A multi-target cross-platform pharmacodynamic data fusion and detection system

Abstract

The application provides a multi-target cross-platform drug efficacy data fusion and detection system, and relates to the technical field of data fusion, and comprises: an inhibition rate acquisition unit, which is used for acquiring standard inhibition rates of a target drug for each target on each detection platform; a weight acquisition unit, which is used for obtaining platform reliability weights of each detection platform according to the standard inhibition rates of the target drug for each target on each detection platform and preset target sensitivity coefficients corresponding to each target; a coefficient determination unit, which is used for obtaining reliability coefficients of each detection platform according to the platform reliability weights of each detection platform and a preset core target combination corresponding to the target drug; and an adaptation degree determination unit, which is used for obtaining a cross-platform data fusion adaptation degree value according to the reliability coefficients of each detection platform and target network features of each detection platform. The application effectively improves the accuracy and reliability of multi-target drug efficacy evaluation.

Description

Technical Field

[0001] The present application relates to the field of data fusion technology, and in particular to a multi-target cross-platform drug efficacy data fusion and detection system. Background Art

[0002] During the development of multi-target drugs, it is necessary to evaluate the drug's effects on multiple targets through various testing platforms (such as in vitro cell experiments, animal models, and molecular docking) to comprehensively determine the drug's actual efficacy. However, the experimental environments and detection principles of different testing platforms vary significantly, resulting in significant dispersion in the efficacy data for the same drug across different testing platforms.

[0003] Specifically, on the one hand, the initial inhibition rate data collected by each platform have dimensional differences due to differences in detection methods, instrument accuracy, etc., and direct fusion will lead to data deviation; on the other hand, different targets have different response sensitivities on different detection platforms, and some platforms may have weaker capabilities to capture the efficacy signals of specific targets. If the platforms are directly given equal weight for data fusion, the true effect of the core target will be obscured; in addition, the existing technology lacks quantitative evaluation of platform reliability, and it is difficult to measure the consistency of data from different detection platforms, resulting in the fusion results being difficult to accurately reflect the actual effect of the drug on the core target combination, increasing the error and uncertainty in the efficacy evaluation of multi-target drugs. Summary of the Invention

[0004] In response to the above technical problems, the present application provides a multi-target cross-platform drug efficacy data fusion and detection system, which at least partially solves the problems existing in the existing technology.

[0005] In a first aspect of the present application, a multi-target cross-platform drug efficacy data fusion and detection system is provided, comprising:

[0006] An inhibition rate acquisition unit is used to obtain the standard inhibition rate of the target drug for each target point on each detection platform; wherein the standard inhibition rate is obtained by standardizing the collected initial inhibition rate; the target drug acts on multiple targets;

[0007] A weight acquisition unit is used to obtain the platform reliability weight of each detection platform based on the standard inhibition rate of the target drug for each target point in each detection platform and the preset target sensitivity coefficient corresponding to each target point;

[0008] A coefficient determination unit, configured to obtain a reliability coefficient for each detection platform based on the platform reliability weight of each detection platform and a preset core target combination corresponding to the target drug;

[0009] The fitness determination unit is used to obtain a cross-platform data fusion fitness value based on the reliability coefficient of each detection platform and the target network characteristics of each detection platform; wherein the target network characteristics of the detection platform include the average target correlation strength and the proportion of strongly correlated target pairs; the absolute value of the correlation coefficient between the targets included in the strongly correlated target pairs is greater than a preset correlation coefficient threshold; the cross-platform data fusion fitness value represents the degree of data consistency between each detection platform.

[0010] This application has at least the following beneficial effects:

[0011] The multi-target cross-platform efficacy data fusion and detection system provided by this application, the inhibition rate acquisition unit eliminates the data deviation caused by dimensional differences between different detection platforms by standardizing the initial inhibition rate, making the inhibition rate data of different detection platforms comparable, and laying a unified data foundation for subsequent weight calculation and data fusion. The weight acquisition unit calculates the platform reliability weight based on the standard inhibition rate and the preset target sensitivity coefficient, fully considering the response differences of different targets on each platform. Since the target sensitivity coefficient reflects the platform's ability to capture specific target signals, the weight obtained based on this can allow platforms that are more sensitive to core targets to occupy a reasonable proportion in the fusion, reducing excessive reliance on low-sensitivity platform data, and making the weight distribution more in line with the actual efficacy detection capabilities of each platform. The coefficient determination unit determines the reliability coefficient based on the platform reliability weight and the preset core target combination, further focusing on the effect of the core target. Finally, the fitness determination unit calculates the cross-platform data fusion fitness value through the reliability coefficient and target network characteristics, realizing a quantitative evaluation of the consistency of data from different detection platforms. Among them, the average target correlation strength and the ratio of strongly correlated targets reflect the synergistic effect pattern between targets. Combined with the reliability coefficient, the fitness value can not only reflect the overall consistency of the data of each platform, but also reflect the stability of the core target combination effect pattern, providing an intuitive basis for judging the credibility of the fusion data, and effectively improving the accuracy and reliability of multi-target drug efficacy evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0013] Figure 1 This is a structural block diagram of the multi-target cross-platform drug efficacy data fusion and detection system provided in the embodiments of the present application. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0015] It should be noted that the terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0016] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on this application, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement an apparatus and / or practice a method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this apparatus and / or practice this method.

[0017] Please refer to Figure 1 As shown, an embodiment of the present application provides a multi-target cross-platform drug efficacy data fusion and detection system 100, which includes:

[0018] The inhibition rate acquisition unit 110 is used to obtain the standard inhibition rate of the target drug for each target point in each detection platform; wherein the standard inhibition rate is obtained by standardizing the collected initial inhibition rate; the target drug acts on multiple targets.

[0019] Specifically, the initial inhibition rate of the target drug for each target on each detection platform is first collected. For example, the target drug: D002, acts on targets X, Y, and Z. The three platforms are the in vitro cell experiment platform, the nude mouse xenograft model platform, and the molecular docking simulation platform. In the in vitro cell experiment platform, the initial inhibition rate of X is 75%, the initial inhibition rate of Y is 40%, and the initial inhibition rate of Z is 20%; the nude mouse xenograft model platform: the initial inhibition rate of X is 60%, the initial inhibition rate of Y is 50%, and the initial inhibition rate of Z is 30%; the molecular docking simulation platform: the initial inhibition rate of X is 85%, the initial inhibition rate of Y is 30%, and the initial inhibition rate of Z is 10%. It should be noted that the initial inhibition rate can be the average of several data points obtained from several experiments of the target drug on a specific detection platform.

[0020] The data normalization process is as follows: first, set the correlation coefficient between targets, for example: XY: 1 (synergy); XZ: 0 (indifference); YZ: -1 (antagonism);

[0021] Secondly, for the inhibition rate, the standardization process is as follows:

[0022] Standard inhibition rate of a target = (initial inhibition rate of the target - minimum inhibition rate of the target on all detection platforms) / (maximum inhibition rate of the target on all detection platforms - minimum inhibition rate of the target on all detection platforms);

[0023] For example, the standard inhibition rate of X in the nude mouse xenograft tumor model platform = (60%-60%) / (85%-60%) = 0; the standard inhibition rate of Y in the in vitro cell experiment platform = (40%-30%) / (50%-30%) = 50%.

[0024] The weight acquisition unit 120 is used to obtain the platform reliability weight of each detection platform according to the standard inhibition rate of the target drug for each target point in each detection platform and the preset target point sensitivity coefficient corresponding to each target point.

[0025] Specifically, the weight acquisition unit includes:

[0026] The coefficient acquisition subunit is used to obtain the preset target sensitivity coefficient corresponding to each target based on the historical data of key known drugs; wherein the preset target sensitivity coefficient represents the response sensitivity of the detection platform to the biological effect of the corresponding target; the key known drug is a known drug that contains at least one key target; the key target is any target acted by the target drug.

[0027] Here, because different detection platforms have different reliability levels for the same target, weighting is necessary to reflect this. For example, a molecular docking simulation platform is more accurate (higher sensitivity) in detecting "binding ability to target X," while a nude mouse xenograft tumor model platform is more sensitive to the "in vivo synergistic effect of X + Y." Therefore, in this example, based on the historical data of key known drugs, a preset target sensitivity coefficient corresponding to each target is obtained. Here, for a single target, based on the historical data of key known drugs, the correlation between the efficacy data of the key known drugs for that target on different detection platforms and the "gold standard" (e.g., core preclinical experimental results) is calculated. The higher the correlation, the higher the sensitivity coefficient (range 0-1).

[0028] The sub-weight acquisition sub-unit is used to obtain the platform reliability weight sub-weight corresponding to each target point acted by the target drug on each detection platform according to the standard inhibition rate of the target drug for each target point on each detection platform and the preset target sensitivity coefficient corresponding to each target point; wherein the platform reliability weight sub-weight is proportional to the standard inhibition rate and the preset target sensitivity coefficient.

[0029] The summing subunit is used to add the platform reliability weight sub-weights corresponding to each target point of the target drug on each detection platform to obtain the platform reliability weight of each detection platform.

[0030] Here, for the target drug, which acts on targets X, Y, and Z, the weight of the target drug in a certain detection platform = the preset target sensitivity coefficient of target X × the inhibition rate of the target drug on target X in the detection platform + the preset target sensitivity coefficient of target Y × the inhibition rate of the target drug on target Y in the detection platform + the preset target sensitivity coefficient of target Z × the inhibition rate of the target drug on target Z in the detection platform.

[0031] The higher the weight of a testing platform, the more reliable its response to the efficacy of the target drug. Conversely, the less reliable it is.

[0032] The coefficient determination unit 130 is used to obtain the reliability coefficient of each detection platform according to the platform reliability weight of each detection platform and the preset core target combination corresponding to the target drug.

[0033] Specifically, the coefficient determination unit includes:

[0034] The coefficient of variation determination subunit is used to obtain the target combination variation coefficient of each detection platform based on the preset core target combination corresponding to the target drug; wherein, the target combination variation coefficient represents the stability of the effect of the detection platform on the preset core target combination.

[0035] Among them, the preset core target combination meets the following conditions: the effect intensity value of each target included is greater than the preset effect intensity value threshold, the corresponding combination contribution is greater than the preset contribution threshold, the absolute value of the correlation coefficient between the included targets is greater than the preset correlation coefficient absolute value threshold, and the pathway co-enrichment rate between the included targets is greater than the preset enrichment threshold.

[0036] First, the drug's effect on each target (e.g., inhibition rate, binding energy) is measured through in vitro experiments or molecular docking techniques to screen for candidate targets that meet the required inhibition criteria. Evaluation metrics include: Inhibition rate (%): The degree of inhibition of the drug on target activity, which must be ≥50% (i.e., half-effective inhibition); Molecular docking binding energy (kcal / mol): This reflects the binding ability of the drug to the target, which must be ≤ -7 kcal / mol (a larger absolute value indicates a stronger binding). For example, if drug D002 has an inhibition rate of 75% and a binding energy of -8.2 kcal / mol on target X, a 60% inhibition rate and a binding energy of -7.5 kcal / mol on target Y, and a 30% inhibition rate and a binding energy of -5.0 kcal / mol on target Z, the initial candidate targets are X and Y.

[0037] Then, through multivariate linear regression analysis, the contribution of each target (or target combination) to comprehensive efficacy indicators (such as tumor reduction rate and cell apoptosis rate) was calculated, and targets with high contributions were retained. Evaluation metrics include: Standardized regression coefficient (β): quantifies the independent contribution of a single target to efficacy; larger absolute values ​​(e.g., ≥0.3) indicate a more significant contribution; Combination contribution (R² increment): The increase in model goodness of fit (R²) after adding a target combination to a single target must be ≥0.2 (i.e., the combination explains at least 20% of the variation in efficacy). For example, if only X is included, R² = 0.4, adding Y = 0.65 (an increment of 0.25), and adding Z = 0.43 (an increment of 0.03) → The contribution of the X + Y combination is significantly higher than that of X + Z, so X + Y is retained first.

[0038] Next, the Pearson correlation coefficient or mutual information entropy is used to calculate the strength of synergistic / antagonistic associations between candidate targets, screening for combinations with clear associations. Evaluation metrics include: Correlation coefficient (r): reflects the linear association between targets, with |r| ≥ 0.5 indicating a strong association (r > 0 indicating synergy, r < 0 indicating antagonism); Pathway co-enrichment rate (%): the proportion of disease-related pathways shared by both targets relative to the total number of pathways, which must be ≥ 30% to ensure biological significance. For example, if the correlation coefficient r between X and Y is 0.65, they are co-enriched in three tumor-related pathways (accounting for 40% of the total pathways); if the r between X and Z is 0.2, with a co-enrichment rate of 10%, the strength of the association and biological significance of X + Y are more significant.

[0039] In summary, the preset core target combination is determined to meet the following conditions: the effect intensity value of each target included is greater than the preset effect intensity value threshold, the corresponding combination contribution is greater than the preset contribution threshold, the absolute value of the correlation coefficient between the included targets is greater than the preset correlation coefficient absolute value threshold, and the pathway co-enrichment rate between the included targets is greater than the preset enrichment threshold.

[0040] In this embodiment, if the target drug corresponds to the preset core target combination X+Y, the target combination coefficient of variation (CV) is calculated using the formula (standard deviation σ / mean μ) × 100%. As an example: For a specific assay platform, the inhibition rate data for targets X and Y is obtained for several samples. Then, the mean of the X+Y combination effect values ​​corresponding to these samples is calculated. The X+Y combination effect value = the inhibition rate of the target drug on target X in that assay platform + the inhibition rate of the target drug on target Y in that assay platform. The standard deviation is then calculated from the mean to obtain the target combination coefficient of variation (CV) for each assay platform.

[0041] The error determination subunit is used to obtain the intra-platform error of each detection platform based on the preset core target combination corresponding to the target drug; wherein the intra-platform error represents the reliability of the transmission of the effect of the detection platform on the preset core target combination.

[0042] Specifically, each detection platform also has a corresponding specific drug indicator. A simple linear regression model is constructed with the X+Y combined effect value as the independent variable (x) and the detection platform-specific drug efficacy indicator (such as cell apoptosis rate) as the dependent variable (y). The predicted y value of each sample among several samples is calculated according to the simple linear regression model, and then the mean square error is obtained based on the actual y value of each sample.

[0043] The coefficient determination subunit is used to obtain the reliability coefficient of each detection platform based on the platform reliability weight of each detection platform, the target combination variation coefficient of each detection platform and the intra-platform error of each detection platform.

[0044] Specifically, the reliability coefficient Ki of the i-th detection platform meets the following characteristics:

[0045] Ki=Wi / MAX(W)×(1-(P / (CV×Y)));

[0046] Where Wi is the platform reliability weight of the i-th detection platform, MAX() is the maximum value determination function; P is the intra-platform error; CV is the coefficient of variation of the target combination; and Y is the fixed coefficient.

[0047] Here, i = 1, 2, ..., n; MAX(W) is the maximum reliability weight of the n detection platforms; CV × Y represents the error threshold, where Y can be a fixed coefficient determined based on statistical laws. A higher reliability coefficient indicates more stable data.

[0048] The target combination coefficient of variation reflects the natural dispersion of target combination data (e.g., fluctuations in the X+Y effect size across several samples), representing the "unavoidable range of fluctuation" determined by the inherent characteristics of the data. Multiplying by the fixed coefficient Y reflects the statistical principle that "model error should not exceed Y (e.g., 30%) of the natural fluctuation in the data." If the model error (MSE) exceeds this ratio, it indicates that the model not only fails to capture the data pattern but also introduces additional error, significantly reducing the reliability of the platform.

[0049] Furthermore, the core function of (P / (CV×Y)) is to normalize model error. When P is significantly less than the error threshold (e.g., P = 0.03, threshold = 0.045 in this example), the ratio is less than 1, indicating that the model error is within the acceptable fluctuation range of the data and that the platform can stably transmit the efficacy signal of the target combination. If P exceeds the error threshold (ratio > 1), this means that the model error has exceeded the reasonable fluctuation of the data itself. In this case, the platform data may contain anomalies (such as detection errors and environmental interference), and its reliability must be compromised. This normalization process eliminates the differences in data magnitude between different detection platforms, making the "degree of error deviation" a metric that can be compared across platforms. The formula uses "1-ratio" to map the result to a reasonable range of 0 to 1, and the value is positively correlated with reliability. As an example: when P = 0 (perfect model fit), the reliability coefficient = 1, indicating that the platform data is completely reliable.

[0050] This implementation uses data stability (CV) to define a reasonable error boundary (threshold). The reliability of the platform's efficacy signal is inferred from the deviation of the model error (P) from this boundary. This avoids the problem of ignoring natural data fluctuations when using P alone, while ensuring objectivity through a fixed coefficient (Y). The resulting coefficient truly reflects the trustworthiness of the platform data, providing a quantifiable quality basis for multi-platform data integration.

[0051] The fitness determination unit 140 is used to obtain a cross-platform data fusion fitness value based on the reliability coefficient of each detection platform and the target network characteristics of each detection platform; wherein the target network characteristics of the detection platform include the average target correlation strength and the proportion of strongly correlated target pairs; the absolute value of the correlation coefficient between the targets included in the strongly correlated target pairs is greater than a preset correlation coefficient threshold; the cross-platform data fusion fitness value represents the degree of data consistency between each detection platform.

[0052] Specifically, the adaptability determination unit includes:

[0053] The association data acquisition subunit is used to obtain the average association strength of the targets and the ratio of strongly associated targets for each detection platform.

[0054] Among them, the average target association strength is the average of the association coefficients of all target pairs (such as XY, XZ, YZ) in the detection platform. The association coefficient is the association coefficient mentioned above. As an example: XY: 1 (synergy); XZ: 0 (indifference); YZ: -1 (antagonism); the average target association strength reflects the intensity of the overall synergistic / antagonistic effect between targets. In one embodiment, the target average association strength value range is -1~1, positive values ​​indicate synergy, negative values ​​indicate antagonism, and the larger the absolute value, the stronger the association.

[0055] The absolute value of the correlation coefficient between the targets included in the strongly correlated target pairs is greater than the preset correlation coefficient threshold; the proportion of strongly correlated target pairs is the proportion of strongly correlated target pairs to the total target pairs (range 0~100%). The proportion of strongly correlated target pairs reflects the density of the correlation between the core targets in the platform. The higher the proportion, the clearer the role of the key targets.

[0056] The correction subunit is used to correct the target average correlation strength and the strong correlation target ratio of the corresponding detection platform according to the reliability coefficient of each detection platform to obtain the corrected target average correlation strength and the corrected strong correlation target ratio.

[0057] Here, corrections are made one by one according to the reliability coefficient of each detection platform. As an example, the specific correction process is: the reliability coefficient of a certain detection platform × the average target association strength of the detection platform = the corrected target average association strength of the detection platform; the reliability coefficient of a certain detection platform × the ratio of strongly associated targets of the detection platform = the corrected associated target ratio of the detection platform.

[0058] This embodiment corrects the original eigenvalues ​​using the platform reliability coefficient, so that the features of platforms with high reliability (such as molecular docking) account for a higher proportion in subsequent calculations, reducing the interference of low-reliability platforms (such as the nude mouse model), reflecting the logic that the higher the data credibility, the greater the weight.

[0059] The fitness determination subunit is used to obtain the cross-platform data fusion fitness value based on the corrected target average association strength and the corrected strong association target ratio of each detection platform.

[0060] Here, we first obtain the average of the corrected target average association strengths and the average of the corrected strongly associated target ratios across all platforms. Next, we calculate the sum of the absolute values ​​of each platform's corrected target average association strengths and the corrected strongly associated target ratios, along with these two averages. Finally, we calculate the similarity score for each platform and the cross-platform data fusion fitness value.

[0061] The similarity score for a given test platform is calculated as 1 minus (the sum of the absolute differences between the test platform and the weighted feature mean divided by 2). The sum of the absolute differences between the platform's corrected average target association strength and the corrected proportion of strongly associated targets, respectively, and the average of the corrected average target association strength and the average of the corrected proportion of strongly associated targets across all test platforms, respectively. The theoretical maximum deviation between the two feature dimensions (corrected average target association strength and corrected proportion of strongly associated targets) is 2. After dividing by 2, the result of the sum of the absolute differences divided by 2 is mapped to the range of 0–1, ensuring that subsequent similarity scores remain between 0 and 1 (which is consistent with the typical range of proportion indicators). Subtracting the deviation ratio from 1 results in a positive correlation between the score and consistency. That is, when a test platform is completely consistent with the overall trend (the sum of the absolute differences = 0), the score is 1, indicating perfect agreement; when the deviation reaches the maximum limit (the sum of the absolute differences = 2), the score is 0, indicating complete disagreement. This forward mapping allows the scores to directly reflect the consistency of the platform data (without additional interpretation), facilitating subsequent comprehensive judgments (for example, by directly comparing the scores of different testing platforms, it can be determined which platform is more in line with the overall trend).

[0062] The final cross-platform data fusion fitness value is the average of the three platforms. This is because the reliability of each detection platform has been adjusted based on its reliability coefficient. Therefore, the final cross-platform data fusion fitness value gives the characteristic values ​​of high-quality detection platforms a higher weight in the overall mean. This also prevents bias from individual platforms from dominating the results. For example, if a platform scores low due to accidental error (e.g., 0.83 in the nude mouse model), but most platforms score higher (0.94, 0.93), the average value still reflects good overall consistency. Conversely, if most platforms score below the threshold, or if a single platform scores very low, the average value will drop significantly, fully reflecting anomalous data. The final comprehensive index (range 0–1) indicates that higher values ​​indicate greater consistency in target network features across multiple platforms.

[0063] In this embodiment, the setting of the cross-platform data fusion fitness value not only highlights the role of high-quality data through weighted processing, but also achieves accurate quantification of consistency through two-dimensional evaluation and standardized transformation. Ultimately, individual deviations are balanced through averaging, and while filtering noise, the stability of the drug action mode in different detection platforms is objectively reflected.

[0064] In an exemplary embodiment of the present application, after the adaptability determination unit, the system further includes:

[0065] A relationship graph generation unit is used to generate a target relationship network graph for each detection platform if the cross-platform data fusion fitness value is less than a preset fitness threshold; wherein the target relationship network graph includes nodes and edges; the nodes in the target relationship network graph are the average inhibition rates of each target acted by the target drug in the corresponding detection platform; the edges are the association relationships between targets; the edges have thickness features and color features, the thickness features represent the association strength; the color features represent the association type.

[0066] Specifically, the core target combination of a multi-target drug (e.g., X+Y+Z) will form a stable association network across different testing platforms (e.g., strong synergy between X and Y, weak antagonism between Y and Z). However, platforms with anomalous data will exhibit "topological structure mutations" that are inconsistent with the majority (e.g., X and Y, which should be strongly synergistic, become antagonistic). By visually comparing the differences in network structure, the anomalous platform can be quickly identified. In this example, targets are used as nodes, and lines represent the associations between targets. The thickness of the line indicates the strength of the association (thicker, stronger), and the color indicates the type of association (red for synergy, blue for antagonism).

[0067] A conversion unit is used to convert all edges in the target relationship network diagram of each detection platform into a triple matrix; wherein each triple matrix is ​​in the form of (first target, second target, association feature of the first target and the second target); the association feature of the first target and the second target is a numerical value; the positive or negative value represents the type of association; and the absolute value of the numerical value represents the strength of the association.

[0068] Specifically, the association relationship of the target combination is converted into a "triplet matrix": the format of the triplet matrix is ​​(first target, second target, association feature of the first and second targets), where the "association feature" is encoded with a numerical value. For example, synergy is positive and antagonism is negative, and the absolute value represents the strength, such as strong synergy = +2, weak synergy = +1, no association = 0, weak antagonism = -1, and strong antagonism = -2. For example, the matrix of an in vitro cell-based assay platform contains (X, Y, +2), (Y, Z, -1); the matrix of a molecular docking assay platform contains (X, Y, +2), (Y, Z, -1); and the matrix of a nude mouse model platform contains (X, Y, -2), (Y, Z, +1).

[0069] The outlier determination unit is used to obtain the outlier value of each detection platform based on the triple matrix of each detection platform; wherein the outlier value is the sum of the differences between the correlation features of each target pair in the corresponding detection platform and the mode of the correlation features of each target pair.

[0070] Specifically, for the same target pair (e.g., X, Y) across all platforms, determine the mode of each correlation feature. The mode of a correlation feature is the correlation feature that appears the most times (e.g., if the correlation feature of X, Y is +2 in 80% of the platforms), then the mode of the correlation feature of X, Y is +2.

[0071] Furthermore, for each platform, the sum of the differences between the association features and the mode of the association features of all target pairs is calculated. For example: if the (X, Y) association feature of a certain detection platform is -2 (different from the mode of the association features of (X, Y): +2), then the difference value is 4 (|+2-(-2)|); if the (Y, Z) association feature is +1 (different from the mode of the association features of (Y, Z): -1), the difference value is 2, and the total difference = 4+2=6.

[0072] The larger the sum of the differences, the more likely the data is anomalous.

[0073] The abnormal platform determination unit is used to determine the detection platform with the largest abnormal value as the abnormal detection platform.

[0074] In this embodiment, the abstract target association data is converted into an intuitive visual network diagram through the relationship graph generation unit, and the association pattern between the targets of each platform is clearly presented with the help of the characteristics of nodes and edges; the conversion unit quantifies the network graph information into a triple matrix, realizes the standardized representation of the association relationship, and provides a unified data basis for subsequent calculations; the outlier determination unit objectively quantifies the degree of deviation of the platform data from the overall trend by calculating the sum of the differences between the association characteristics of each platform and the mode; the abnormal platform determination unit accurately locates the abnormal platform based on the outlier value. The whole process not only retains the intuitive characteristics of multi-target association, but also avoids subjective judgment bias through quantitative calculation, thereby efficiently and accurately identifying the abnormal detection platform from the cross-platform data, providing a strong guarantee for the reliability of multi-target drug efficacy evaluation.

[0075] In an exemplary embodiment of the present application, after the fitness determination unit, the system further includes:

[0076] The fusion completion unit is used to determine that the fusion is completed and there is no abnormal detection platform if the cross-platform data fusion fitness value is less than a preset fitness threshold.

[0077] In this embodiment, when the cross-platform data fusion fitness value is less than the preset fitness threshold, it is directly determined that the fusion is completed and there is no abnormal detection platform. This setting objectively confirms the consistency of the data of each platform in the core target association pattern through the quantitative indicator of the fitness value, without the need to perform additional abnormality investigation process. It avoids unnecessary verification of platforms that conform to the overall trend and reduces the consumption of system computing resources. It can also quickly complete data fusion and output reliable results, providing an efficient and reliable conclusion basis for the cross-platform comprehensive evaluation of the efficacy of multi-target drugs.

[0078] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A multi-target cross-platform drug efficacy data fusion and detection system, characterized by: The system comprises: An inhibition rate acquisition unit is used to obtain the standard inhibition rate of the target drug for each target point on each detection platform; wherein the standard inhibition rate is obtained by standardizing the collected initial inhibition rate; the target drug acts on multiple targets; A weight acquisition unit is used to obtain the platform reliability weight of each detection platform based on the standard inhibition rate of the target drug for each target point in each detection platform and the preset target sensitivity coefficient corresponding to each target point; A coefficient determination unit, configured to obtain a reliability coefficient for each detection platform based on the platform reliability weight of each detection platform and a preset core target combination corresponding to the target drug; The fitness determination unit is used to obtain a cross-platform data fusion fitness value based on the reliability coefficient of each detection platform and the target network characteristics of each detection platform; wherein the target network characteristics of the detection platform include the average target correlation strength and the proportion of strongly correlated target pairs; the absolute value of the correlation coefficient between the targets included in the strongly correlated target pairs is greater than a preset correlation coefficient threshold; the cross-platform data fusion fitness value represents the degree of data consistency between each detection platform.

2. The multi-target cross-platform drug efficacy data fusion and detection system according to claim 1, characterized in that: The weight acquisition unit includes: The coefficient acquisition subunit is used to obtain the preset target sensitivity coefficient corresponding to each target based on the historical data of key known drugs; wherein the preset target sensitivity coefficient represents the sensitivity of the detection platform to the biological effect of the corresponding target; a key known drug is a known drug containing at least one key target; and a key target is any target acted by the target drug; The sub-weight acquisition sub-unit is used to obtain the platform reliability weight sub-weight corresponding to each target of the target drug on each detection platform based on the standard inhibition rate of the target drug for each target on each detection platform and the preset target sensitivity coefficient corresponding to each target; wherein the platform reliability weight sub-weight is proportional to the standard inhibition rate and the preset target sensitivity coefficient; The summing subunit is used to add the platform reliability weight sub-weights corresponding to each target point of the target drug on each detection platform to obtain the platform reliability weight of each detection platform.

3. The multi-target cross-platform drug efficacy data fusion and detection system according to claim 1, characterized in that: The coefficient determination unit includes: The coefficient of variation determination subunit is used to obtain the target combination variation coefficient of each detection platform based on the preset core target combination corresponding to the target drug; wherein the target combination variation coefficient represents the stability of the detection platform's effect on the preset core target combination; An error determination subunit is used to obtain the intra-platform error of each detection platform based on the preset core target combination corresponding to the target drug; wherein the intra-platform error represents the reliability of the detection platform in transmitting the effect of the preset core target combination; The coefficient determination subunit is used to obtain the reliability coefficient of each detection platform based on the platform reliability weight of each detection platform, the target combination variation coefficient of each detection platform and the intra-platform error of each detection platform.

4. The multi-target cross-platform drug efficacy data fusion and detection system according to claim 3, characterized in that: The preset core target combination meets the following conditions: the effect intensity value of each target included is greater than the preset effect intensity value threshold, the corresponding combination contribution is greater than the preset contribution threshold, the absolute value of the correlation coefficient between the included targets is greater than the preset correlation coefficient absolute value threshold, and the pathway co-enrichment rate between the included targets is greater than the preset enrichment threshold.

5. The multi-target cross-platform drug efficacy data fusion and detection system according to claim 3, characterized in that: The reliability coefficient K of the i-th detection platform i Meet the following characteristics: K i =W i / MAX(W)×(1-(P / (CV×Y))); Among them, W i is the platform reliability weight of the i-th detection platform, MAX() is the maximum value determination function; P is the intra-platform error; CV is the target combination variation coefficient; Y is the fixed coefficient.

6. The multi-target cross-platform drug efficacy data fusion and detection system according to claim 1, characterized in that: The fitness determination unit includes: The association data acquisition subunit is used to obtain the average association strength of the targets and the ratio of strongly associated targets for each detection platform; A correction subunit, configured to correct the target average association strength and the strong correlation target ratio of the corresponding detection platform according to the reliability coefficient of each detection platform, so as to obtain a corrected target average association strength and a corrected strong correlation target ratio; The fitness determination subunit is used to obtain the cross-platform data fusion fitness value based on the corrected target average association strength and the corrected strong association target ratio of each detection platform.

7. The multi-target cross-platform drug efficacy data fusion and detection system according to claim 1, characterized in that: After the fitness determination unit, the system further includes: A relationship graph generating unit is configured to generate a target relationship network graph for each detection platform if the cross-platform data fusion fitness value is less than a preset fitness threshold; wherein the target relationship network graph includes nodes and edges; the nodes in the target relationship network graph are the mean inhibition rates of each target acted by the target drug in the corresponding detection platform; the edges are the association relationships between the targets; the edges have thickness features and color features, the thickness features represent the strength of the association; and the color features represent the type of association; A conversion unit is configured to convert all edges in the target relationship network diagram of each detection platform into a triple matrix; wherein each triple matrix is ​​in the form of (first target, second target, correlation feature between the first target and the second target); the correlation feature between the first target and the second target is a numerical value; the positive or negative value of the numerical value indicates the type of correlation; and the absolute value of the numerical value indicates the strength of the correlation; an outlier determination unit, configured to obtain an outlier value for each detection platform based on the triple matrix of each detection platform; wherein the outlier value is the sum of the differences between the correlation feature of each target pair in the corresponding detection platform and the mode of the correlation feature of each target pair; The abnormal platform determination unit is used to determine the detection platform with the largest abnormal value as the abnormal detection platform.

8. The multi-target cross-platform drug efficacy data fusion and detection system according to claim 1, characterized in that: After the fitness determination unit, the system further includes: The fusion completion unit is used to determine that the fusion is completed and there is no abnormal detection platform if the cross-platform data fusion fitness value is less than a preset fitness threshold.

Images