around cell / drug i in virtual space. q It refers to the distribution of all cells / drugs in virtual space, d ij This represents the distance between cell / drug i and cell / drug j in virtual space; for example, when M=3, it is the Euclidean distance in three-dimensional space. i m The coordinates of the m-th dimension of cell / drug i are y. j m f(d) is the m-th dimension coordinate of cell / drug j; ij ) is determined by distance d ij The defined density function, for example, can be defined as: (a and b are both constants) etc.

[0169] Based on the aforementioned dimensionality reduction loss function, the coordinates of cells and drugs in virtual space are solved by minimizing the dimensionality reduction loss function. The dimensionality reduction loss function can be updated using methods such as gradient descent to update cell / drug coordinates and calculate the gradient direction of each cell / drug at the current coordinates. For example, in two-dimensional or three-dimensional space, the density function is defined as... When the coordinates are obtained, they can be iterated using the gradient descent method according to the following formula:

[0170]

[0171] Where C is the dimensionality reduction loss function, y i Let y be the current coordinate of cell / drug i in a certain dimension. j Given the current coordinates of cell / drug j in this dimension, the gradient direction is used as the coordinate update direction. The cell / drug coordinates are updated with a fixed step size or different step sizes. After multiple iterations, the virtual spatial coordinates of the control group and the drug group are finally obtained, so as to more intuitively present the interaction relationships between cells, cells and drugs, and drugs.

[0172] The interaction networks of the control group and the drug group can be constructed based on a distance matrix of virtual spatial coordinates. Cell-cell, cell-drug, and drug-drug interaction networks can be built by setting the same or different distance thresholds. For example, when constructing the cell-cell interaction network, a cell distance threshold can be selected such that each cell interacts with an average of h cells, resulting in an intercellular interaction network. Here, h is a positive integer greater than 0, such as 3 or 5. For each cell, the distance to its h-th closest cell is calculated, and the median of these distances is taken as the threshold. After obtaining the threshold, for each pair of cells, if their distance is less than the specified threshold, they are considered to interact, thus obtaining the cell interaction network.

[0173] As another embodiment, after obtaining the drug interaction network, the effects of drugs with specific intrinsic activities on cell phenotypes and characteristics can also be considered, which requires iterative processing. First, based on the current drug interaction network, the target cell set of the drug is obtained, and the original single-cell data of the target cells is perturbed, for example, by upregulating the expression levels of target features and target pathway-related features, to simulate the effect of the drug. Then, all steps in steps 103 to 104 above are repeated on the perturbed expression matrix.

[0174] As one aspect of the embodiments of this specification, the construction of the dimensionality reduction loss function also includes,

[0175] The size of the drug attribute data is added as a constraint to the dimensionality reduction loss function.

[0176] In this embodiment, since the volumes of cells and drugs typically differ, constraints can be imposed on the size of the drug volume. Therefore, different d values ​​can be added to the constraint settings. ij Threshold. For example, in cell-cell interactions, the distance d between cell i and cell j in virtual space. ij ≥r cc Where i≠j, r cc The sum of the radii between cell i and cell j; the distance d between cell i and drug j in virtual space during cell-drug interaction. ij ≥r cb Where i≠j, r cb Let be the sum of the radii between cell i and drug j. Furthermore, if saturation constraints are applied, flexible constraints can be added during this spatial constraint process. For example, when there are m drugs whose distance to cell i is less than a distance threshold d, then the distance between cell i and other drugs j should satisfy d. ij >>r cb In drug-drug interactions, the distance d between drug i and drug j in virtual space... ij≥r bb Where i≠j, r bb This is the sum of the radii between drug i and drug j. Constraints can also be applied to the entire mutual space, for example, by setting |y... i m |≤R, where R is the radius of the entire interaction three-dimensional space.

[0177] As one aspect of the embodiments of this specification, the construction of the dimensionality reduction loss function also includes,

[0178] The density function in the dimensionality reduction loss function is determined based on the affinity in the drug property data.

[0179] In this embodiment, since the kinetic behaviors of cells and drugs under affinity mediation may differ, the density functions f(d) of cell-cell, cell-drug, and drug-drug interactions are... ij The density function (DFC) can have different definitions. For example, the cell-to-cell density function can be defined as... The cell-drug density function can be defined as follows: The drug-drug density function can be defined as follows: (a, b, e, f, g, and h are all constants).

[0180] As one aspect of the embodiments of this specification, the evaluation of the drug based on the said difference further includes,

[0181] Analyze the differences before and after drug addition to obtain the drug's pharmacological and toxicological potential.

[0182] In this embodiment, drug design, evaluation, and screening can be carried out based on the differences before and after drug addition. Application scenarios include, but are not limited to, drug sensitivity evaluation, high-throughput drug screening, combination drug recommendation, drug target discovery, off-target effects, toxicity effect evaluation, and damage mechanism analysis of pathogenic factors. The following are specific application examples based on the drug evaluation methods described in this specification:

[0183] Efficacy evaluation of immune checkpoint inhibitors: Immune checkpoint inhibitors can kill tumor cells by activating exhausted T cells and promoting the infiltration of non-exhausted T cells into tumor cells. Therefore, after simulating the addition of immune checkpoint inhibitors, the number of interactions between non-exhausted T cells and tumor cells in the treatment group and the control group can be counted. By comparing the two groups, it can be determined whether the added immune checkpoint inhibitors promoted the immune infiltration of non-exhausted T cells into tumor cells, thus evaluating the efficacy of immune checkpoint inhibitors.

[0184] Drug toxicity evaluation: Drug overdose is often the cause of toxicity. By simulating the addition of different doses of drug, the number of drug interactions with target cells and non-target cells in the drug administration group is statistically compared, and the cell-cell interactions between the control group and the drug administration group are compared. On the one hand, the safe dosage range of the drug can be calculated to evaluate the safety of the drug. On the other hand, the target cells of the drug added in excess and the efficacy of the drug on cell interactions can be statistically analyzed to evaluate the possible mechanisms of drug-induced toxicity.

[0185] Drug design: By simulating key factors affecting drug efficacy, such as drug type, affinity, target, and intrinsic activity, we can provide insights for the design of target drugs. For example, non-cellular drugs like bispecific antibodies can target different targets with their two arms to improve target cell specificity or bridge different cell types. However, balancing the affinity of the two arms of a bispecific antibody is a challenge in the design of such drugs. The method of this invention can simulate a series of bispecific antibodies with different affinities for their arms, evaluate the on-target / off-target cell ratios of these antibodies, and find the affinity region with the lowest off-target ratio, thereby designing the optimal bispecific antibody drug. The challenge in designing cellular drugs like CAR-T lies in how to modify T cells, including which CAR to assemble or alter the original gene expression pattern, to improve CAR-T efficacy and reduce toxicity. The method of this invention can evaluate CAR-T drugs assembled with single or combined CARs targeting different targets, or evaluate the efficacy and toxicity of CAR-T drugs that overexpress / knock down different gene sets, in order to design the most potentially effective CAR-T drugs.

[0186] Drug screening: For individual patients, the optimal drug and treatment dose can be screened by simulating the therapeutic effects of different marketed drugs at different doses.

[0187] Drug target discovery: For specific types of diseases, drugs that broadly mimic different receptors or ligand proteins in the human body can be used to observe the effects of perturbing different targets, either individually or in combination, on the disease microenvironment, thereby discovering new potential drug therapeutic targets.

[0188] Drug delivery efficiency evaluation: Many drug targets are not only highly expressed in lesion areas but also widely present in various organs and tissues of the human body. Drug delivery efficiency can be calculated by statistically analyzing the ratio of drug entering the lesion area to drug entering other normal organs and tissues. By integrating single-cell data from normal organs and tissues and the lesion area, and using the method proposed in this patent, drug addition is simulated, and the amount of drug binding to target cells in the lesion area and the ratio of the amount of drug binding to other normal organs and tissues are calculated to evaluate drug delivery efficiency.

[0189] Combination therapy recommendations: By constructing evaluation models on specific patients and statistically analyzing the changes in the interaction between different drug combinations and target cells, combination therapy recommendations can be made for patients.

[0190] like Figure 3 The diagram shown is another flowchart of a drug analysis method according to an embodiment of this specification. This diagram illustrates how the acquired single-cell data are used as a control group and a drug group, respectively, and the drug is evaluated by comparing the differences between the two groups. The method specifically includes:

[0191] Step 301: Obtain single-cell data.

[0192] In this step, single-cell data refers to the single-cell feature matrix, which is used as the control group.

[0193] Step 302, drug simulation, forming a drug group.

[0194] In this step, the main characteristics of the simulated drug are simulated, such as the drug's dosage / concentration, the drug's target site, the drug's affinity for the target site, the drug's volume, and the drug's form (non-cellular or cellular).

[0195] Step 303: For the drug group, obtain the interaction strength matrix of each component.

[0196] This requires obtaining the cell-cell interaction strength matrix, the cell-drug interaction strength matrix, and the drug-drug interaction strength matrix.

[0197] For the control group, the interaction strength matrix of each component is obtained.

[0198] Among them, the cell-cell interaction strength matrix is ​​obtained.

[0199] Step 304: For the drug group, sparsify the interaction strength matrix.

[0200] Among them, the cell-cell interaction strength matrix, cell-drug interaction strength matrix, and drug-drug interaction strength matrix are sparsified.

[0201] For the control group, the interaction strength matrix is ​​sparsified.

[0202] Among them, the cell-cell interaction strength matrix is ​​sparsified.

[0203] Step 305: For the drug group, obtain the interaction network of each component.

[0204] This involves acquiring cell-cell interaction networks, cell-drug interaction networks, and drug-drug interaction networks. Considering the impact of drugs with specific intrinsic activities on cell phenotypes and characteristics, it is necessary to return to step 303 to re-acquire the interaction strength matrix of each component of the drug group for iterative processing.

[0205] For the control group, the interaction network of each component was obtained.

[0206] Among these, the cell-cell interaction network was obtained.

[0207] Step 306: Compare the interaction network of the control group with the interaction network of the drug group to evaluate the effect of the drug.

[0208] This specification also discloses an embodiment that calculates the intercellular interaction strength matrix based on single-cell transcriptome data and known receptor-ligand interaction relationships, and constructs an evaluation model in three-dimensional coordinates using a nonlinear dimensionality reduction method. Based on this, and combined with pharmacodynamic theory, it allocates drug concentration, drug, receptor / target, receptor-ligand affinity, and intrinsic activity to evaluate the drug's impact on cell relationships, phenotype, and fate. The drugs that can be evaluated include not only vaccines, antibodies, blood components, somatic cells, gene therapy, tissues, small molecule compounds, recombinant therapeutic proteins, and cell therapy drugs, but also potential pathogenic factors such as viruses, bacteria, fungi, and allergens. Typical simulation examples include... Figures 4A-4D The diagram illustrates the simulation of different types of drugs and pathogenic factors in the evaluation model based on single-cell transcriptomics in the embodiments of this specification. It uses single-cell datasets from screened melanoma patients (Tirosh et al., Science, 2016) and complete glioma patients (Abdelfattah et al., Nature Communications, 2022) to simulate antibody drugs (e.g.) Figure 4A As shown), virus simulation (such as) Figure 4B (as shown) and CAR-T cell drug mimicry (such as Figure 4C Perturbation simulations (as shown), and mixed simulations of viruses and viral antibodies (as shown) Figure 4D (As shown).

[0209] In addition to simulating different types of drugs, the examples in this specification can also simulate the pharmacological and toxicological effects of specific drugs under different conditions (such as dosage / concentration, receptor affinity). For example, Figures 5A-5DThe diagram shown is a simulation of the dose-response relationship of PD1 antibody in the embodiments of this specification. The data used is a single-cell dataset of melanoma patients selected from screening (Tirosh et al., Science, 2016) (Ren et al., Cell Research, 2020). Perturbation simulations of PD1 antibody drug at different concentrations (25, 50, 100, 250, 500, 1000 units) were performed. The affinity weighting coefficient between PD1 antibody and target was 20, and each cell could bind a maximum of 10 antibodies. Figures 5A-5C The images show the effects of 100 units of antibody (unsaturated dose), 250 units of antibody (saturated dose), and 500 units of antibody (supersaturated dose). Figure 5D The results showed that as the concentration of PD1 antibody in the environment increased, more target cells bound to PD1 antibody, and the trend of change showed a typical dose-response S-shaped curve. At low and high concentrations of PD1 antibody, the binding efficiency between the antibody and the target site on the target cell was low, while the binding efficiency was high when the PD1 antibody was at an intermediate concentration. When the PD1 antibody was close to saturation, the rate of increase in binding efficiency slowed down significantly or stopped changing.

[0210] like Figures 6A-6D The figure shown is a simulation diagram of the affinity-pharmacodynamic relationship of PD1 antibody in the embodiments of this specification. Using the same dataset and hyperparameters, 500 units of PD1 antibody were added to simulate the perturbation of PD1 antibody with different target affinities (weighting coefficients include 1, 5, 10, 20, 40, 80, and 160). Figures 6A-6C The images show the effects of low-affinity, medium-affinity, and high-affinity antibodies. Figure 6D The results showed that, with the number of antibodies remaining constant, the higher the affinity coefficient between the PD1 antibody and the target, the higher the efficiency of the binding between the PD1 antibody and the target cells.

[0211] Figure 7The diagram illustrates the prediction of each patient's sensitivity to the immune checkpoint inhibitor (ICB) PD1 antibody based on single-cell data from renal cell carcinoma patients (Kevin et al., Cancer Cell, 2021), and compares the consistency between the predicted and actual results. R represents patient sensitivity to ICB in clinical treatment, and NR represents patient resistance to ICB in clinical treatment. The results divide each patient's single-cell data into a control group and a treatment group. The control group only calculates cell-cell interaction strength, while the treatment group simulates the addition of PD1 antibody to the above dataset, with each patient receiving 50 units of PD1 antibody. Since the pharmacological mechanism of ICB is to activate exhausted T cells and promote the infiltration of non-exhausted T cells into tumor cells, after completing the simulation process described in the embodiments of this specification, it is observed whether the antibody, after capturing exhausted T cells, can promote the immune infiltration of non-exhausted T cells into tumor cells. The interaction count (TMI) between non-PD1-expressing T cells and tumor cells in the treatment group is counted. pert ), and the number of interactions between T cells that do not express PD1 and tumor cells in the control group (TMI). Naive ), and calculate TMI pert Compared to TMI Naive The change factor (TTA).

[0212]

[0213] If the TTA is greater than 0, it indicates that after simulating the addition of a specific dose of PD1 antibody, the immune infiltration of tumor cells by non-antibody-target T cells is enhanced, suggesting that the patient may be a potential ICB-sensitive patient. If the TTA is less than 0, it indicates that after simulating the addition of a specific dose of PD1 antibody, the enhancement of immune infiltration of tumor cells by non-antibody-target T cells is weakened, suggesting that the patient may be a potential ICB-resistant patient. The results show that the TTA in real ICB-sensitive renal cell carcinoma patients is much higher than the TTA in real resistant patients, demonstrating that the examples in this specification can better simulate real clinical results.

[0214] Figure 8A and Figure 8B The diagram illustrates the prediction of each patient's sensitivity to the immune checkpoint inhibitor (ICB) PD1 antibody based on single-cell data from clinical breast cancer patients (Ayse et al., Nature Medicine, 2021), and compares the consistency between the predicted and actual results. Each patient was simulated with 10 different doses of PD1 antibody (equal to 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, and 100% of the total cell count). Each antibody had a target affinity weighting coefficient of 8000, a virtual antibody abundance of 1, and no upper limit on the number of antibodies bound per cell. Figure 8A The results showed that TTA was significantly higher in breast cancer patients who were truly ICB-sensitive than in those who were truly resistant. Figure 8B The results showed that the TTA index calculated using the method described in this specification had reliable predictive performance (AUC = 0.88) in predicting the sensitivity of breast cancer patients to ICB therapy, and was highly consistent with clinical results.

[0215] Figure 9 The diagram illustrates the evaluation of potential new immunotherapy targets for an ICB-resistant breast cancer patient (BIOKEY_8) from the aforementioned breast cancer dataset (Ayse et al., Nature Medicine, 2021) according to an embodiment of this specification. Single-cell data from this patient were divided into a control group and a treatment group. The control group only calculated cell-cell interactions, while the treatment group simulated antibodies targeting 1066 targets. After completing the aforementioned simulation process of this embodiment, the TTA of the 1066 target antibodies was calculated and ranked. The results are as follows... Figure 9 As shown, the top 5 targets and 8 typical tumor immunotherapy targets are marked. The highest-ranked TTA targets include CD44, which has been reported as a potential therapeutic target for various tumors in recent years. In addition, targets ERBB3 and ERBB2 also have high rankings, both of which are important therapeutic targets for breast cancer.

[0216] Figure 10A and Figure 10B The diagram illustrates an embodiment of this specification evaluating potential combination therapies that could reduce resistance to PD1 antibodies in all ICB-resistant breast cancer patients from the aforementioned breast cancer dataset (Ayse et al., Nature Medicine, 2021). Single-cell data from these patients were divided into a control group and a treatment group. The control group only had cell-cell interactions calculated, while the treatment group simulated the combination of 82 marketed antibody drugs targeting different targets with PD1 antibodies. After completing the simulation process described above, the TTAs of the 82 antibody-PD1 antibody combination therapies were calculated and ranked. The results are as follows: Figure 10A As shown, the reported PD-1 combination therapy targets ERBB2 and IL2RA / B are ranked, with both having a TTA greater than 0, especially ERBB2, which ranks highly. Figure 10B The study showed that some patients who were initially unresponsive to PD1 antibodies (BIOKEY_7, BIOKEY_17, and BIOKEY_30) exhibited significantly improved responsiveness after combination therapy. The simulation results largely align with clinical findings, and the high-scoring predictions can provide insights into potential new combination therapies.

[0217] Figure 11The diagram illustrates simulations of PD-L1 antibody monotherapy and PD-L1 combined with VEGF antibody therapy using a single-cell dataset of 160 hepatocellular carcinoma patients (Xue et al., Nature, 2022). The dosage of PD-L1 antibody added to each patient was 20% of the patient's total cell count. The affinity weighting coefficient between each antibody and the target was 20, and each cell bound a maximum of 10 antibodies. The results showed that PD-L1 antibody monotherapy for hepatocellular carcinoma was less effective (median TTA close to 0), while the combination of PD-L1 and VEGF antibodies was more effective, exhibiting a significant synergistic effect. These results are consistent with clinical findings.

[0218] Figures 12A-12C The diagram illustrates a CAR-T drug delivery simulation performed on a single-cell dataset (Abdelfattah et al., Nature Communications, 2022) from a glioma patient (patient ID: ndGBM-09) according to an embodiment of this specification. T cells from this patient sample were used as feature data for the simulated CAR-T. A virtual feature "CAR" was added to the feature data matrix, with EGFR as the target and CAR abundance equal to the maximum expression value of each cell feature. To verify the effect of IFNγR expression in glioma cells on the infiltration of tumor cells by CAR-T cells, three parallel samples with IFNγR expression levels of 10, 50, and 100 were manually set, and drug delivery simulations were performed for each sample. Figure 12A The results showed that when the IFNγR expression level in glioma cells was low, CAR-T cells hardly infiltrated the glioma cell region, but Figure 12B As shown, with the increase of IFNγR abundance in glioma cells, the number of CAR-T cells infiltrating the glioma cell region increased significantly. Figure 12C The results showed that the killing effect of CAR-T on immune cells was positively correlated with the level of IFNγR. This result indicates that the killing effect of CAR-T on glioma depends on the expression level of IFNγR in glioma cells. Changes in the expression level of IFNγR in glioma cells may be one of the mechanisms leading to CAR-T resistance. This result is consistent with the conclusions of previously reported literature.

[0219] like Figure 13 The diagram shown is a schematic diagram of a drug evaluation device according to an embodiment of this specification. The diagram illustrates a modular structure for implementing the above method through a computer program or software. The units and modules can be implemented by software, or by a general-purpose chip or a dedicated chip. The device specifically includes: a single-cell feature data acquisition unit 1301, a drug simulation unit 1302, an interaction strength calculation unit 1303, a difference calculation unit 1304, and an evaluation unit 1305.

[0220] The single-cell feature data acquisition unit 1301 is used to acquire feature data of a single cell.

[0221] The drug simulation unit 1302 is used for drug simulation to obtain the characteristic data and attribute data of the drug.

[0222] The interaction strength calculation unit 1303 is used to combine the characteristic data of the drug with the characteristic data of the single cell to calculate the interaction strength between the drug and the single cell;

[0223] The difference calculation unit 1304 is used to analyze the interaction strength by combining the biological direct attribute data to obtain the difference after the drug is added;

[0224] The evaluation unit 1305 is used to evaluate the drug based on the difference.

[0225] The drug evaluation device described in the embodiments of this specification simulates drug delivery efficiency, targeting, and biological activity through affinity-mediated cell-drug self-organized reactions. It offers advantages such as direct application from clinical samples, low cost, and rapid computation. Therefore, it has broad application prospects in clinical diagnosis and treatment, including drug recommendation, drug target recommendation, virtual drug design, and efficacy evaluation, and is expected to significantly shorten drug development cycles, save development costs, and reduce development risks.

[0226] like Figure 14 The diagram illustrates the structure of a computing device according to an embodiment of this specification. The methods described in the above embodiments can all be run on the computing device of this embodiment. The computing device 1402 may include one or more processing devices 1404, such as one or more central processing units (CPUs), each of which can implement one or more hardware threads. The computing device 1402 may also include any storage resource 1406 for storing any kind of information, such as code, settings, data, etc. Without limitation, for example, the storage resource 1406 may include any one or more combinations of the following: any type of RAM, any type of ROM, flash memory, hard disk, optical disk, etc. More generally, any storage resource can use any technology to store information. Furthermore, any storage resource can provide volatile or non-volatile retention of information. Furthermore, any storage resource can represent a fixed or removable component of the computing device 1402. In one case, when the processing device 1404 executes associated instructions stored in any storage resource or combination of storage resources, the computing device 1402 can perform any operation of the associated instructions. The computing device 1402 also includes one or more drive mechanisms 1408 for interacting with any storage resources, such as hard disk drive mechanisms, optical disk drive mechanisms, etc.

[0227] The computing device 1402 may also include an input / output module 1410 (I / O) for receiving various inputs (via input device 1412) and providing various outputs (via output device 1414). A specific output mechanism may include a presentation device 1416 and an associated graphical user interface (GUI) 1418. The computing device 1402 may also include one or more network interfaces 1420 for exchanging data with other devices via one or more communication links 1422. One or more communication buses 1424 couple the components described above together.

[0228] Communication link 1422 can be implemented in any way, such as via a local area network, a wide area network (e.g., the Internet), a point-to-point connection, or any combination thereof. Communication link 1422 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.

[0229] This specification also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the above steps.

[0230] The computer device provided in the embodiments of this specification can also implement the methods of the foregoing embodiments of this specification.

[0231] This specification also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the above-described method.

[0232] This specification also provides computer-readable instructions, wherein when a processor executes the instructions, the program therein causes the processor to perform the methods as described in the foregoing embodiments of this specification.

[0233] It should be understood that in the various embodiments of this specification, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this specification.

[0234] It should also be understood that, in the embodiments of this specification, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this specification generally indicates that the preceding and following related objects have an "or" relationship.

[0235] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this specification can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this specification.

[0236] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0237] In the several embodiments provided in this specification, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, devices, or units, or they may be electrical, mechanical, or other forms of connection.

[0238] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments described in this specification, depending on actual needs.

[0239] Furthermore, the functional units in the various embodiments of this specification can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0240] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this specification, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this specification. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0241] This specification uses specific embodiments to illustrate the principles and implementation methods of this specification. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this specification. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this specification. Therefore, the content of this specification should not be construed as a limitation of this specification.

Claims

1. A virtual drug evaluation method, characterized in that... include, Obtain feature data of single cells; Drug simulation is used to obtain the characteristic data and attribute data of the drug. The interaction strength between the drug and the single cell is calculated by combining the characteristic data of the drug with the characteristic data of the single cell. The interaction strength is analyzed by combining the property data of the drug to obtain the difference after the drug is added; The drug is evaluated based on the differences described; The calculation of the interaction strength between the drug and the single cell by combining the drug's characteristic data with the single cell's characteristic data further includes... The characteristic data of the single cells were used as a control group. The drug feature matrix formed by the feature data of the drug is concatenated with the single cell feature matrix formed by the feature data of the single cell to form drug group data. Calculate the cell-cell interaction strength matrix A in the control group. cc ;Calculate the cell-cell interaction strength matrix A in the drug group. cc Cell-drug interaction strength matrix A cb Drug-drug interaction strength matrix A bb .

2. The virtual drug evaluation method according to claim 1, characterized in that, Drug simulation, obtaining the characteristic data and attribute data of the drug, further includes, Perform non-cellular drug simulation and cellular drug simulation to obtain the characteristic data and attribute data of the non-cellular drug, and obtain the characteristic data and attribute data of the cellular drug.

3. The virtual drug evaluation method according to claim 2, characterized in that, The simulation of non-cellular drugs, which yields characteristic data and attribute data of the non-cellular drugs, further includes, The virtual features of the non-cellular drug are set to form a set of virtual features that serve as the feature data of the non-cellular drug; Based on the virtual characteristics of the non-cellular drug, the attribute data of the non-cellular drug are obtained.

4. The virtual drug evaluation method according to claim 3, characterized in that, The attribute data of the non-cellular drugs shall include at least one or a combination of the following: Targeting, affinity, concentration, size, saturation, and intrinsic activity; among these, In the background network of receptor-ligand interactions, the virtual feature set of the non-cellular drug and the binding relationship of the drug's target receptor-ligand are added to obtain the targeting of the non-cellular drug. Add affinity weighting coefficients to the receptor-ligand binding relationships between the virtual feature set and non-cellular drug targets; The concentration of the non-cellular drug is obtained by setting the number of non-cellular drug data and the abundance of the virtual feature set; The non-cellular drug is defined as a molecule with radius r that can rotate freely in space, and the size of the non-cellular drug is obtained. The upper limit of the amount of non-cellular drug that binds to the target cells of the non-cellular drug is set to obtain the saturation. Based on the effect of the non-cellular drug on the drug target, the intrinsic activity is obtained by altering the downstream phenotype and characteristics of the target corresponding to the non-cellular drug.

5. The virtual drug evaluation method according to claim 2, characterized in that, The cell-based drug simulation, which obtains the characteristic data and attribute data of the cell-based drug, further includes, Obtain characteristic data for each cell type of the cell-based drug; Based on the characteristics of the cell-based drug, the attribute data of the cell-based drug are obtained.

6. The virtual drug evaluation method according to claim 5, characterized in that, Obtaining the characteristic data of each cell type for the cell-based drug also includes, Set the abundance of the feature data.

7. The virtual drug evaluation method according to claim 5, characterized in that, The attribute data of the cell-based drugs shall include at least one or a combination of the following: Targeting, affinity, concentration, size, saturation, and intrinsic activity; among these, The targeting of the cell-based drug is obtained by matching the feature set in the real cell-based drug with the target of the cell-based drug. Based on the receptor-ligand binding relationship between the feature set and the cell drug target, the affinity weighting coefficient is obtained; The number of cell-based drug data points is set to obtain the concentration of the cell-based drug; The size of the cell-based drug is obtained based on its radius r. The upper limit of the cell-based drug binding to the target cells of the cell-based drug is set to obtain the saturation. Based on the effect of the cell-based drug on the drug target, the intrinsic activity is obtained by altering the downstream phenotype and characteristics of the target corresponding to the cell-based drug.

8. The virtual drug evaluation method according to claim 7, characterized in that, Obtaining the characteristic data of each cell type for the cell-based drug also includes, A virtual feature set is added to the feature data of each cell in the cell-based drug to form the feature data of the cell-based drug.

9. The virtual drug evaluation method according to claim 8, characterized in that, When the feature data of the cell-based drug includes a virtual feature set, obtaining the attribute data of the cell-based drug based on its features further includes, By adding the feature set of the cell-based drug to the background network of receptor-ligand interactions and the binding relationship between the drug's target receptor and ligand, the targeting of the cell-based drug can be obtained. Add affinity weighting coefficients to the receptor-ligand binding relationship between feature data and cell-based drug targets; The number of cell-based drug data points is set to obtain the concentration of the cell-based drug; The size of the cell-based drug is obtained based on its radius r. The upper limit of the cell-based drug binding to the target cells of the cell-based drug is set to obtain the saturation. Based on the effect of the cell-based drug on the drug target, the intrinsic activity is obtained by altering the downstream phenotype and characteristics of the target corresponding to the cell-based drug.

10. The virtual drug evaluation method according to claim 1, characterized in that, Before analyzing the interaction strength by combining the property data of the drug, the following steps are also included: The interaction strength is sparsified based on the drug's property data.

11. The virtual drug evaluation method according to claim 10, characterized in that, The sparsification of the interaction strength based on the drug's property data further includes, For the control group, only the top N cells with the highest interaction strength with each cell were retained, and the interaction strength of other cells was set to 0. For the drug group, in the cell-cell interaction strength matrix A cc In the middle, only the cell-cell interaction strength matrix A is retained. cc The interaction strength of the top N cells with the highest interaction strength with each cell is set to 0, and the interaction strength of other cell-cell interactions is set to 0. Cell-cell interactions with strength less than the maximum value of cell-drug interaction strength are also set to 0. In the cell-drug interaction strength matrix A cb In this process, based on the saturation of the drug, only the n cells with the highest interaction strength with each drug are retained, while the interaction strength of other cells-drugs is set to 0. When these n cells bind to m drugs, their interaction strength with other drugs will be set to 0. In the drug-drug interaction strength matrix A bb In this context, all elements are set to values ​​greater than or equal to 0.

12. The virtual drug evaluation method according to claim 11, characterized in that, The sparsification of the interaction strength based on the drug's property data further includes, In the drug group, in the drug-drug interaction strength matrix A bb In the process, only the interaction strengths of the top N drugs with the highest interaction strength in the drug-drug interaction strength matrix are retained, and the interaction strengths of other drug-drug interactions are set to 0. Cell-cell interaction strengths that are less than the maximum value of drug-drug interaction strengths are also set to 0. or Based on the saturation of the drugs, only the n drugs with the highest interaction strength between each drug are retained, while the interaction strength of other drugs is set to 0.

13. The virtual drug evaluation method according to claim 1, characterized in that, The interaction strength is analyzed by combining the property data of the drug, and the differences before and after the addition of the drug are further included. The interaction strength matrix A in the drug group cc A cb A bb By splicing the data and comparing the interaction strength matrices of the drug group and the control group, the differences before and after the addition of the drug can be obtained.

14. The virtual drug evaluation method according to claim 13, characterized in that, Analyzing the interaction strength using the drug's property data reveals that the differences before and after drug addition also include: Based on the interaction strength, the interaction network of the control group and the interaction network of the drug group are obtained; By comparing the interaction network of the control group and the interaction network of the drug group, the differences before and after the addition of the drug were obtained.

15. The virtual drug evaluation method according to claim 1, characterized in that, Analyzing the interaction strength using the drug's property data reveals that the differences before and after drug addition also include: Calculate the probability matrix of the interaction strength based on the interaction strength; Construct a dimensionality reduction loss function; The virtual spatial coordinates of cells or drugs in the control group and drug group are obtained using the probability matrix and the dimensionality reduction loss function; Based on the virtual space coordinates, the interaction network of the control group and the interaction network of the drug group are obtained; By comparing the interaction network of the control group and the interaction network of the drug group, the differences before and after the addition of the drug were obtained.

16. The virtual drug evaluation method according to claim 15, characterized in that, The construction of the dimensionality reduction loss function also includes, The size of the drug's attribute data is added as a constraint to the dimensionality reduction loss function.

17. The virtual drug evaluation method according to claim 15, characterized in that, The construction of the dimensionality reduction loss function also includes, The density function in the dimensionality reduction loss function is determined based on the affinity in the drug's property data.

18. The virtual drug evaluation method according to claim 1, characterized in that, The drug is further included in the evaluation based on the aforementioned differences. Analyze the differences before and after drug addition to obtain the drug's pharmacological and toxicological potential.

19. The application of the virtual drug evaluation method according to any one of claims 1-18 in drug design.

20. The application of the virtual drug evaluation method according to any one of claims 1-18 in drug screening.

21. The application of the virtual drug evaluation method according to any one of claims 1-18 in drug target discovery.

Images