A method and system for predicting the intervention response of fritillaria wallichii extract based on multi-dimensional biological data

Abstract

This invention relates to the field of data analysis and discloses a method and system for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data. By extracting response change characteristics from different data dimensions of historical samples and analyzing their sequential occurrence, direction of change, and cross-dimensional correlations, the method identifies correlation features formed by the continuous propagation of the same response process. By constructing a response evidence propagation chain, the method merges gene, protein, metabolic, and phenotypic response characteristics within the same propagation chain, compressing cross-dimensional repetitive representations into single response evidence while retaining independent response evidence, thus avoiding duplicate inclusion and aberrant amplification of the same response information. Based on the merged target response evidence, a mapping relationship between biological state and the intervention response of Fritillaria cirrhosa extract is established, reducing the overestimation of response degree caused by cross-dimensional repetitive evidence and improving the accuracy and reliability of intervention response prediction.

Description

Technical Field

[0001] This invention relates to the field of data analysis, and more specifically to a method and system for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data. Background Technology

[0002] Predictive intervention responses to Fritillaria cirrhosa extract based on multidimensional biological data typically involves collecting multidimensional biological data of research subjects before and after receiving Fritillaria cirrhosa extract intervention. The correlation between pre-intervention biological state and post-intervention response is analyzed to establish a data model for predicting intervention effects. Specifically, gene expression data, protein expression data, metabolic data, biochemical index data, and phenotypic state data of research subjects are first obtained, and a correspondence between different dimensions of data is established according to sample number and collection time. Subsequently, the data of each dimension are standardized, feature extracted, and correlation analyzed to obtain characteristic information that can characterize the initial biological state of research subjects and changes in state after intervention. Based on this, a mapping relationship between biological state characteristics and intervention response is established according to the pre-intervention multidimensional biological characteristics, Fritillaria cirrhosa extract intervention conditions, and actual intervention response results of historical samples. This mapping relationship is then used to analyze new research subjects, outputting the corresponding intervention response category, response degree, or response probability, thereby achieving early prediction of the intervention effect of Fritillaria cirrhosa extract.

[0003] However, in the aforementioned multidimensional biological data fusion prediction process, different data dimensions do not necessarily provide independent response information. Because there are usually continuous biological regulatory and information transmission relationships between gene expression, protein expression, metabolic changes, and phenotypic changes, the same upstream biological change may propagate step-by-step along different biological levels, forming response features in multiple data dimensions. For example, a change in the same gene expression may further cause changes in the corresponding protein level, and through changes in protein function, cause synchronous changes in downstream metabolites and phenotypic states. In this case, although multiple dimensions all show changes supporting the same intervention response, these changes may actually originate from the same response process. Existing prediction methods usually focus more on whether different dimensions jointly support a certain response result, lacking further distinction of the information source relationship and propagation correlation between various response features. This easily leads to the continuous representation of the same response process in different dimensions being repeatedly included in the prediction as multiple independent pieces of evidence, causing the same response information to be counted multiple times and abnormally amplified. Consequently, the degree or probability of the intervention response from Fritillaria cirrhosa extract is overestimated, affecting the accuracy and reliability of the final prediction results. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method and system for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data, so as to solve the problems existing in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data, the method comprising: S1: Obtain multidimensional biological data corresponding to historical samples and intervention response results of Fritillaria cirrhosa extract, extract response change features in different data dimensions, and construct a cross-dimensional response feature set according to the sample correspondence; S2: Based on the cross-dimensional response feature set, analyze the sequential occurrence relationship, change direction relationship and cross-dimensional correlation relationship among different response change features, identify the correlation features formed by the continuous propagation of the same response change, and construct the response evidence propagation chain accordingly; S3: Based on the response evidence propagation chain, multiple response features belonging to the same propagation chain are merged into evidence, the repeated representations of the same response process in different data dimensions are compressed into a single response evidence, and response features that do not form a propagation association are retained as independent response evidence, thus obtaining the target response evidence set. S4: Based on the target response evidence set, establish a mapping relationship between multidimensional biological state and the intervention response of Fritillaria cirrhosa extract, and use the mapping relationship to perform response analysis on the sample to be predicted, so as to obtain the prediction result of the intervention response of Fritillaria cirrhosa extract.

[0006] The present invention also provides a system for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data, comprising: a response feature module: acquiring multidimensional biological data corresponding to historical samples and the intervention response results of Fritillaria cirrhosa extract, extracting response change features in different data dimensions, and constructing a cross-dimensional response feature set according to the sample correspondence; Propagation Chain Module: Based on a cross-dimensional response feature set, this module analyzes the sequential occurrence, direction of change, and cross-dimensional correlation among different response change features, identifies the associated features formed by the continuous propagation of the same response change, and constructs a response evidence propagation chain accordingly. Response Evidence Module: Based on the response evidence propagation chain, multiple response features belonging to the same propagation chain are merged into evidence, the repeated representations of the same response process in different data dimensions are compressed into a single response evidence, and response features that do not form a propagation association are retained as independent response evidence, thus obtaining the target response evidence set; Prediction module: Based on the target response evidence set, a mapping relationship is established between the multidimensional biological state and the intervention response of Fritillaria cirrhosa extract, and the mapping relationship is used to perform response analysis on the sample to be predicted to obtain the prediction result of the intervention response of Fritillaria cirrhosa extract.

[0007] This invention extracts response change features from different data dimensions of historical samples in a unified manner, and further analyzes the sequential occurrence, direction of change, and cross-dimensional correlation among these response features. This enables the system to identify correlation features formed by the continuous propagation of the same upstream response change along different biological levels, avoiding the direct assumption of multiple independent response evidence based solely on simultaneous changes in multiple dimensions. By constructing a response evidence propagation chain and merging multiple response features within the same propagation chain, the repetitive representations of the same response process across different dimensions such as genes, proteins, metabolism, and phenotypes are compressed into a single response evidence, while retaining independent response evidence that has not formed a propagation correlation. This reduces the duplication and amplification of the same response information. Based on this, the merged target response evidence set is used to establish a mapping relationship between multidimensional biological states and the intervention response of Fritillaria cirrhosa extract. This allows the final prediction results to more realistically reflect the degree of support of independent response sources for the intervention effect, reducing the overestimation of response degree or probability caused by cross-dimensional repetitive evidence, and improving the accuracy and reliability of Fritillaria cirrhosa extract intervention response prediction. Attached Figure Description

[0008] Figure 1 This is a flowchart of a method for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data according to the present invention. Detailed Implementation

[0009] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The method and system for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data involved in the present invention are not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0010] like Figure 1 As shown, this invention provides a method for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data, comprising the following steps: S1: Obtain multidimensional biological data corresponding to historical samples and the intervention response results of Fritillaria cirrhosa extract. Extract response change features from different data dimensions and construct a cross-dimensional response feature set according to the sample correspondence. Specifically: Gene expression data, protein expression data, metabolic data, and phenotypic status data of each historical sample at each collection time before and after intervention with Fritillaria cirrhosa extract were obtained, and the intervention response results of Fritillaria cirrhosa extract for each historical sample were obtained. The biological data of each dimension were arranged according to the sample number and collection time. For each historical sample, the change of the same detection index in each data dimension between adjacent collection times is calculated, and the change is normalized according to the detection value before intervention to obtain the time series change sequence corresponding to each detection index. Based on the time-series change sequence corresponding to each detection indicator, the collection time, deviation direction, and cumulative change magnitude from the first deviation to the last collection time are extracted. The first deviation collection time, deviation direction, and cumulative change magnitude are used together as the response change characteristics of the corresponding detection indicator. Each response change feature is associated with its corresponding sample number, data dimension, and detection index. The response change features corresponding to different data dimensions are aggregated according to the same sample number to obtain a cross-dimensional response feature subset corresponding to each historical sample. A one-to-one correspondence is established between the cross-dimensional response feature subsets corresponding to each historical sample and the corresponding intervention response results of Fritillaria cirrhosa extract. The cross-dimensional response feature subsets are then aggregated according to the sample number to construct a cross-dimensional response feature set.

[0011] It should be specifically noted that multidimensional biological data refers to data used to characterize the state changes of historical samples during the intervention of Fritillaria cirrhosa extract at different biological levels. This includes gene expression data, protein expression data, metabolic data, and phenotypic state data. Specifically, gene expression data refers to the expression levels of each target gene in historical samples obtained through gene expression detection, used to record the expression changes caused by the intervention of Fritillaria cirrhosa extract at the gene level; protein expression data refers to the expression levels or contents of each target protein in historical samples obtained through protein expression detection, used to record the protein layer changes corresponding to the changes in target gene expression; metabolic data refers to the concentrations or relative contents of each target metabolite in historical samples obtained through metabolite detection, used to record the downstream metabolic state changes corresponding to the changes in protein function; and phenotypic state data refers to at least one of cell activity, proliferation level, inflammatory factor content, or tissue state indicators obtained through cell or tissue state detection, used to record the detectable biological state changes ultimately corresponding to the above molecular-level changes. For each historical sample, the above data were collected once before the addition of Fritillaria cirrhosa extract. The results of this test were used as the pre-intervention baseline values ​​for each indicator. The same indicators were repeatedly tested at multiple preset collection times after the addition of Fritillaria cirrhosa extract, such as 2 hours, 6 hours, 12 hours and 24 hours after the intervention. This ensures that each indicator has a pre-intervention value and multiple post-intervention values. Meanwhile, the Fritillaria cirrhosa extract intervention response result refers to the historical sample response result determined based on the actual degree of improvement relative to the pre-intervention baseline state at the last collection time. It is used to represent the actual intervention effect that has occurred in the historical sample and serves as a known result for establishing response prediction relationships in the future. After obtaining the above data, gene, protein, metabolic, and phenotypic data belonging to the same historical sample are mapped together by sample number. Then, consecutive detection values ​​of the same indicator are arranged according to the collection time. For example, if the expression levels of gene A in a historical sample before intervention and at 2, 6, 12, and 24 hours after intervention are 100, 98, 80, 62, and 55, respectively, the changes between adjacent collection times are calculated sequentially, using the pre-intervention expression level of 100 as the normalization benchmark. This forms a temporal change sequence of gene A over intervention time. Protein B, metabolite C, and phenotypic indicator D are processed in the same way. Subsequently, from each temporal change sequence, the collection time at which the indicator first deviates from the pre-intervention state, the direction of increase or decrease relative to the pre-intervention state, and the cumulative change magnitude from the first deviation to the last collection time are determined. These three pieces of information are used as the response change characteristics of the indicator.For example, if gene A shows no significant change after 2 hours but begins to decline after 6 hours, then 6 hours is defined as the first response time of gene A, the decline is defined as the direction of change, and the cumulative decline from 6 hours to the last collection time is calculated. If protein B, which is related to this gene, begins to decline after 12 hours, and downstream metabolite C begins to decline after 24 hours, then response change characteristics of the protein layer and the metabolic layer are formed respectively. Finally, the response change characteristics of all detection indicators in the same historical sample are recorded in the form of "sample number - data dimension - detection indicator - first response time - direction of change - cumulative change magnitude" to form a cross-dimensional response feature subset of the historical sample. Then, all cross-dimensional response feature subsets of historical samples are matched one-to-one with the previously determined intervention response results of Fritillaria cirrhosa extract to form a cross-dimensional response feature set. The advantage of this approach is that subsequent analyses no longer deal with isolated original gene, protein, and metabolic test values ​​with different dimensions, but rather with directly comparable "response timing, response direction, and response magnitude." Therefore, when it is subsequently discovered that a gene changes first, the corresponding protein changes in the same direction, and the related metabolites change in the same direction at a later time, it is possible to further determine whether these response evidences, which appear to belong to three data dimensions, are actually the result of a continuous propagation of the same response process.

[0012] S2: Based on a cross-dimensional response feature set, analyze the sequential occurrence, directional relationship, and cross-dimensional correlation among different response change features, identify the associated features formed by the continuous propagation of the same response change, and construct a response evidence propagation chain accordingly. Specifically: For each historical sample in the cross-dimensional response feature set, the response change features in gene expression dimension, protein expression dimension, metabolism dimension and phenotypic state dimension are extracted according to the data dimension corresponding to each response change feature. They are then arranged in order of first deviation from the acquisition time corresponding to each response change feature to obtain the cross-dimensional response time series corresponding to each historical sample. Based on cross-dimensional response time series, response change features located in different data dimensions and with different first deviations from the acquisition time are paired up. The difference between the first deviations from the acquisition time of the two response change features in each pair is calculated. The corresponding change direction combination is determined according to the increasing or decreasing direction of the two response change features to obtain cross-dimensional response feature pairs. For cross-dimensional response feature pairs with the same detection index in different historical samples, the number of historical samples with the same order of appearance and the same direction of change of the two detection indexes is counted, and the proportion of the number of such historical samples to the total number of historical samples that simultaneously contain the two detection indexes is calculated. The proportion is used as the propagation consistency of the corresponding cross-dimensional response feature pair. Cross-dimensional response feature pairs with a propagation consistency greater than or equal to a preset propagation consistency threshold are identified as candidate propagation feature pairs. Based on the first deviation of the two response change features in each candidate propagation feature pair from the acquisition time, the response change feature with the earlier first deviation from the acquisition time is identified as the preceding response feature, and the response change feature with the later first deviation from the acquisition time is identified as the following response feature, thus establishing a cross-dimensional propagation relationship from the preceding response feature to the following response feature. According to the detection index of the preceding response features and the following response features in each cross-dimensional propagation relationship, the preceding response features of the following cross-dimensional propagation relationship and the following response features of the preceding cross-dimensional propagation relationship with the same detection index are connected in sequence, and arranged in the direction of increasing deviation from the first acquisition time to obtain the candidate response propagation chain. For each candidate response propagation chain, the propagation consistency degree corresponding to all cross-dimensional propagation relationships that make up the candidate response propagation chain is extracted. The smallest propagation consistency degree is taken as the link propagation consistency degree of the corresponding candidate response propagation chain. The candidate response propagation chain with a link propagation consistency degree greater than or equal to the preset link consistency degree threshold is determined as the response evidence propagation chain.

[0013] Specifically, for each historical sample, response change features are first extracted in the gene expression, protein expression, metabolism, and phenotypic state dimensions. These features are then sorted chronologically based on the "first deviation from the collection time" determined in step S1, thus obtaining the order in which different response changes occur within the sample. For example, in a historical sample, gene A first decreases 2 hours after intervention, protein B first decreases 6 hours, metabolite C first decreases 12 hours, and phenotypic index D first improves 24 hours, forming a temporal arrangement of "gene A—protein B—metabolite C—phenotype D". Subsequently, response change features located in different data dimensions and with different first deviation times are paired, such as forming cross-dimensional response feature pairs like "gene A—protein B", "gene A—metabolite C", and "protein B—metabolite C". The first deviation time difference between the two features and their combination of change directions are recorded, such as "first decrease—then decrease" or "first increase—then decrease", to describe whether a stable sequential and directional relationship exists between the two response changes. Since accidental synchronous changes in a single historical sample cannot directly indicate that two features belong to the same propagation process, further statistical analysis is performed on cross-dimensional response features with the same combination of detection indicators across all historical samples. For example, for the combination "gene A—protein B", the number of historical samples that simultaneously detected gene A and protein B is counted, and then the number of samples in which gene A always changes before protein B and both always maintain the same direction of change is counted. The ratio of these two is taken as the propagation consistency of the feature pair. When this propagation consistency reaches a preset threshold, it indicates that the phenomenon of "gene A changing first, protein B changing later, and the directional relationship being stable" is not an accidental result in a single sample, but rather a recurrence in multiple historical samples. Therefore, this feature pair is identified as a candidate propagation feature pair, and a cross-dimensional propagation relationship is established from the earlier response feature to the later response feature according to the time of the first deviation. Subsequently, the cross-dimensional propagation relationships are linked end-to-end. For example, if "gene A → protein B" and "protein B → metabolite C" have been identified, and the preceding characteristic protein B in the second relationship and the following characteristic protein B in the first relationship are the same detection indicator, then the two relationships are linked as "gene A → protein B → metabolite C". If "metabolite C → phenotype D" further exists, then the links are continued to form a complete candidate response propagation chain. Considering that the reliability of a propagation chain depends on its weakest segment, the propagation consistency of each cross-dimensional propagation relationship in the candidate propagation chain is extracted, and the minimum value is taken as the link propagation consistency of the candidate propagation chain. Only when this minimum value still reaches the preset link consistency threshold is it determined as the final response evidence propagation chain.The above processing avoids directly assuming a propagation relationship based solely on changes in multiple dimensions within a single sample. Instead, it utilizes the consistency of "time sequence and direction of change" across multiple historical samples to verify the stability of cross-dimensional relationships. By connecting segments sequentially to form a complete propagation path, it identifies continuous evidence that may originate from the same response process, such as "gene changes—protein changes—metabolic changes—phenotypic changes." This reduces the problem of the same response information being repeatedly calculated and amplified abnormally across different data dimensions.

[0014] S3: Based on the response evidence propagation chain, multiple response features belonging to the same propagation chain are merged into single response evidence by compressing repeated representations of the same response process across different data dimensions. Response features that do not form a propagation link are retained as independent response evidence, resulting in the target response evidence set. Specifically: For each historical sample, the response change features in the cross-dimensional response feature subset are matched with the detection indicators in the response evidence propagation chain. The response change features that can be continuously matched according to the order of the response evidence propagation chain are assigned to the same propagation evidence group, and the response change features that do not match any response evidence propagation chain are assigned to an independent evidence group. For each propagation evidence group, the first-end response feature and the last-end response feature are determined according to the order of the propagation chain of response evidence. The first deviation acquisition time corresponding to the first-end response feature is extracted as the response start time of the propagation evidence group. The difference between the first deviation acquisition time corresponding to the last-end response feature and the first deviation acquisition time corresponding to the first-end response feature is extracted as the cross-dimensional propagation time of the propagation evidence group. Extract the cumulative change magnitude corresponding to each response change feature in the transmission evidence group, and divide each cumulative change magnitude by the maximum cumulative change magnitude of the detection index to which the response change feature belongs in all historical samples to obtain the standard change magnitude corresponding to each response change feature. According to the order of the response evidence propagation chain, the ratio of the standard change amplitude corresponding to two adjacent response change characteristics is calculated in turn, and the ratio of each standard change amplitude is arranged in the order of the corresponding cross-dimensional propagation relationship to form a propagation amplitude sequence. The propagation amplitude sequence, together with the response start time and the cross-dimensional propagation duration, is determined as the propagation state characteristic of the corresponding propagation evidence group. Delete the independent evidence identifiers of multiple response change features in the same propagation evidence group, and set a unified evidence identifier for the propagation evidence group. Record the detection index arrangement, propagation state characteristics and unified evidence identifier corresponding to the response evidence propagation chain as a merged response evidence, so that the same response evidence propagation chain forms only one merged response evidence in the corresponding historical sample. For each response change characteristic in the independent evidence group, retain its first deviation time, change direction and cumulative change magnitude, and set an independent evidence identifier to form the corresponding independent response evidence; According to the sample number, the merged response evidence and independent response evidence corresponding to each historical sample are collected and a correspondence is established with the corresponding Fritillaria cirrhosa extract intervention response results to obtain the target response evidence set.

[0015] It should be specifically noted that the core of reorganizing the evidence for the response change features in each historical sample is not to delete gene, protein, or metabolic data, but to change the "evidence counting method" when these features participate in subsequent intervention response prediction, so that multiple cross-dimensional features belonging to the same propagation process participate in the establishment of subsequent mapping relationships only as one response evidence. Specifically, for each historical sample, the cross-dimensional response feature subset obtained in S1 is first matched item by item with the response evidence propagation chain obtained in S2. For example, if S2 has determined a response evidence propagation chain of "gene A → protein B → metabolite C", then it is checked whether the current historical sample simultaneously has response change features corresponding to gene A, protein B and metabolite C, and whether the first deviation time of the three maintains the chain order of gene A earlier than protein B and protein B earlier than metabolite C. When the above continuous matching conditions are met, gene A, protein B and metabolite C are no longer treated as three independent response evidences, but are instead classified into the same propagation evidence group. For response change features that have not entered any response evidence propagation chain, such as another metabolite E undergoing a significant change, but there is no stable propagation relationship that can connect it with other dimensional features, then metabolite E is classified into the independent evidence group. After grouping, it is necessary to retain the effective differences inherent in the same propagation process. Therefore, it is not simply a matter of merging multiple features in the propagation evidence group into a single value, but rather a re-description of the entire propagation process: the first deviation of the response feature at the forefront of the propagation chain from the collection time is taken as the response start time. For example, if gene A changes for the first time 2 hours after intervention, then 2 hours is recorded as the response start time for this propagation evidence group. The difference between the first deviation of the terminal response feature and the head response feature is used to represent the cross-dimensional propagation duration. For example, if protein B changes at 6 hours and metabolite C changes at 12 hours, then the propagation process extends from 2 hours to 12 hours, and the cross-dimensional propagation duration is 10 hours. Since gene expression level, protein content, and metabolite concentration have different dimensions, the cumulative change amplitude of the three cannot be directly compared. Therefore, normalization is performed using the maximum cumulative change amplitude of the corresponding detection index in all historical samples. For example, if the cumulative change amplitude of gene A in the current sample is 40, while the maximum cumulative change amplitude of gene A in all historical samples is 80, then the standard change amplitude of gene A is 0.5. The standard change amplitudes of protein B and metabolite C are obtained in the same way. Subsequently, following the propagation order of "gene A → protein B → metabolite C", the standard change amplitudes of adjacent levels are compared to obtain the amplitude transmission relationships of "protein B relative to gene A" and "metabolite C relative to protein B". A propagation amplitude sequence is then formed according to the propagation order to record whether the intensity of change remains, decays, or increases when the same response process propagates from upstream to downstream. This propagation amplitude sequence, together with the response start time and cross-dimensional propagation duration, serves as the propagation state characteristics of the entire propagation evidence set.After the above processing, the individual identifiers of gene A, protein B, and metabolite C as independent evidence are removed. A unified evidence identifier is assigned to the same propagation evidence group containing all three, and the detection index arrangement of "gene A → protein B → metabolite C" and its corresponding propagation status characteristics are recorded as a single merged response evidence. This "merging" does not mean discarding the three original response features, but rather retaining their propagation order and amplitude change information, while stipulating that they represent only one response source in subsequent predictions. In contrast, for response change features such as metabolite E in the independent evidence group, their initial deviation time from the collection point, direction of change, and cumulative amplitude of change are retained, and independent evidence identifiers are assigned to each, making them each an independent response evidence. Finally, using the sample number as an index, all merged and independent response evidence from each historical sample are aggregated and correlated with the actual intervention response results of Fritillaria cirrhosa extract obtained for that historical sample, forming a target response evidence set. Through this process, for example, the original "decline in gene A, decrease in protein B, decrease in metabolite C, and decrease in metabolite E" might be considered as four pieces of evidence supporting a high response using traditional methods. However, this step identifies the first three as belonging to the same propagation process and treats them as a merged response evidence, while only considering metabolite E as another independent response evidence. Therefore, the final result is two independent response sources, rather than four surface response features. At the same time, the propagation state features retain the time and amplitude transmission information within the first three items, thus avoiding the loss of effective information caused by simply deleting relevant features. It also prevents the continuous representation of the same upstream change in multiple dimensions such as genes, proteins, and metabolism from being counted repeatedly. This provides a predictive basis after eliminating duplicate evidence for S4 to establish the mapping relationship between multidimensional biological states and the intervention response of Fritillaria cirrhosa extract, thereby reducing the overestimation of the intervention response level or response probability caused by the accumulation of cross-dimensional duplicate evidence.

[0016] S4: Based on the target response evidence set, establish a mapping relationship between multidimensional biological states and the intervention response of Fritillaria cirrhosa extract, and use the mapping relationship to perform response analysis on the sample to be predicted, obtaining the prediction results of the intervention response of Fritillaria cirrhosa extract. Specifically: Based on the intervention response results of Fritillaria cirrhosa extract corresponding to the historical samples, the historical samples in the target response evidence set were divided into high response sample group, medium response sample group and low response sample group, and the merged response evidence and independent response evidence corresponding to each historical sample were extracted respectively. For each response sample group, the number of times each merged response evidence and each independent response evidence appears in the response sample group is counted, and then divided by the total number of historical samples in the response sample group to obtain the intra-group occurrence ratio of each response evidence; the intra-group occurrence ratio of the same response evidence in the high response sample group, medium response sample group and low response sample group is arranged according to the response level to form the response distribution sequence of the corresponding response evidence. Establish a correspondence between the response distribution sequence corresponding to each response evidence and the unified evidence identifier or independent evidence identifier corresponding to that response evidence, and associate and store each response evidence identifier, the corresponding response distribution sequence and the corresponding response level to construct a response mapping relationship between multidimensional biological state and the intervention response of Fritillaria cirrhosa extract. Gene expression data, protein expression data, metabolic data, and phenotypic state data of the sample to be predicted before intervention with Fritillaria cirrhosa extract were obtained. The normalized difference between the detection value of each detection indicator and the baseline value of the same detection indicator before intervention in the target response evidence set was calculated to obtain the multidimensional initial state features of the sample to be predicted. The multidimensional initial state features of the sample to be predicted are matched with the multidimensional initial state features of each historical sample in the target response evidence set before intervention, and the proportion of the number of detection indicators that meet the preset difference range to the total number of common detection indicators is calculated. The proportion is determined as the state matching degree between the sample to be predicted and the corresponding historical sample. Historical samples with a state matching degree greater than or equal to a preset state matching threshold are selected as target historical samples. The merged response evidence and independent response evidence corresponding to each target historical sample are extracted, and the extracted response evidence is merged according to the same evidence identifier to obtain the candidate response evidence corresponding to the sample to be predicted. Based on the response distribution sequence corresponding to each candidate response evidence, the sum of the proportions of candidate response evidence appearing in the high response sample group, medium response sample group and low response sample group are counted respectively. The response level corresponding to the response sample group with the largest sum of the proportions of appearance in the group is determined as the target response level of the sample to be predicted, and the intervention response prediction results of Fritillaria cirrhosa extract are obtained.

[0017] It should be specifically explained that, based on merging cross-dimensional repeated response features into single response evidence and retaining independent response evidence, a predictive relationship is further established using these historical sample data after eliminating repeated evidence, and this predictive relationship is applied to new samples to be predicted. Specifically, firstly, based on the actual intervention results of Fritillaria cirrhosa extract obtained for each historical sample, the historical samples are divided into three categories: high response, medium response, and low response. For example, according to pre-set response evaluation criteria, samples whose post-intervention improvement meets the criteria for high response can be classified into the high response sample group, those that meet the criteria for medium response can be classified into the medium response sample group, and the rest can be classified into the low response sample group. Subsequently, the merged response evidence and independent response evidence formed by S3 are extracted from each response sample group, and the frequency of each type of response evidence in that group is counted. For example, if a certain aggregated response evidence, "gene A → protein B → metabolite C," appears 80 times in 100 high-response samples, 32 times in 80 medium-response samples, and 10 times in 100 low-response samples, then its within-group occurrence proportions in the three response levels are 0.80, 0.40, and 0.10, respectively. This forms the response distribution sequence corresponding to this response evidence, which indicates in which intervention response level the response evidence is more likely to appear. The same method is used for independent response evidence. Then, the evidence identifier of each response evidence is associated and stored with its corresponding high, medium, and low response distribution results. This allows the system to directly obtain the support relationship of the evidence for different response levels when a response evidence is subsequently identified, thus forming the mapping basis between response evidence and the actual intervention results of Fritillaria cirrhosa extract. For the sample to be predicted, since it has not yet received intervention from Fritillaria cirrhosa extract, its response changes after intervention cannot be directly obtained. Therefore, gene expression data, protein expression data, metabolic data, and phenotypic state data before intervention are collected first, and normalized according to the same detection indicators and data processing methods as historical samples to form multidimensional initial state features of the sample to be predicted. Subsequently, these multidimensional initial state features are compared item by item with the multidimensional initial state features of each historical sample before intervention. For example, if there are 50 comparable detection indicators between the sample to be predicted and a certain historical sample, and the normalized difference of 42 of the detection indicators is within the preset difference range, then the state matching degree between the two is 42 divided by 50, i.e., 0.84. When the state matching degree reaches the preset state matching threshold, it indicates that the overall biological state of the historical sample before intervention is relatively close to that of the sample to be predicted, and therefore it is identified as the target historical sample.Subsequently, instead of directly using the original multidimensional features of these target historical samples for prediction, the merged response evidence and independent response evidence corresponding to them after S3 processing are extracted, and response evidence with the same evidence identifier is merged to obtain a set of candidate response evidence applicable to the initial state of the current sample to be predicted. Finally, the response distribution sequence corresponding to each of these candidate response evidences is read, and the proportion of occurrence of high, medium and low responses within the group is accumulated respectively. For example, if the cumulative support results of candidate evidence for high, medium and low responses are 4.6, 2.8 and 1.5 respectively, then the high response level with the largest cumulative result is determined as the target response level of the sample to be predicted, thus obtaining the intervention response prediction result of Fritillaria cirrhosa extract. Through the above processing, the response judgment of the sample to be predicted is not directly based on "how many genes, proteins or metabolic indicators are associated". Instead, it first selects truly comparable samples through the initial state of historical samples, and then calls the response evidence that has been merged with repeated evidence to determine the level. Therefore, it can avoid the repeated inclusion of multiple representations formed by the same upstream response process in different biological dimensions into the prediction results. At the same time, it retains the actual support of independent response sources for different response levels, making the established mapping relationship closer to the real independent response structure. This reduces the problem of high response misjudgment or inflated response probability caused by the accumulation of cross-dimensional repeated evidence, and improves the accuracy, stability and interpretability of the prediction of the intervention response of Fritillaria cirrhosa extract.

[0018] This invention also provides a system for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data, comprising: Response Feature Module: Acquire multidimensional biological data corresponding to historical samples and intervention response results of Fritillaria cirrhosa extract, extract response change features in different data dimensions, and construct a cross-dimensional response feature set according to the sample correspondence; Propagation Chain Module: Based on a cross-dimensional response feature set, this module analyzes the sequential occurrence, direction of change, and cross-dimensional correlation among different response change features, identifies the associated features formed by the continuous propagation of the same response change, and constructs a response evidence propagation chain accordingly. Response Evidence Module: Based on the response evidence propagation chain, multiple response features belonging to the same propagation chain are merged into evidence, the repeated representations of the same response process in different data dimensions are compressed into a single response evidence, and response features that do not form a propagation association are retained as independent response evidence, thus obtaining the target response evidence set; Prediction module: Based on the target response evidence set, a mapping relationship is established between the multidimensional biological state and the intervention response of Fritillaria cirrhosa extract, and the mapping relationship is used to perform response analysis on the sample to be predicted to obtain the prediction result of the intervention response of Fritillaria cirrhosa extract.

[0019] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data, characterized in that, The method includes: S1: Obtain multidimensional biological data corresponding to historical samples and intervention response results of Fritillaria cirrhosa extract, extract response change features in different data dimensions, and construct a cross-dimensional response feature set according to the sample correspondence; S2: Based on the cross-dimensional response feature set, analyze the sequential occurrence relationship, change direction relationship and cross-dimensional correlation relationship among different response change features, identify the correlation features formed by the continuous propagation of the same response change, and construct the response evidence propagation chain accordingly; S3: Based on the response evidence propagation chain, multiple response features belonging to the same propagation chain are merged into evidence, the repeated representations of the same response process in different data dimensions are compressed into a single response evidence, and response features that do not form a propagation association are retained as independent response evidence, thus obtaining the target response evidence set. S4: Based on the target response evidence set, establish a mapping relationship between multidimensional biological state and the intervention response of Fritillaria cirrhosa extract, and use the mapping relationship to perform response analysis on the sample to be predicted, so as to obtain the prediction result of the intervention response of Fritillaria cirrhosa extract.

2. The method for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data according to claim 1, characterized in that: The steps for constructing a cross-dimensional response feature set based on sample correspondence are as follows: Gene expression data, protein expression data, metabolic data, and phenotypic status data of each historical sample at each collection time before and after intervention with Fritillaria cirrhosa extract were obtained, and the intervention response results of Fritillaria cirrhosa extract for each historical sample were obtained. The biological data of each dimension were arranged according to the sample number and collection time. For each historical sample, the change of the same detection index in each data dimension between adjacent collection times is calculated, and the change is normalized according to the detection value before intervention to obtain the time series change sequence corresponding to each detection index. Based on the time-series change sequence corresponding to each detection indicator, the collection time, deviation direction, and cumulative change magnitude from the first deviation to the last collection time are extracted. The first deviation collection time, deviation direction, and cumulative change magnitude are used together as the response change characteristics of the corresponding detection indicator. Each response change feature is associated with its corresponding sample number, data dimension, and detection index. The response change features corresponding to different data dimensions are aggregated according to the same sample number to obtain a cross-dimensional response feature subset corresponding to each historical sample. A one-to-one correspondence is established between the cross-dimensional response feature subsets corresponding to each historical sample and the corresponding intervention response results of Fritillaria cirrhosa extract. The cross-dimensional response feature subsets are then aggregated according to the sample number to construct a cross-dimensional response feature set.

3. The method for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data according to claim 1, characterized in that, The steps for identifying associated features formed by the continuous propagation of the same response change, and constructing a response evidence propagation chain accordingly, are as follows: For each historical sample in the cross-dimensional response feature set, the response change features in gene expression dimension, protein expression dimension, metabolism dimension and phenotypic state dimension are extracted according to the data dimension corresponding to each response change feature. They are then arranged in order of first deviation from the acquisition time corresponding to each response change feature to obtain the cross-dimensional response time series corresponding to each historical sample. Based on cross-dimensional response time series, response change features located in different data dimensions and with different first deviations from the acquisition time are paired up. The difference between the first deviations from the acquisition time of the two response change features in each pair is calculated. The corresponding change direction combination is determined according to the increasing or decreasing direction of the two response change features to obtain cross-dimensional response feature pairs. For cross-dimensional response feature pairs with the same detection index in different historical samples, the number of historical samples with the same order of appearance and the same direction of change of the two detection indexes is counted, and the proportion of the number of such historical samples to the total number of historical samples that simultaneously contain the two detection indexes is calculated. The proportion is used as the propagation consistency of the corresponding cross-dimensional response feature pair. Cross-dimensional response feature pairs with a propagation consistency greater than or equal to a preset propagation consistency threshold are identified as candidate propagation feature pairs. Based on the first deviation of the two response change features in each candidate propagation feature pair from the acquisition time, the response change feature with the earlier first deviation from the acquisition time is identified as the preceding response feature, and the response change feature with the later first deviation from the acquisition time is identified as the following response feature, thus establishing a cross-dimensional propagation relationship from the preceding response feature to the following response feature.

4. The method for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data according to claim 3, characterized in that, The steps to construct a chain of evidence propagation for a response also include: According to the detection index of the preceding response features and the following response features in each cross-dimensional propagation relationship, the preceding response features of the following cross-dimensional propagation relationship and the following response features of the preceding cross-dimensional propagation relationship with the same detection index are connected in sequence, and arranged in the direction of increasing deviation from the first acquisition time to obtain the candidate response propagation chain. For each candidate response propagation chain, the propagation consistency degree corresponding to all cross-dimensional propagation relationships that make up the candidate response propagation chain is extracted. The smallest propagation consistency degree is taken as the link propagation consistency degree of the corresponding candidate response propagation chain. The candidate response propagation chain with a link propagation consistency degree greater than or equal to the preset link consistency degree threshold is determined as the response evidence propagation chain.

5. The method for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data according to claim 1, characterized in that: The steps to obtain the target response evidence set are as follows: For each historical sample, the response change features in the cross-dimensional response feature subset are matched with the detection indicators in the response evidence propagation chain. The response change features that can be continuously matched according to the order of the response evidence propagation chain are assigned to the same propagation evidence group, and the response change features that do not match any response evidence propagation chain are assigned to an independent evidence group. For each propagation evidence group, the first-end response feature and the last-end response feature are determined according to the order of the propagation chain of response evidence. The first deviation acquisition time corresponding to the first-end response feature is extracted as the response start time of the propagation evidence group. The difference between the first deviation acquisition time corresponding to the last-end response feature and the first deviation acquisition time corresponding to the first-end response feature is extracted as the cross-dimensional propagation time of the propagation evidence group. Extract the cumulative change magnitude corresponding to each response change feature in the transmission evidence group, and divide each cumulative change magnitude by the maximum cumulative change magnitude of the detection index to which the response change feature belongs in all historical samples to obtain the standard change magnitude corresponding to each response change feature.

6. The method for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data according to claim 1, characterized in that: The steps to obtain the predicted intervention response results of Fritillaria cirrhosa extract are as follows: Based on the intervention response results of Fritillaria cirrhosa extract corresponding to the historical samples, the historical samples in the target response evidence set were divided into high response sample group, medium response sample group and low response sample group, and the merged response evidence and independent response evidence corresponding to each historical sample were extracted respectively. For each response sample group, the number of times each merged response evidence and each independent response evidence appears in the response sample group is counted, and then divided by the total number of historical samples in the response sample group to obtain the intra-group occurrence ratio of each response evidence; the intra-group occurrence ratio of the same response evidence in the high response sample group, medium response sample group and low response sample group is arranged according to the response level to form the response distribution sequence of the corresponding response evidence. Establish a correspondence between the response distribution sequence corresponding to each response evidence and the unified evidence identifier or independent evidence identifier corresponding to that response evidence, and associate and store each response evidence identifier, the corresponding response distribution sequence and the corresponding response level to construct a response mapping relationship between multidimensional biological state and the intervention response of Fritillaria cirrhosa extract. Gene expression data, protein expression data, metabolic data, and phenotypic state data of the sample to be predicted before intervention with Fritillaria cirrhosa extract were obtained. The normalized difference between each detection index and the pre-intervention detection value of the corresponding detection index in the target response evidence set was calculated to obtain the multidimensional initial state features of the sample to be predicted.

7. The method for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data according to claim 6, characterized in that: Other studies that obtained intervention response prediction results for Fritillaria cirrhosa extract include: The multidimensional initial state features of the sample to be predicted are matched with the multidimensional initial state features of each historical sample in the target response evidence set before intervention, and the proportion of the number of detection indicators that meet the preset difference range to the total number of common detection indicators is calculated. The proportion is determined as the state matching degree between the sample to be predicted and the corresponding historical sample. Historical samples with a state matching degree greater than or equal to a preset state matching threshold are selected as target historical samples. The merged response evidence and independent response evidence corresponding to each target historical sample are extracted, and the extracted response evidence is merged according to the same evidence identifier to obtain the candidate response evidence corresponding to the sample to be predicted. Based on the response distribution sequence corresponding to each candidate response evidence, the sum of the proportions of candidate response evidence appearing in the high response sample group, medium response sample group and low response sample group are counted respectively. The response level corresponding to the response sample group with the largest sum of the proportions of appearance in the group is determined as the target response level of the sample to be predicted, and the intervention response prediction results of Fritillaria cirrhosa extract are obtained.

8. A system for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data, used to implement the method for predicting the intervention response of Fritillaria cirrhosa extract based on multidimensional biological data as described in any one of claims 1-7, characterized in that: The system includes: Response Feature Module: Acquire multidimensional biological data corresponding to historical samples and intervention response results of Fritillaria cirrhosa extract, extract response change features in different data dimensions, and construct a cross-dimensional response feature set according to the sample correspondence; Propagation Chain Module: Based on a cross-dimensional response feature set, this module analyzes the sequential occurrence, direction of change, and cross-dimensional correlation among different response change features, identifies the associated features formed by the continuous propagation of the same response change, and constructs a response evidence propagation chain accordingly. Response Evidence Module: Based on the response evidence propagation chain, multiple response features belonging to the same propagation chain are merged into evidence, the repeated representations of the same response process in different data dimensions are compressed into a single response evidence, and response features that do not form a propagation association are retained as independent response evidence, thus obtaining the target response evidence set; Prediction module: Based on the target response evidence set, a mapping relationship is established between the multidimensional biological state and the intervention response of Fritillaria cirrhosa extract, and the mapping relationship is used to perform response analysis on the sample to be predicted to obtain the prediction result of the intervention response of Fritillaria cirrhosa extract.

Images