Simulation method and device based on intervention response and computer device

Abstract

The application relates to a simulation method and device based on intervention response and a computer device. The method comprises the following steps: receiving an intervention simulation request, wherein the intervention simulation request carries state information of a target object and intervention information; based on the state information and the intervention information, a search instruction corresponding to the target object is generated, and based on the search instruction, searching is performed in a stage information database and an intervention mechanism database to obtain a candidate search result; based on the candidate search result, the current search instruction is updated to obtain an updated search instruction; based on the updated search instruction, the step of searching based on the search instruction in the stage information database and the intervention mechanism database to obtain a candidate search result is re-executed until a preset simulation stop condition is met, and intervention response simulation information of the target object is output, thereby effectively improving the reliability and accuracy of a simulation result.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a simulation method, apparatus and computer equipment based on intervention response. Background Technology

[0002] With the continuous development of computer technology, digital applications in fields such as precision medicine, translational medicine, and drug development are becoming increasingly widespread. The demand for dynamic simulation and prediction of the physiological state evolution of virtual research subjects under external interventions such as drugs and surgery is also continuously increasing.

[0003] In related technologies, when simulating and predicting virtual research objects, a single retrieval based on a knowledge base is usually performed. The simulation results are obtained immediately based on the user's query conditions and then directly output. In-depth simulation and deduction cannot be performed, and the accuracy of the intervention response simulation results is low. Summary of the Invention

[0004] Therefore, it is necessary to provide an intervention response-based simulation method, apparatus, and computer equipment that can improve the accuracy of intervention response simulation in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a simulation method based on intervention response, including:

[0006] Receive an intervention simulation request, the intervention simulation request carrying the state information of the target object and intervention information;

[0007] Based on the status information and the intervention information, a retrieval instruction corresponding to the target object is generated, and a retrieval is performed in the stage information database and the intervention mechanism database based on the retrieval instruction to obtain candidate retrieval results;

[0008] Based on the candidate search results, the current search instruction is updated to obtain the updated search instruction;

[0009] Based on the updated retrieval instructions, the steps of re-executing the stage information database and intervention mechanism database based on the retrieval instructions to obtain candidate retrieval results are repeated until the preset simulation stop condition is met, and the intervention response simulation information of the target object is output.

[0010] In one embodiment, the method further includes:

[0011] Obtain the indicator parameters from the status information in the candidate search results;

[0012] When the indicator parameter exceeds the preset threshold corresponding to the significant event, the current simulation inference link of the target object is verified and a simulation verification result is generated. The simulation inference link includes the process of the target object being inferred and updated from the initial state information to the current state information.

[0013] If the simulation verification results meet the rollback triggering rules, a target rollback node is determined among each historical state node, and the state information of the target object is rolled back to the target rollback node.

[0014] In one embodiment, verifying the current simulation projection link of the target object and generating simulation verification results includes:

[0015] Based on the index parameters that exceed the preset threshold of significant events, related state nodes are extracted from the simulation and inference link, and key causal paths are constructed according to the order of the related state nodes.

[0016] Based on the key causal path, a re-search is performed in the stage information database and the intervention mechanism database to obtain multi-source evidence;

[0017] The evidence information from the multi-source evidence is compared with the historical evidence in the simulation and deduction link to determine the consistency result;

[0018] The causal relationships corresponding to the simulation and deduction links are compared with the mechanism relationships in the intervention mechanism database to determine the mechanism conflict results;

[0019] Based on the consistency results and the conflict results of the mechanism, the confidence level of the simulation inference link is evaluated, and simulation verification results are generated.

[0020] In one embodiment, the stage information database stores multiple stage triples corresponding to historical research subjects. Each stage triple includes the pre-intervention state, intervention operation, and post-intervention state of each historical research subject. The step of searching the stage information database and the intervention mechanism database based on the search command to obtain candidate search results includes:

[0021] The search command is matched with the pre-intervention state and intervention operation in the stage triplet, and candidate stage triplets are determined based on the matching results;

[0022] Based on the search instructions, a search is performed in the intervention mechanism database to determine candidate intervention mechanism information;

[0023] The candidate stage triples and the candidate intervention mechanism information are determined as candidate retrieval results.

[0024] In one embodiment, the candidate search results include multiple data groups, each data group including status information and / or intervention mechanism information. The step of updating the current search instruction based on the candidate search results to obtain an updated search instruction includes:

[0025] Based on the simulation task requirement data, the data groups in the candidate search results are sorted and the weight of each data group is determined.

[0026] By combining the simulation task objective and the weights of each data group, the state change information of the target object is determined;

[0027] Based on the state change information, the current search instruction is updated to obtain the updated search instruction.

[0028] In one embodiment, updating the current search instruction based on the state change information to obtain the updated search instruction includes:

[0029] Obtain current simulation association information, which includes at least one of simulation task requirement data, simulation task objectives, simulation inference links, and current status information;

[0030] The simulation association information, the state change information, and the current search instruction are fused together to obtain an updated search instruction. The fusion process includes at least one of the following: generation based on template mapping, generation through rule combination, or generation based on model.

[0031] In one embodiment, before the step of performing a search based on the search instruction to obtain candidate search results, the method further includes:

[0032] Obtain historical data of the subjects to be studied, including clinical records, experimental records, and treatment operation records;

[0033] Based on the time sequence, the historical data of the object under study is divided into multiple time periods;

[0034] For each of the defined time periods, the historical data of the subject under study is parsed to extract the stage triples for that time period. The stage triples include the pre-intervention state of the subject under study at the beginning of the time period, the intervention operation performed during the time period, and the post-intervention state at the end of the time period.

[0035] The stage triples are structured and then stored in the stage information database.

[0036] In one embodiment, the intervention response simulation information includes the state information of the target object when the preset simulation stop condition is met, the state information sequence corresponding to each round of multiple iterations, and the intervention information sequence.

[0037] Secondly, this application also provides a simulation device based on intervention response, comprising:

[0038] The receiving module is used to receive intervention simulation requests, which carry the status information of the target object and intervention information.

[0039] The first generation module is used to generate a retrieval instruction corresponding to the target object based on the status information and the intervention information, and to perform a retrieval in the stage information database and the intervention mechanism database based on the retrieval instruction to obtain candidate retrieval results;

[0040] The update module is used to update the current search instruction based on the candidate search results to obtain the updated search instruction;

[0041] The first determining module is used to re-execute the step of searching in the stage information database and intervention mechanism database based on the updated search instruction to obtain candidate search results, until the preset simulation stop condition is met, and output the intervention response simulation information of the target object.

[0042] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0043] Receive an intervention simulation request, the intervention simulation request carrying the state information of the target object and intervention information;

[0044] Based on the status information and the intervention information, a retrieval instruction corresponding to the target object is generated, and a retrieval is performed in the stage information database and the intervention mechanism database based on the retrieval instruction to obtain candidate retrieval results;

[0045] Based on the candidate search results, the current search instruction is updated to obtain the updated search instruction;

[0046] Based on the updated retrieval instructions, the steps of re-executing the stage information database and intervention mechanism database based on the retrieval instructions to obtain candidate retrieval results are repeated until the preset simulation stop condition is met, and the intervention response simulation information of the target object is output.

[0047] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0048] Receive an intervention simulation request, the intervention simulation request carrying the state information of the target object and intervention information;

[0049] Based on the status information and the intervention information, a retrieval instruction corresponding to the target object is generated, and a retrieval is performed in the stage information database and the intervention mechanism database based on the retrieval instruction to obtain candidate retrieval results;

[0050] Based on the candidate search results, the current search instruction is updated to obtain the updated search instruction;

[0051] Based on the updated retrieval instructions, the steps of re-executing the stage information database and intervention mechanism database based on the retrieval instructions to obtain candidate retrieval results are repeated until the preset simulation stop condition is met, and the intervention response simulation information of the target object is output.

[0052] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0053] Receive an intervention simulation request, the intervention simulation request carrying the state information of the target object and intervention information;

[0054] Based on the status information and the intervention information, a retrieval instruction corresponding to the target object is generated, and a retrieval is performed in the stage information database and the intervention mechanism database based on the retrieval instruction to obtain candidate retrieval results;

[0055] Based on the candidate search results, the current search instruction is updated to obtain the updated search instruction;

[0056] Based on the updated retrieval instructions, the steps of re-executing the stage information database and intervention mechanism database based on the retrieval instructions to obtain candidate retrieval results are repeated until the preset simulation stop condition is met, and the intervention response simulation information of the target object is output.

[0057] The aforementioned simulation method, apparatus, and computer equipment based on intervention response can first receive an intervention simulation request, which carries the state information and intervention information of the target object. Then, based on the state and intervention information, a retrieval instruction corresponding to the target object can be generated. A search is then performed in a stage information database and an intervention mechanism database based on the retrieval instruction to obtain candidate search results. Based on the candidate search results, the current retrieval instruction is updated to obtain an updated retrieval instruction. Then, based on the updated retrieval instruction, the steps of searching in the stage information database and intervention mechanism database again based on the retrieval instruction to obtain candidate search results are repeated until a preset simulation stop condition is met, at which point the intervention response simulation information of the target object is output. Thus, by receiving simulation requests carrying state and intervention information, generating retrieval instructions accordingly, searching in the stage information database and intervention mechanism database, and iteratively updating the retrieval instructions based on the search results, the retrieval and instruction update operations are repeatedly executed until the stop condition is met, and then the simulation results are output. This achieves continuous and dynamic simulation and deduction of the state evolution of the target object under intervention, effectively improving the reliability and accuracy of the intervention simulation results. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart illustrating a simulation method based on intervention response in one embodiment;

[0060] Figure 2 This is a flowchart illustrating the verification of the simulation process in one embodiment;

[0061] Figure 3 This is a schematic diagram illustrating the process of updating a search command in one embodiment;

[0062] Figure 4 This is a flowchart illustrating the process of constructing a phase information database in another embodiment;

[0063] Figure 5 This is a structural block diagram of a simulation device based on intervention response in one embodiment;

[0064] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0066] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0067] In one embodiment, such as Figure 1 As shown, a simulation method based on intervention response is provided. This embodiment illustrates the application of this method to a terminal. It can be understood that this method can also be applied to a server, and also to a system including a terminal and a server, and implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0068] Step 102: Receive intervention simulation request. The intervention simulation request carries the status information of the target object and intervention information.

[0069] The intervention simulation request can be a request used to trigger a virtual research subject to perform an intervention simulation operation. It can be input or manually triggered by users such as researchers and clinicians through human-computer interaction interfaces, terminals, or other devices. When the terminal receives the operation, it can receive the intervention simulation request, parse the intervention simulation request, apply corresponding intervention measures to the target object, and perform simulation deduction to obtain the possible results under the action of the intervention information.

[0070] It is understood that the target object in this application is a virtual simulation object, which can be of various types, such as human body, animal, tissue, organ, cell, organoid, chromosome, gene locus or other affected object, etc. This application does not limit this.

[0071] Furthermore, the state information of virtual simulation objects can be represented in various forms, such as one-dimensional vectors, feature maps, structured text, or multi-layered state matrices containing relationships. Alternatively, the state information can be layered according to its state structure, such as intervention and exposure layer states, molecular and regulatory layer states, cellular and microstructural layer states, tissue, organ, and system layer states, and phenotypic and functional structural layer states.

[0072] The intervention and exposure layer status can include the intervention input status, the exposure status of the drug, compound or other active agent, the intervention intensity status, the distribution, delivery, diffusion or localization status, the clearance, degradation, inactivation or removal status, the target binding status, the occupation of the action area or the action interface association status, etc.

[0073] Molecular and regulatory states can include key biomarker states, gene expression states, transcriptional, translational or epigenetic regulation states, enzyme activity states, receptor activation states, signaling pathway activation or inhibition states, metabolic load, toxic load or stress load states, nucleic acid editing, gene regulation or chromosome structural changes states, etc.

[0074] Cellular and microstructural states can include cell activity, cell proliferation, differentiation, apoptosis or necrosis, organelle function, cell population composition, microenvironment, and tissue microstructure or subcellular structure.

[0075] The state of tissues, organs, and systems can include the state of tissue integrity, the state of organ function, the state of local damage or repair, the state of circulation, respiration, metabolism, neural regulation, immune regulation or endocrine regulation, the state of multi-system coupling, and the state of homeostasis maintenance or imbalance.

[0076] Phenotypic and functional outcome states may include beneficial response states, unexpected response states, symptom, sign, or behavioral states, structural alteration states, functional improvement or deterioration states, risk level states, survival states, end-stage states, or other key outcome states. This application does not limit these.

[0077] In addition, intervention information can be of various types, such as the name, category, number or identification information of the intervention, the composition, structure, description or target information of the intervention, the known or candidate target, the site of action, the pathway or region of action, the intensity, frequency, route, duration or timing of the intervention, information on single or combined interventions, and information on metabolic, transport, diffusion, editing, regulation, repair, stimulation or other processes related to the intervention. This application does not limit the types of intervention information.

[0078] Optionally, the intervention simulation request may also include basic information about the target object and information about control variables.

[0079] The basic information of the target object may include at least one of the following: the type, source, number or classification information of the target object; the basic status, stage status, developmental status, disease course status or treatment stage information of the target object; the compositional characteristics, structural characteristics, functional characteristics or abnormal characteristics of the target object; the functional status, tissue status, cellular status, molecular status or genetic status information of organs related to the target object; genetic background, metabolic characteristics, immune characteristics, expression characteristics or regulatory characteristics; previous intervention history, exposure history, treatment history or historical status migration information; special object characteristics, including special populations, special animal models, special cell lines, organoid models, specific chromosomal regions or specific gene loci, etc. This application does not limit these.

[0080] In addition, the information on control variables can be of various kinds, such as environmental conditions, time conditions, circadian rhythms or observation sequence, dietary status, nutritional status or culture conditions, parallel intervention measures or combined treatment measures, external exposure conditions, stress conditions or experimental conditions, observation period, sampling period or simulation period, key target response type, risk event type or state change indicator, etc. This application does not limit it.

[0081] Therefore, in this embodiment of the application, the terminal can parse the received intervention simulation request to determine the state information of the target object and the intervention information to be applied to the target object carried in the request, so as to perform subsequent simulation and deduction of the target object.

[0082] Step 104: Based on the status information and intervention information, generate the retrieval instruction corresponding to the target object, and perform a retrieval in the stage information database and intervention mechanism database based on the retrieval instruction to obtain candidate retrieval results.

[0083] There are several ways to generate retrieval instructions. For example, state information and intervention information can be generated through template mapping to generate retrieval instructions corresponding to the target object; state information and intervention information can be combined through rule combination to generate corresponding retrieval instructions; state information and intervention information can be processed with the help of large language models, deep learning models, etc., to generate corresponding retrieval instructions; or state information and intervention information can be processed through candidate expansion methods such as external knowledge graphs or thesaurus expansion to generate corresponding retrieval instructions.

[0084] This application does not limit the use of any one of the above methods or a combination of multiple methods.

[0085] In addition, the phase information database can store information before and after various interventions, such as medical records or experimental records corresponding to multiple historical research subjects, and this application does not limit this.

[0086] Furthermore, the data sources for intervention mechanism databases can be diverse, including pharmacological mechanism databases, toxicological mechanism databases, compound databases, databases related to metabolism, transport, distribution, or clearance, drug instruction manual databases, operating procedure databases, or intervention instruction databases, molecular pathway knowledge bases, gene regulation knowledge bases, signaling network knowledge bases, organ medicine knowledge bases, physiological and pathological mechanism knowledge bases, cellular process mechanism knowledge bases, surgical mechanism databases, physical therapy mechanism databases, gene editing mechanism databases, molecular regulatory mechanism databases, literature evidence, research abstracts, experimental conclusions, or other mechanistic data. Therefore, intervention mechanism databases store the theoretical knowledge and objective laws of action corresponding to various intervention mechanisms. Information on various mechanism principles and their sources can be retrieved from intervention mechanism databases, and this application does not limit this.

[0087] Optionally, the candidate search results may include only candidate stage information retrieved from the stage information database, or only candidate intervention mechanism information retrieved from the intervention mechanism database, or may include both candidate stage information and candidate intervention mechanism information. This application does not limit this.

[0088] Optionally, the stage information database may also store multiple stage triples corresponding to historical research subjects. Each stage triple can include the pre-intervention state, intervention action, and post-intervention state of each historical research subject, and each stage triple can correspond to one historical research record. Therefore, in this embodiment, a search command can be matched with the pre-intervention state and intervention action in the stage triples. Based on the matching results, candidate stage triples are determined. Then, based on the search command, a search is performed in the intervention mechanism database to determine candidate intervention mechanism information. Finally, the candidate stage triples and candidate intervention mechanism information are determined as candidate search results.

[0089] Specifically, semantic similarity can be calculated between the search command and the pre-intervention state and intervention operation in the stage triplet, and candidate stage triplets can be determined based on the semantic similarity results. For example, the stage triplet with the highest semantic similarity result can be determined as a candidate stage triplet, or the stage triplet with a semantic similarity result greater than a preset threshold can be determined as a candidate stage triplet, or the semantic similarity results can be arranged in descending order, and the top N stage triplets can be determined as candidate stage triplets, etc. This application does not limit this.

[0090] Subsequently, semantic similarity can be calculated based on the search instructions and the intervention mechanism information stored in the intervention mechanism database. Candidate intervention mechanism information is then determined based on the semantic similarity results. It is understood that the candidate intervention mechanism information retrieved from the intervention mechanism database can be used to characterize the interaction mechanism between the intervention information and the target object, such as the absorption, distribution, metabolism, and excretion patterns of the intervention factor, as well as target binding kinetic parameters, or it can be information related to various specific interventions. This candidate intervention mechanism information can be one or multiple pieces, and this application does not limit this.

[0091] Therefore, in this embodiment of the application, after determining the retrieval instruction, a retrieval can be performed in the stage information database to obtain candidate stage triples, and candidate intervention mechanism information can be obtained by retrieving from the intervention mechanism database. The candidate stage triples and the candidate intervention mechanism information together constitute the candidate retrieval results. The candidate retrieval results can be used to characterize the possible changes in the effect and the effect of the target object based on the intervention mechanism information, etc. This application does not limit this.

[0092] Step 106: Based on the candidate search results, update the current search instruction to obtain the updated search instruction.

[0093] It is understandable that candidate search results can include potential state change information, evolutionary history, etc. of the target object. Therefore, the state information in the search instruction can be updated based on the state information in the candidate search results, and the intervention information in the search instruction can be updated based on the intervention mechanism information in the candidate search results, generating an updated search instruction. This search instruction can be used for the next round of intervention simulation.

[0094] For example, if candidate search results indicate a decrease in the activity of a metabolic enzyme and a deterioration in liver condition, the generated search requests could include those for liver risk mechanisms, metabolic regulation inhibition mechanisms, combined intervention interactions, and gastrointestinal reaction mechanisms. As another example, when the target is a cell model, if candidate search results indicate a decrease in the activity of the target pathway but a simultaneous increase in stress markers, the generated search requests could include those for cellular stress mechanisms, mitochondrial dysfunction mechanisms, pathway compensation activation mechanisms, and unexpected cell fate change mechanisms. Furthermore, when the target is a gene editing model, if candidate search results indicate an increased success rate of target site editing but an increased risk of structural abnormalities, the generated search requests could include those for editing efficiency mechanisms, off-target mechanisms, DNA damage response mechanisms, and chromosome structural change mechanisms.

[0095] Alternatively, if the current simulation round is high-risk, search requests related to risk, abnormal structural changes, unexpected responses, and similar high-risk stage triples can be generated first; if the current simulation round is target response priority, search requests related to disease staging, target sites, key pathways, and changes in response indicators can be generated first.

[0096] It should be noted that the above examples are merely illustrative and should not be construed as limiting the methods for updating search requests in the embodiments of this application.

[0097] It is understood that the state information of the target object can also be called state element. Based on the candidate retrieval results and combined with the current state information of the target object, the state information of the target object itself can be updated. The update range can be one or more state elements in a single state level, or it can involve updating multiple state levels at the same time, etc. This application does not limit this.

[0098] Understandably, if the candidate search results include new state nodes, new organ reaction chains, new toxicological pathways, or new disease progression features, then corresponding state elements can be added to the target object. If the numerical change of a certain state element is small over multiple rounds, has low relevance to the current simulation task, or contributes little to the simulation task objective, it can be downgraded or deactivated. If subsequent candidate search results show that the state element is relevant again, it can be reactivated. Thus, state elements can be dynamically managed based on one or more factors such as relevance, change magnitude, significant event triggering, evidence weight, macroscopic disease progression matching strength, and resource conditions, thereby enabling the addition, deletion, downgrading, deactivation, or reactivation of state elements. This application does not limit this.

[0099] Step 108: Based on the updated retrieval instructions, re-execute the steps of searching in the stage information database and intervention mechanism database according to the retrieval instructions to obtain candidate retrieval results until the preset simulation stop condition is met, and output the intervention response simulation information of the target object.

[0100] The preset simulation stopping conditions can be various, such as the achievement of the target response, the recovery of key functions to a stable state, the achievement of a stable platform in a risky state, the evolution of an abnormal state to a stable end, the triggering of a significant end event, the end of the simulation cycle, or the state changes being lower than a preset threshold for several consecutive rounds. This application does not limit these conditions.

[0101] Furthermore, the intervention response simulation information can take many forms, such as the state information of the target object when the preset simulation stopping condition is met, the state information sequence corresponding to multiple iterations in each round, and the intervention information sequence. The state information sequence can be understood as a sequence obtained by arranging the state information from multiple iterations in each round of simulation in chronological order; it can also be called a state trajectory. The intervention information sequence can be the intervention effect during multiple iterations in each round of simulation, arranged in chronological order. This application does not limit this.

[0102] Optionally, the intervention response simulation information may also include significant event state trajectories, structural or functional state change results, evidence support chains, rollback and error correction records, uncertainty descriptions, risk warnings, etc., which are not limited in this application.

[0103] Therefore, in this embodiment, after receiving the updated search instruction, a new search can be performed in the stage information database and intervention mechanism database based on the updated search instruction to obtain candidate search results for the next simulation evolution stage. After each round of simulation iteration or state update, it can be determined whether the current state of the target object or the deduction link meets the preset simulation stop condition. If not, a newly generated search instruction is given for updating again, and the above steps are repeated. If the condition is met, the simulation is stopped, and the intervention response simulation information corresponding to the target is output. Thus, through iterative updates of search instructions and cyclic retrieval, multi-stage continuous simulation deduction of the target object under intervention is realized. The entire process of state evolution under intervention can be completely simulated, thereby significantly improving the accuracy and reliability of intervention response simulation, as well as the continuity, controllability, realism, and interpretability of drug response analysis, providing an efficient and reliable foundation for scenarios such as drug development and treatment plan evaluation.

[0104] In the aforementioned simulation method based on intervention response, an intervention simulation request is first received, carrying the target object's state information and intervention information. Then, based on the state and intervention information, a retrieval instruction corresponding to the target object is generated. A search is then performed in the stage information database and the intervention mechanism database based on the retrieval instruction to obtain candidate search results. Based on the candidate search results, the current retrieval instruction is updated to obtain an updated retrieval instruction. Then, based on the updated retrieval instruction, the steps of searching in the stage information database and the intervention mechanism database again based on the retrieval instruction to obtain candidate search results are repeated until a preset simulation stopping condition is met, at which point the intervention response simulation information of the target object is output. Thus, by receiving simulation requests carrying state and intervention information, generating retrieval instructions accordingly, searching in the stage information database and the intervention mechanism database, and iteratively updating the retrieval instructions based on the search results, the retrieval and instruction update operations are repeatedly executed until the stopping condition is met, at which point the simulation results are output. This achieves continuous and dynamic simulation and deduction of the target object's state evolution under intervention, effectively improving the reliability and accuracy of the intervention simulation results.

[0105] In one exemplary embodiment, the simulation process can also be verified during the simulation process, such as... Figure 2 As shown, the simulation method based on intervention response includes steps 202 to 206. Wherein:

[0106] Step 202: Obtain the indicator parameters from the status information in the candidate search results.

[0107] Since the candidate search results are determined based on search instructions within the stage information database and intervention mechanism database, they may include corresponding status information and status change information. The terminal can parse and process the status information in the candidate search results to determine the indicator parameters within that status information.

[0108] The indicator parameters can be quantitative values ​​or state representations reflecting the physiological, pathological, physical, or biochemical characteristics of the target object. Examples include drug concentration in the blood, concentration of specific proteins such as the liver function indicator alanine aminotransferase (ALT), target occupancy rate, signaling pathway activation level, apoptosis rate, tissue damage percentage, or specific phenotypic scores. This application does not limit these parameters.

[0109] Step 204: When the index parameter exceeds the preset threshold corresponding to the significant event, the current simulation deduction link of the target object is verified and the simulation verification result is generated. The simulation deduction link includes the process of the target object being deduced and updated from the initial state information to the current state information.

[0110] Significant events can be pre-defined key physiological, pathological, or safety events, such as severe drug toxicity reactions, target oversaturation, or physiological homeostasis imbalance. The preset thresholds corresponding to these significant events can be used as boundaries for judging abnormal or critical turning points. When a certain indicator parameter exceeds the preset threshold corresponding to a significant event, it indicates that there may be an abnormal outcome or potential logical deviation in the current simulation process. Therefore, a global tracing and verification can be performed on the entire simulation chain of the target object, from its initial state through multiple iterations to the current state, generating simulation verification results to check whether logical breaks or irrationalities have occurred in certain steps of the simulation process.

[0111] Optionally, the specific steps for verifying the simulation projection link can be as follows: First, based on the index parameters that exceed the preset threshold of significant events, extract the associated state nodes from the simulation projection link, and construct the key causal path according to the order between the associated state nodes. Then, based on the key causal path, re-search in the stage information database and the intervention mechanism database to obtain multi-source evidence. After that, compare the evidence information in the multi-source evidence with the historical evidence in the simulation projection link to determine the consistency result. Then, compare the causal relationship corresponding to the simulation projection link with the mechanism relationship in the intervention mechanism database to determine the mechanism conflict result. Based on the consistency result and the mechanism conflict result, evaluate the confidence of the simulation projection link and generate the simulation verification result.

[0112] Among them, the associated state nodes can be historical state nodes that are directly or indirectly associated with index parameters that exceed the preset threshold of significant events in the simulation and deduction link. The key causal path can be a state evolution link formed by connecting the associated state nodes in chronological order and causal relationship. Multi-source evidence can refer to similar historical cases retrieved from the stage information database for each node in the key causal path, i.e., empirical evidence, as well as intervention mechanism information such as pharmacology, toxicology and other literature principles retrieved from the intervention mechanism database, i.e., theoretical evidence.

[0113] Understandably, we can first acquire indicator parameters that exceed a preset threshold for significant events, such as elevated ALT concentration, and then use this to trace back to identify other state variables that significantly contribute to changes in the value of this indicator parameter within the simulation chain. For example, we can identify neutrophil percentage, cumulative drug exposure, or specific receptor occupancy rates that are closely related to elevated ALT levels. Then, the specific state elements of the identified state variables can be designated as associated state nodes.

[0114] Subsequently, the extracted associated state nodes can be arranged sequentially according to the time sequence of simulation iterations, and directional connections between nodes can be established using state transition rules, preset system biological pathway topology relationships, or attention mechanism weights in deep neural networks. This will connect the discrete associated state nodes into a directional chain with temporal and causal evolution orientation, i.e., a key causal path, which is not limited in this application.

[0115] Subsequently, a search can be performed in the stage information database based on the constructed key causal pathway to determine whether there are historical clinical studies or experimental records with a similarity greater than a similarity threshold to that key causal pathway, serving as empirical evidence. Then, based on the key causal pathway, a match can be made in the intervention mechanism database. For example, based on whether oxidative stress leads to hepatocyte apoptosis, the corresponding pharmacological, toxicological, or chemical reaction mechanism principles can be retrieved from the intervention mechanism database as theoretical evidence. Finally, the retrieved historical clinical and experimental records, along with the corresponding principles and mechanisms, are combined to constitute multi-source evidence.

[0116] Subsequently, the literature and clinical case data relied upon by the simulation chain during each round of simulation iterations can be extracted and used as historical evidence. Newly retrieved multi-source evidence is then compared with this historical evidence. If the deviation of identical attribute values ​​or identical parameter values ​​is within a preset range and the semantic similarity is higher than a set threshold, the consistency result is determined to be high or successful. If the deviation of attribute values ​​or identical parameter values ​​exceeds a preset range or the conclusions deviate, the consistency result is determined to be low or unsuccessful.

[0117] Next, the logic between two or more adjacent related state nodes in the critical causal path can be converted into causal rules and causal relationships, such as state A increasing leading to state B decreasing. The principles controlling the interaction between state A and state B can be found in the intervention mechanism database. Then, the direction of the simulated causal relationship, positive and negative feedback logic, etc., can be compared with the standard intervention mechanism relationship in the database to see if they are the same as or violate each other. For example, if the standard intervention mechanism is A activating B, while the simulated causal relationship is A increasing leading to B being suppressed, then they can be considered opposite, indicating a logical conflict. Therefore, if the comparison finds that the causal relationship in the simulated link violates known rules or laws, then a logical conflict can be considered to exist, and the mechanism conflict result can be determined as "mechanism conflict exists," which may also include specific conflict nodes; if the comparison finds that the causal relationship in the simulated link is the same as or consistent with the standard intervention mechanism, then the logic can be considered correct, and the mechanism conflict result can be determined as "mechanism conflict does not exist."

[0118] If the consistency result is high and the mechanism conflict result is no mechanism conflict, then the confidence level of the simulation deduction link can be determined to be high, and the simulation verification result is successful. Further simulations can then be performed based on the current state node. If the consistency result is low, or the mechanism conflict result is a mechanism conflict, then the confidence level of the simulation deduction link can be determined to be low, and the simulation verification result is a verification failure. Further identification of conflict nodes and inconsistent nodes leading to the low confidence level, or identification of state nodes with low consistency results or mechanism conflict results, etc., is not limited in this application.

[0119] Step 206: If the simulation verification results meet the rollback triggering rules, determine the target rollback node among each historical state node, and roll back the state information of the target object to the target rollback node.

[0120] Among them, the rollback trigger rule can be a pre-set rule or condition used to determine whether the current simulation path needs to be terminated, which can be a simulation verification failure.

[0121] In addition, the target backtracking node can be a state node in the generated historical state node sequence that has real and reliable data and is not affected by logical conflicts. For example, it can be the previous stable state, the state before the previous significant event, the previous evidence bifurcation point, or other nodes. The above nodes can be determined by data analysis or by using a large language model that has been trained. This application does not limit this.

[0122] Optionally, in the case of simulation at the human or animal level, if it is determined that the function of a critical organ is rapidly deteriorating, but after re-searching it is found that the deterioration trend is supported by only a single low-quality piece of evidence and is inconsistent with the evolution direction of the stage tuple of similar target objects, it can be determined that the simulation verification result has failed. In this case, it can be rolled back to the target rollback node and the simulation inference path can be corrected.

[0123] In another specific scenario, when the target object is a cell-level simulation, if it is determined that the target pathway is significantly suppressed, but after re-searching, an unexpected stress pathway is abnormally activated and there is a clear conflict between the relevant evidence chains, it can be determined that the simulation verification result has failed. The simulation inference link can be marked as a low-confidence link and backtracked to the target backtracking node.

[0124] Alternatively, if the target is gene editing, and the editing of the target site is confirmed to be successful, but the evidence strength for off-target risk links is insufficient, alternative searches and confidence reassessments can be performed on off-target related inferences to avoid directly outputting unreliable conclusions.

[0125] Optionally, the above verification process can also be carried out with the help of a large language model. For example, the relevant information can be processed by a large language model to determine whether the evidence is consistent, whether the mechanism is conflicting, the confidence result, etc. This application does not limit this.

[0126] The above methods not only enable continuous state simulation, but also link-level verification of significant events. Furthermore, when low-confidence or conflicting inferences are discovered, backtracking and path correction can be performed, thereby improving the stability, interpretability, and credibility of the simulation results.

[0127] It is understood that in this application, the identification and rollback of significant events is not simply a matter of canceling actions. Since intervention simulation is essentially a dynamic prediction within an unknown search space, if the index parameters exceed the significance threshold, it can be assumed that the simulation process may have deviated or triggered abnormal evolution, potentially entering a stage of error distortion and out-of-bounds accumulation. Therefore, in this embodiment, upon detecting that the index parameters exceed the significance threshold, the simulation chain verification can be triggered, and the calculation state can be rolled back to a historical rollback node that has passed the prior constraint verification. This effectively avoids cascading deviations caused by disturbances in high-iteration-depth simulations, fundamentally ensuring the stability and global convergence of complex dynamic intervention simulation processes.

[0128] In this embodiment, by identifying target rollback nodes in each historical state node and rolling back the state information of the target object to the target rollback node, such as rolling back to the logical fork point before the significant event is triggered, the continuation of erroneous inferences is effectively avoided, the accumulation of invalid simulation data is avoided, and real-time tracing and dynamic correction of simulation inference paths are realized. By rebranching the inference, the logical consistency and scientific rationality of subsequent inference results can be ensured, thereby significantly improving the accuracy and reliability of intervention response simulation in complex environments.

[0129] Understandably, to ensure high-confidence results even when faced with complex and variable inputs during the intervention simulation process, this embodiment organically combines steps such as initial input state, data acquisition, intermediate simulation result calculation, significant event identification, link verification backoff, and final convergence determination to form a closed-loop long simulation link in this application. Since this long simulation link has complete logic, its specific execution sequence is not limited to the sequence presented in this embodiment and can be flexibly adjusted or parallelized according to the actual computing architecture. For example, while obtaining intermediate simulation results, the data source for the next iteration can be retrieved in parallel, or the triggering conditions for significant event identification and convergence determination can be merged and executed synchronously. Regardless of how the execution order of the steps is adjusted, the data closed loop and dynamic feedback effect formed on the overall topology can be maintained, all falling within the scope of the long simulation link.

[0130] In this embodiment, the indicator parameters in the status information of the candidate search results can be obtained first. When the indicator parameters exceed the preset threshold corresponding to the significant event, the current simulation inference link of the target object is verified, and a simulation verification result is generated. The simulation inference link includes the process of the target object being updated from the initial state information to the current state information. When the simulation verification result meets the rollback triggering rule, the target rollback node is determined in each historical state node, and the state information of the target object is rolled back to the target rollback node. Thus, during the simulation inference process, the state information can be detected, and when a significant event is triggered, the simulation verification result can be determined. When the rollback triggering rule is met, the state information of the target object is rolled back to that node. This allows for timely identification of abnormal situations in the simulation process, investigation of inference defects through link verification, and correction of erroneous paths through the rollback mechanism, preventing the continuous propagation of erroneous inferences and effectively improving the stability of long-term simulation operation and the accuracy and reliability of simulation results.

[0131] In one exemplary embodiment, such as Figure 3 As shown, step 106 includes steps 302 to 306. Wherein:

[0132] Step 302: Based on the simulation task requirement data, sort the data groups in the candidate search results and determine the weight of each data group.

[0133] The simulation task requirement data can be understood as additional constraint information that may affect the simulation process and simulation results, such as one or more of the following: environmental conditions, time conditions, observation period, culture conditions, parallel intervention measures, external exposure conditions, experimental conditions, etc. This application does not limit this.

[0134] In addition, the candidate search results can be the entire set of data retrieved from the stage information database and the intervention mechanism database. It can include multiple data groups, and each data group can include candidate triples and corresponding candidate intervention mechanism information, etc. This application does not limit this.

[0135] Furthermore, after obtaining the candidate search results, the data groups can be sorted according to the relevance between the simulation task requirements data and each data group. For example, the relevance can be determined by calculating the semantic matching degree, or it can be determined by neural network models, large language models, etc. This application does not limit this.

[0136] It is understandable that the greater the correlation between each data group in the candidate search results and the simulation task requirement data, the greater the weight of that data group and the higher its ranking; conversely, the smaller the correlation between the two, the smaller the weight of that data group and the lower its ranking.

[0137] For example, if the simulation task requires a focus on risk events, then data sets related to risks such as toxicity, side effects, and organ damage can be given higher weights and prioritized. If the simulation task requires a focus on target response, then data sets related to efficacy and improvement of target indicators can be given higher weights. If the simulation task requires a focus on mechanism explanation, then data sets related to pathways of action, molecular mechanisms, and regulatory networks can be given higher weights, and so on. This application does not impose any limitations on these aspects.

[0138] Therefore, in this embodiment of the application, the importance ranking of each data group can be dynamically adjusted based on the simulation task requirements, so that the subsequent state update process is more in line with user needs.

[0139] Step 304: Combine the simulation task objective and the weights of each data group to determine the state change information of the target object.

[0140] The simulation task objective can also be called simulation objective information or simulation objective outcome direction. It can characterize the expected final state or evaluation dimension of the simulation. For example, it can include task priorities such as prioritizing beneficial responses, prioritizing risk responses, prioritizing mechanism explanations, as well as output granularity requirements, state trajectory requirements, etc. This application does not limit these.

[0141] In addition, the state change information can be the state changes that the target object may undergo during the simulation process, which can be deduced based on the candidate retrieval results and the simulation task objective. It can include the state information that has changed, the direction of the change, and the specific value, etc. This application does not limit this.

[0142] Understandably, after determining the weights of each data group, it is possible to further combine the simulation task objectives to determine the potential state changes of the target object during the simulation process.

[0143] For example, the terminal can prioritize the numerical information contained in data groups with higher weights; or, it can combine the priorities in the simulation task objectives to further determine the beneficial parts that need to be retained or the risky parts that need to be avoided in state changes. For instance, if the simulation task objective prioritizes risk response, then the terminal can prioritize risk-related state changes such as elevated liver enzymes or worsening toxicity indicators, and determine the corresponding risk state change information, etc. This application does not limit this.

[0144] Step 306: Based on the state change information, update the current search instruction to obtain the updated search instruction.

[0145] Optionally, during the process of updating the current search command, the current state change information can be used to update or replace the corresponding parameters, keywords or search constraints in the search command without having to reconstruct a completely new search command from scratch, thereby effectively improving the efficiency of search command generation.

[0146] Optionally, incremental retrieval can be performed based on the current state change information, and updated retrieval instructions can be generated through synonym replacement, similar replacement, etc. This application does not limit this.

[0147] Therefore, in this embodiment of the application, the retrieval instruction can be updated based on the current state change, and a closed-loop logic of retrieval and update can be performed again. The updated retrieval instruction is used as the input for the next round of simulation, and a new round of retrieval is performed in the dual database, thereby improving the accuracy and reliability of the simulation link in terms of time axis and causal relationship.

[0148] It is understood that the simulation-related information may include at least one of the following: simulation task requirement data, simulation task objectives, simulation deduction links, and current status information. When updating the retrieval instruction, the simulation-related information can be obtained first, and then the simulation-related information, status change information, and the current instruction can be fused together to obtain the updated retrieval instruction. The fusion process includes at least one of the following: generation based on template mapping, generation through rule combination, or generation based on model.

[0149] In the process of fusion processing, any one of the above simulation-related information can be fused with the state change information and the current search instruction to obtain an updated search instruction; or multiple combinations of the above simulation-related information can be fused with the state change information and the current search instruction to obtain an updated search instruction. This application does not limit this.

[0150] In addition, multiple retrieval instruction templates for simulation scenarios can be pre-built. Each template can be set with one or more parameters to be filled. During fusion processing, simulation-related information and current state change information can be used as parameters to fill in the preset positions of the templates and combined with the current retrieval instructions to obtain the updated retrieval instructions.

[0151] Alternatively, by using preset rules and logic, conditions can be concatenated with simulation-related information, current state change information, and current search instructions, and filter words can be added or deleted or logical operators can be adjusted to obtain updated search instructions.

[0152] Alternatively, simulation-related information, current state change information, and current search instructions can be used as inputs, and a trained machine learning model or deep learning model can be used to process the data to obtain updated search instructions.

[0153] It should be noted that the above examples are merely illustrative and should not be construed as limiting the methods for generating updated search instructions in the embodiments of this application.

[0154] The following section explains the updating of the above search instructions using different simulation scenarios.

[0155] For example, if the simulation task aims to evaluate the efficacy and safety of anti-tumor intervention regimen A, and after the current simulation round, the obtained state change information indicates that the virtual target object has experienced a rapid increase in transaminase levels and hepatotoxicity risk indicators exceeding the safety threshold, the terminal can, based on this state change information and integrating the current simulation deduction chain, determine that further in-depth deduction is needed for this abnormal liver state. Then, based on rule combinations or template mapping, the current search command can be parameterized and appended to add content related to liver injury transmission mechanisms, idiosyncratic hepatotoxicity mechanisms, and drug-metabolic interactions. This generates an updated search command containing hepatotoxicity targets and metabolic enzyme inhibition pathways, allowing for dynamic acquisition of mechanistic evidence supporting the evolution of toxicity and risk correction in the next simulation search round.

[0156] Alternatively, if the simulation target is to evaluate the response of a specific cell population to intervention by a candidate compound, and the state change information obtained in the current simulation round indicates that although the activity of the target signaling pathway was successfully inhibited, it was accompanied by an abnormally high level of reactive oxygen species (ROS) and a decrease in mitochondrial membrane potential, indicating a tendency for cell damage. The terminal can fuse this state change information with the simulation task target through a large model to determine that the current state has entered an unexpected cellular stress state. The terminal can then update the current search instructions by introducing search keywords related to cellular stress damage, apoptosis cascades, and bypass compensation mechanisms to obtain updated search instructions targeting cell fate.

[0157] By integrating simulation-related information and state change information, the instruction upgrade mechanism in this embodiment ensures that each updated retrieval instruction carries a global perspective, guaranteeing that the deduction link can always be anchored on the predetermined logical trajectory and point to the final simulation target outcome. This effectively avoids the waste of simulation computing resources caused by retrieval instruction drift or invalid retrieval evidence.

[0158] Therefore, in this embodiment of the application, by integrating simulation-related information and state change information during the process of updating the retrieval instructions, each updated retrieval instruction can have a global perspective, ensuring that the inference link always leads to the final simulation task goal, thereby effectively avoiding the waste of computing resources caused by invalid evidence and improving the accuracy and reliability of the simulation.

[0159] In this embodiment, the data groups in the candidate search results are first sorted based on the simulation task requirement data to determine the weight of each data group. Then, the state change information of the target object is determined by combining the simulation task objective and the weight of each data group. Finally, the current search instruction is updated based on the state change information to obtain the updated search instruction. Therefore, by integrating the simulation task requirement data, the simulation task objective, and the candidate search results during the update process, the updated search instruction becomes more comprehensive and reliable. This ensures that the simulation chain always leads towards the final simulation task objective, effectively guaranteeing the orderly conduct of multiple rounds of recursive simulation and improving the accuracy and consistency of the simulation.

[0160] In one exemplary embodiment, such as Figure 4 As shown, a simulation method based on intervention response is provided, which includes steps 402 to 408. Wherein:

[0161] Step 402: Obtain the historical data of the subject of study. The historical data includes clinical records, experimental records, and treatment operation records.

[0162] The research object can be a virtual subject in the simulation process, which can be a virtual patient, animal model, cell group, organoid, tissue organ or specific biological system, etc. This application does not limit this.

[0163] In addition, the sources of historical data may include at least one of the following, such as human clinical record data, animal experiment record data, electronic medical record texts, inpatient progress notes, outpatient follow-up records, medical orders, medication records, treatment records, nursing records or other intervention records, laboratory data, testing data, examination data, monitoring data or sampling data, surgical records, interventional treatment records, physiotherapy records, rehabilitation records, gene editing records, cell processing operation records or other operation records, vital signs data, behavioral records, phenotypic records, pathological records or imaging records, experimental records of tissues, organs, cells, organoids, chromosomes, gene loci or other target objects, preclinical trial records, real-world research data or other longitudinal process data, and other data that can characterize the state transition relationship of the target object before and after intervention, etc., which are not limited in this application.

[0164] Understandably, after obtaining raw record data from multiple data sources, multi-source data cleaning and preprocessing can be performed. Preprocessing can include data format alignment, data deduplication, missing value detection and filling, unit standardization, etc., to obtain standardized historical record data with clear structure and complete time.

[0165] Step 404: Based on the time sequence, the historical data of the object under study is divided into multiple time periods.

[0166] In this context, a time segment refers to a historical record divided according to key clinical events, experimental changes, or observational nodes. There are various ways to divide a time segment, such as using a fixed duration or segmenting the historical record based on a preset time span. Alternatively, it can be segmented based on key nodes such as significant changes in state or shifts in intervention strategies. For example, if a historical record contains key nodes or related semantic keywords such as drug dosage adjustments, introduction of new drugs, commencement of dietary intervention, or the appearance of new pathological signs, these can be used as time segment boundaries to split the historical record into multiple time segments. Alternatively, multiple historical records of the same subject can be organized according to a state evolution timeline, and then divided into multiple time segments based on fixed time windows, key intervention events, significant indicator changes, structural change nodes, functional change nodes, treatment plan change nodes, or combinations thereof. This application does not limit this approach.

[0167] Step 406: For each time period, the historical data of the research subjects is analyzed to extract the stage triples for each time period. The stage triples include the pre-intervention state of the research subjects at the beginning of the time period, the intervention operation performed during the time period, and the post-intervention state at the end of the time period.

[0168] The phase triple can be used to characterize the causal logic transition within the time phase. It can include the pre-intervention state of the subject at the beginning of the current time phase, the intervention performed during the current time phase, and the post-intervention state at the end of the current time phase.

[0169] The pre-intervention state in the phase triad can include one or more of the following: basic attributes, structural state, functional state, molecular state, cellular state, organ state, systemic state, past exposure status, and current risk level of the target object. The intervention operation, also referred to as intervention action or intervention operation, can include one or more of the following: drug administration, compound treatment, combined intervention, dose or intensity adjustment, supportive therapy, surgery, intervention, physical therapy, rehabilitation, gene editing, nucleic acid regulation, cell treatment, discontinuation, replacement, or other action operations. The post-intervention state can include one or more of the following: molecular state, cellular state, tissue state, organ state, functional changes, structural changes, risk changes, target response changes, and end-stage trends in subsequent phases. This application does not limit this.

[0170] Furthermore, for each defined time stage, the terminal can perform semantic understanding on the text, tables, or time-series data within that time stage based on Natural Language Processing (NLP), Named Entity Recognition (NER), and large language models. This allows it to identify and extract physiological information, phenotypic information, or indicator values ​​corresponding to the start time of that time stage, serving as the pre-intervention state; extract the treatment plan, dosage, and experimental conditions implemented during that time stage, serving as the corresponding intervention operation; and extract the physiological information, phenotypic information, or indicator values ​​corresponding to the end time of that time stage, serving as the corresponding post-intervention state. Finally, the extracted pre-intervention state, intervention operation, and post-intervention state are combined and correlated to form a complete causal relationship chain, i.e., a stage triplet.

[0171] For example, by analyzing historical data, the resulting stage triplet could be: (Target subject in middle-aged or elderly cancer state, mild liver dysfunction, abnormal basal metabolism, under continuous anti-tumor intervention, combined with supportive treatment) - (Continuous intervention, intensity maintenance, accompanied by remission treatment) - (Mildly elevated liver enzymes, increased gastrointestinal discomfort, mild improvement in target response indicators). Alternatively, if the target is a cell or organoid, the stage triplet constructed by analyzing the corresponding historical data could be: (Target cell population in abnormal proliferation state, high expression of specific signaling pathways, high basal stress level) - (Treatment with candidate compounds, setting the concentration and duration of action) - (Decreased cell activity, suppressed target pathway activity, elevated stress markers). Or, if the target is a gene locus or chromosomal region, the stage triplet constructed by analyzing the historical data could be: (Target locus in unedited state, abnormal activation of related regulatory networks) - (Execution of gene editing operation and application of corresponding delivery conditions) - (Target locus editing completed, target expression level decreased, accompanied by changes in off-target risk).

[0172] It should be noted that the above examples are merely illustrative and should not be construed as limiting the stage triples in the embodiments of this application.

[0173] Step 408: After structuring the stage triples, store them in the stage information database.

[0174] This allows for various structured processing of stage triples, including field mapping, standard terminology normalization, time alignment, outlier cleaning, entity alignment, and manual validation. For example, non-standard clinical descriptions or laboratory terminology in stage triples, such as "high blood pressure" or "slightly elevated transaminase levels," can be mapped and uniformly replaced with standard vocabulary or numerical ranges, such as "stage 1 hypertension" or "ALT: 85 U / L," thereby minimizing semantic ambiguity.

[0175] If identical stage triples are extracted from the historical records of the same or different subjects under study, they can be deduplicated and merged, and their frequency of occurrence in the database can be accumulated. A corresponding weight value can then be assigned based on the frequency of occurrence. If there are data points with identical pre-intervention states and intervention actions, but conflicting post-intervention states, these can be recorded as parallel probability branches, thus preserving diverse inference possibilities.

[0176] Afterwards, the processed triples can be encapsulated according to a unified data field format and stored in a stage information database. This stage information database can be a distributed graph database, a relational database, or a vector retrieval database, etc. This application does not limit this.

[0177] Therefore, in this embodiment, the stage information database constructed through the above steps, together with the intervention mechanism database, constitutes a dual data source combination for simulation. The stage triples stored in the stage information database can be used to characterize the global temporal evolution trajectory and state transition path of the target object, and can be used to provide baseline state parameters for simulation deduction; the intervention mechanism information stored in the intervention mechanism database, such as the absorption and distribution parameters of intervention factors and target binding kinetic parameters, can be used to characterize the local perturbation characteristics at the molecular pathway level. Through joint retrieval instructions, this application associates and combines the overall evolution path of the target object with the specific intervention mechanism. If the baseline state parameters are not considered, the predicted index parameters of the intervention simulation may deviate from actual laws due to the lack of constraints from the overall baseline state; if the specific intervention parameters are not considered, the intervention simulation may fail to reflect the dynamic changes in the state of the target object caused by external intervention measures. Therefore, the dual data source combination of this application establishes a mutual constraint verification mechanism between the global state and local intervention actions, which can effectively avoid illogical calculation paths in the intervention simulation process, significantly reduce the scope of invalid retrieval and calculation, effectively eliminate prediction branches that do not conform to actual laws, and maximize the accuracy and reliability of the simulation output.

[0178] Optionally, the stage information database can also include a vector similarity retrieval layer, a keyword connection retrieval layer, and an evidence scoring layer. The vector similarity retrieval layer can be used to vectorize the features of stage information tuples and virtual target objects in multiple dimensions to retrieve similar interventions, similar objects, similar stage tuples, similar reaction chains, and similar risk scenarios. The keyword connection retrieval layer can be used to retrieve terms such as mechanism terms, toxicology terms, instruction manual terms, treatment operation terms, and organ response terms. The evidence scoring layer can be used to store and evaluate the evidence level, source type, time fit, population fit, and historical effectiveness score of individual clinical / experimental records.

[0179] Optionally, the intervention mechanism database can also include the same vector similarity retrieval layer, keyword connection retrieval layer, and evidence scoring layer. The vector similarity retrieval layer can be used to embed and encode molecular structures and pathway topologies to retrieve intervention reaction chains or alternative pathways with similar chemical structures or similar topological cascade reactions. The intervention mechanism database can also include a structured relationship graph layer, which can be used in the form of a knowledge graph or attribute graph to express possible multi-level relationship links, such as multi-level relationship links between intervention methods, target sites, metabolites, independent variable molecules / pathways, effector organs, physiological states, and toxicological outcomes.

[0180] Therefore, in this embodiment of the application, by transforming messy real-world clinical records or experimental records into standardized stage triples and storing them in the stage information database, not only are multidimensional details in the historical records preserved, but also the causal evolution logic between intervention measures and state changes is preserved. This provides a reliable data foundation for dynamic retrieval of the dual databases, retrieval instruction updates, and deduction convergence in the subsequent intervention simulation process.

[0181] In this embodiment, historical data of the research subjects can be acquired first. This historical data includes clinical records, experimental records, and treatment operation records. Then, based on chronological order, the historical data of the research subjects is divided into multiple time stages. For each time stage, the historical data of the research subjects is parsed to extract stage triples. The stage triples include the research subjects' pre-intervention state at the beginning of the time stage, the intervention operations performed during the time stage, and the post-intervention state at the end of the stage. The stage triples are then structured and stored in a stage information database. Thus, by processing the historical data, structured stage triples can be obtained and stored in the stage information database. This not only preserves the state information in the original data but also retains the causal relationships between the data, providing a reliable data foundation for subsequent intervention simulations based on the stage information database.

[0182] The intervention response-based simulation method provided in this application can be applied to intervention response simulation systems in any scenario. The following is a brief explanation of the intervention response-based simulation process provided in this application, taking a candidate anti-tumor intervention A as an example.

[0183] The intervention simulation task can be to simulate a candidate anti-tumor intervention plan A on a middle-aged or elderly virtual target subject. The virtual target subject has a basic tumor status, mild liver function abnormalities, and basic metabolic abnormalities. The intervention plan adopts a continuous implementation method. The simulation task requirements are to focus on target response indicators, liver risk, and unexpected gastrointestinal responses.

[0184] For ease of description, the simulation system based on intervention response can be simply referred to as the system. Upon receiving an intervention simulation request, the system can parse it and extract relevant information about the target object. This includes intervention information such as the identifier of intervention plan A, implementation conditions, and whether the implementation method is continuous; basic target object information such as age characteristics (middle-aged and elderly) and underlying disease status (tumor); initial state information such as liver function status (mildly abnormal) and metabolic status (basal metabolic abnormality) as dynamic indicators serving as the starting point for simulation; control variable information such as the observation period and key indicator types (e.g., target response, liver risk, gastrointestinal reactions); and simulation task requirements such as priority for target response and priority for risk response.

[0185] Then, based on the intervention simulation requirements, a multi-layered state structure of the virtual target object can be initialized. Depending on the simulation task type and the key risks, the following state layers and corresponding state elements can be selected in this embodiment: the intervention and exposure layer state can include intervention input state, drug exposure level, target binding state, etc.; the tissue, organ, and system layer state can include liver state, gastrointestinal state, etc.; the phenotypic and functional outcome layer state can include hepatotoxicity risk level, gastrointestinal risk level, etc.

[0186] Understandably, the stage information tuple database was pre-constructed before the simulation began by sorting, dividing, and structuring the longitudinal process data of the historical research object. It stores multiple stage triples. The intervention mechanism database stores pharmacological, toxicological, pathway, and regulatory mechanisms related to various intervention measures.

[0187] The system can retrieve stage information tuples corresponding to longitudinal process data similar to the current target object from the existing stage information tuple database. Specifically, state information and intervention information can be fused to generate a retrieval instruction corresponding to the target object, which is then searched in the stage information database to obtain candidate stage triples. Alternatively, based on the basic information of the current target object, such as middle-aged or elderly, tumor status, mild liver function abnormalities, or basal metabolic abnormalities, a similarity search can be performed in the stage information tuple database to obtain stage triples matching the current simulation scenario as candidate stage triples. Furthermore, the system can retrieve the action path, potential risk links, related structural or functional mechanisms, and descriptive constraint information related to intervention plan A from the intervention mechanism database to obtain candidate intervention mechanism information.

[0188] Subsequently, the system can update one or more state elements in the multi-layered state structure based on the candidate search results, forming the current simulation state. After the first round of state updates, the system reads the liver function state, gastrointestinal state, intervention exposure state, risk level state, and the key event chains that have occurred in the current simulation state, and generates multiple search requests for the next round of simulation based on the simulation task objectives. These search requests can be single or multiple, and may include, for example, search requests for liver risk mechanisms, gastrointestinal response mechanisms, metabolic regulation mechanisms, and disease target-related mechanisms.

[0189] Subsequently, the system can perform incremental searches in the aforementioned stage information database and intervention mechanism database based on the generated search request. If the search results are insufficient, knowledge completion operations such as synonym expansion, upstream and downstream pathway expansion, and similar object stage expansion can be used to supplement available evidence.

[0190] Subsequently, the system can continue to update one or more state elements in the multi-layered state structure based on the incremental retrieval results, forming a state evolution trajectory. Then, according to the iterative loop of updating retrieval instructions - obtaining candidate retrieval results - updating state information - updating retrieval instructions, new retrieval requests are continuously generated based on the latest state, new evidence is obtained, and the state of the target object is continuously updated, realizing continuous and dynamic deduction of the intervention response process.

[0191] Understandably, during the recursive deduction process, the system continuously monitors various status indicators. When the system detects a rapid increase in the liver function risk score exceeding a preset threshold, it identifies this as a significant event.

[0192] After identifying a significant event, the system can verify the simulation-driven chain that triggered it. For example, it can extract the associated state nodes leading to a rapid increase in liver function risk scores from the simulation-driven chain and construct a key causal path. Then, it can re-retrieve multi-source evidence from the stage information tuple database and the intervention mechanism database. Next, it can perform a consistency assessment between the re-retrieved multi-source evidence and the historical evidence in the simulation-driven chain to determine whether there are consistency or inconsistencies between evidence from different sources, or between new evidence and existing evidence. Further analysis can then be conducted to determine whether there are logical contradictions between the causal relationships in the current chain and the mechanistic relationships in the intervention mechanism database. Based on the consistency assessment results and the mechanism conflict analysis results, the confidence level of the chain is re-evaluated, generating simulation verification results.

[0193] If simulation verification results indicate that the current deduction chain suffers from insufficient evidence, logical conflicts, inconsistent mechanisms, or a confidence level below a preset threshold, a rollback operation is triggered. In this embodiment, if a deduction chain is found to be supported only by a low-confidence mechanism and conflicts with the macroscopic regression direction reflected by the stage information tuples of similar target objects, the system rolls back to the stable state before the occurrence of the significant event, performs a weight reduction on the corresponding chain, and regenerates the retrieval request before continuing the simulation. Alternatively, it can roll back to other target rollback nodes, such as the previous stable state, the previous evidence bifurcation point, or other preset nodes; this application does not limit this.

[0194] During multi-round iterative simulation, if a predetermined observation period is reached, the system can determine that the preset simulation termination condition is met, stop the simulation, and output the results. In this embodiment, the output intervention response simulation results may include the final simulation outcome, such as the achievement of the target response and the occurrence of risk events; key state trajectories, such as the trajectories of liver function status, gastrointestinal status, and tumor response status over time; key impact events, such as event nodes that cause significant changes in status; structural or functional state changes; evidence support chains, such as evidence sources supporting each round of state updates; rollback and error correction records; uncertainty explanations; risk warnings, etc., which are not limited in this application.

[0195] Optionally, the status can be updated first based on the candidate search results, and then the search instructions can be updated; alternatively, the search instructions can be updated first based on the candidate search results, and then the status can be updated; or the search instructions and status can be updated simultaneously based on the candidate search results. This application does not limit this approach.

[0196] Therefore, in this embodiment of the application, when performing intervention simulation operations, the conclusion output is no longer limited to static rules or one-time retrieval results. Instead, by constructing a multi-layer state structure of the virtual target object and combining a stage information tuple database with an intervention mechanism database, the state evolution process of the target object under different intervention measures can be continuously represented, dynamically updated, and recursively deduced.

[0197] Furthermore, by adaptively generating retrieval requests based on the current state during the simulation process and performing incremental retrieval and knowledge completion, the relevance of evidence acquisition to the current inference stage can be improved, thereby enhancing the contextual consistency and mechanism coherence of the simulation process. Moreover, since this application verifies upon identifying significant events and corrects low-confidence inference links through evidence re-retrieval, consistency assessment, mechanism conflict analysis, confidence re-evaluation, and backtracking error correction mechanisms, the probability of continued propagation of erroneous links can be reduced, improving the stability, credibility, and interpretability of the simulation results.

[0198] Therefore, this application can more effectively support response simulation, risk assessment and mechanism analysis for drug intervention, surgical intervention, physical therapy intervention, gene editing intervention and other intervention measures, and is applicable to complex state simulation scenarios of human body, animal, tissue, cell and other multi-level target objects.

[0199] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0200] Based on the same inventive concept, this application also provides an intervention-response-based simulation apparatus for implementing the intervention-response-based simulation method described above. The solution provided by this apparatus is similar to the solution described in the above method; therefore, the specific limitations in one or more intervention-response-based simulation apparatus embodiments provided below can be found in the limitations of the intervention-response-based simulation method described above, and will not be repeated here.

[0201] In one exemplary embodiment, such as Figure 5 As shown, a simulation device 500 based on intervention response is provided, including: a receiving module 510, a first generating module 520, an updating module 530, and a first determining module 540, wherein:

[0202] The receiving module 510 is used to receive an intervention simulation request, which carries the status information of the target object and intervention information.

[0203] The first generation module 520 is used to generate a retrieval instruction corresponding to the target object based on the status information and the intervention information, and to perform a retrieval in the stage information database and the intervention mechanism database based on the retrieval instruction to obtain candidate retrieval results;

[0204] The update module 530 is used to update the current search instruction based on the candidate search results to obtain the updated search instruction.

[0205] The first determining module 540 is used to re-execute the step of searching in the stage information database and intervention mechanism database based on the updated search instruction to obtain candidate search results, until the preset simulation stop condition is met, and output the intervention response simulation information of the target object.

[0206] In one embodiment, the device further includes:

[0207] The acquisition module is used to acquire the indicator parameters in the status information of the candidate search results;

[0208] The second generation module is used to verify the current simulation deduction link of the target object when the indicator parameter exceeds the preset threshold corresponding to the significant event, and generate simulation verification results. The simulation deduction link includes the process of the target object being deduced and updated from the initial state information to the current state information.

[0209] The second determining module is used to determine the target rollback node among each historical state node when the simulation verification result meets the rollback triggering rule, and roll back the state information of the target object to the target rollback node.

[0210] In one embodiment, the second generation module is specifically used for:

[0211] Based on the index parameters that exceed the preset threshold of significant events, related state nodes are extracted from the simulation and inference link, and key causal paths are constructed according to the order of the related state nodes.

[0212] Based on the key causal path, a re-search is performed in the stage information database and the intervention mechanism database to obtain multi-source evidence;

[0213] The evidence information from the multi-source evidence is compared with the historical evidence in the simulation and deduction link to determine the consistency result;

[0214] The causal relationships corresponding to the simulation and deduction links are compared with the mechanism relationships in the intervention mechanism database to determine the mechanism conflict results;

[0215] Based on the consistency results and the conflict results of the mechanism, the confidence level of the simulation inference link is evaluated, and simulation verification results are generated.

[0216] In one embodiment, the stage information database stores multiple stage triples corresponding to historical research subjects. Each stage triple includes the pre-intervention state, intervention operation, and post-intervention state of each historical research subject. The first determining module 540 is specifically used for:

[0217] The search command is matched with the pre-intervention state and intervention operation in the stage triplet, and candidate stage triplets are determined based on the matching results;

[0218] Based on the search instructions, a search is performed in the intervention mechanism database to determine candidate intervention mechanism information;

[0219] The candidate stage triples and the candidate intervention mechanism information are determined as candidate retrieval results.

[0220] In one embodiment, the candidate search results include multiple data groups, each data group including status information and / or intervention mechanism information, and the update module 530 includes:

[0221] The first determining unit is used to sort the data groups in the candidate retrieval results based on the simulation task requirement data and determine the weight of each data group.

[0222] The second determining unit is used to determine the state change information of the target object by combining the simulation task objective and the weights of each data group;

[0223] The update unit is used to update the current search instruction based on the state change information to obtain the updated search instruction.

[0224] In one embodiment, the updating unit is specifically used for:

[0225] Obtain current simulation association information, which includes at least one of simulation task requirement data, simulation task objectives, simulation inference links, and current status information;

[0226] The simulation association information, the state change information, and the current search instruction are fused together to obtain an updated search instruction. The fusion process includes at least one of the following: generation based on template mapping, generation through rule combination, or generation based on model.

[0227] In one embodiment, the first generation module 520 is further configured to:

[0228] Obtain historical data of the subjects to be studied, including clinical records, experimental records, and treatment operation records;

[0229] Based on the time sequence, the historical data of the object under study is divided into multiple time periods;

[0230] For each of the defined time periods, the historical data of the subject under study is parsed to extract the stage triples for that time period. The stage triples include the pre-intervention state of the subject under study at the beginning of the time period, the intervention operation performed during the time period, and the post-intervention state at the end of the time period.

[0231] The stage triples are structured and then stored in the stage information database.

[0232] In one embodiment, the intervention response simulation information includes the state information of the target object when the preset simulation stop condition is met, the state information sequence corresponding to each round of multiple iterations, and the intervention information sequence.

[0233] The modules in the aforementioned intervention-response-based simulation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the operations corresponding to each module.

[0234] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements an intervention-response-based simulation method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0235] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0236] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0237] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0238] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0239] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0240] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0241] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0242] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A simulation method based on intervention response, characterized in that, The method includes: Receive an intervention simulation request, the intervention simulation request carrying the state information of the target object and intervention information; Based on the status information and the intervention information, a retrieval instruction corresponding to the target object is generated, and a retrieval is performed in the stage information database and the intervention mechanism database based on the retrieval instruction to obtain candidate retrieval results; Based on the candidate search results, the current search instruction is updated to obtain the updated search instruction; Based on the updated retrieval instructions, the steps of re-executing the stage information database and intervention mechanism database based on the retrieval instructions to obtain candidate retrieval results are repeated until the preset simulation stop condition is met, and the intervention response simulation information of the target object is output.

2. The method according to claim 1, characterized in that, The method further includes: Obtain the indicator parameters from the status information in the candidate search results; When the indicator parameter exceeds the preset threshold corresponding to the significant event, the current simulation inference link of the target object is verified and a simulation verification result is generated. The simulation inference link includes the process of the target object being inferred and updated from the initial state information to the current state information. If the simulation verification results meet the rollback triggering rules, a target rollback node is determined among each historical state node, and the state information of the target object is rolled back to the target rollback node.

3. The method according to claim 2, characterized in that, The step of verifying the current simulation and deduction link of the target object and generating simulation verification results includes: Based on the index parameters that exceed the preset threshold of significant events, related state nodes are extracted from the simulation and inference link, and key causal paths are constructed according to the order of the related state nodes. Based on the key causal path, a re-search is performed in the stage information database and the intervention mechanism database to obtain multi-source evidence; The evidence information from the multi-source evidence is compared with the historical evidence in the simulation and deduction link to determine the consistency result; The causal relationships corresponding to the simulation and deduction links are compared with the mechanism relationships in the intervention mechanism database to determine the mechanism conflict results; Based on the consistency results and the conflict results of the mechanism, the confidence level of the simulation inference link is evaluated, and simulation verification results are generated.

4. The method according to claim 1, characterized in that, The stage information database stores stage triples corresponding to multiple historical research subjects. Each stage triple includes the pre-intervention state, intervention action, and post-intervention state of each historical research subject. The search is performed in the stage information database and the intervention mechanism database based on the search command to obtain candidate search results, including: The search command is matched with the pre-intervention state and intervention operation in the stage triplet, and candidate stage triplets are determined based on the matching results; Based on the search instructions, a search is performed in the intervention mechanism database to determine candidate intervention mechanism information; The candidate stage triples and the candidate intervention mechanism information are determined as candidate retrieval results.

5. The method according to claim 1, characterized in that, The candidate search results include multiple data groups, each data group including status information and / or intervention mechanism information. The updated search instruction is obtained based on the candidate search results, including: Based on the simulation task requirement data, the data groups in the candidate search results are sorted and the weight of each data group is determined. By combining the simulation task objective and the weights of each data group, the state change information of the target object is determined; Based on the state change information, the current search instruction is updated to obtain the updated search instruction.

6. The method according to claim 5, characterized in that, The step of updating the current search instruction based on the state change information to obtain the updated search instruction includes: Obtain current simulation association information, which includes at least one of simulation task requirement data, simulation task objectives, simulation inference links, and current status information; The simulation association information, the state change information, and the current search instruction are fused together to obtain an updated search instruction. The fusion process includes at least one of the following: generation based on template mapping, generation through rule combination, or generation based on model.

7. The method according to claim 1, characterized in that, Before the step of performing a search based on the search instruction to obtain candidate search results, the method further includes: Obtain historical data of the subjects to be studied, including clinical records, experimental records, and treatment operation records; Based on the time sequence, the historical data of the object under study is divided into multiple time periods; For each of the defined time periods, the historical data of the subject under study is parsed to extract the stage triples for that time period. The stage triples include the pre-intervention state of the subject under study at the beginning of the time period, the intervention operation performed during the time period, and the post-intervention state at the end of the time period. The stage triples are structured and then stored in the stage information database.

8. The method according to claim 1, characterized in that, The intervention response simulation information includes the state information of the target object when the preset simulation stop condition is met, the state information sequence corresponding to each round of multiple iterations, and the intervention information sequence.

9. A simulation device based on intervention response, characterized in that, The device includes: The receiving module is used to receive intervention simulation requests, which carry the status information of the target object and intervention information. The first generation module is used to generate a retrieval instruction corresponding to the target object based on the status information and the intervention information, and to perform a retrieval in the stage information database and the intervention mechanism database based on the retrieval instruction to obtain candidate retrieval results; The update module is used to update the current search instruction based on the candidate search results to obtain the updated search instruction; The first determining module is used to re-execute the step of searching in the stage information database and intervention mechanism database based on the updated search instruction to obtain candidate search results, until the preset simulation stop condition is met, and output the intervention response simulation information of the target object.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

Images