A system, apparatus, method, and storage medium for screening disturbance response objects.

Abstract

This disclosure provides a system, apparatus, method, and storage medium for screening perturbation response objects. The system may include a processor configured to: for each of at least two preset perturbation conditions, determine first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition; acquire an object anomaly probability; obtain second expression data of at least one cell under the perturbation condition based on the object anomaly probability and the first expression data of each cell under the perturbation condition; further, perform loss calculation based on the second expression data corresponding to the at least two perturbation conditions to adjust the object anomaly probability, obtain a target anomaly probability, and screen perturbation response objects from at least one object based on the target anomaly probability. The technical solution of this disclosure can improve the screening accuracy of perturbation response objects.

Description

Technical Field

[0001] This disclosure relates to the biological field, and in particular to a system, apparatus, method, and storage medium for screening disturbance response objects. Background Technology

[0002] Simulating and predicting biological responses to perturbations is an important task in systems biology. In recent years, with the development of sequencing technology, research on biological perturbation responses has penetrated to the single-cell level, and a major aspect of exploring the effects of perturbations on cells is the screening of perturbation-responsive genes.

[0003] However, the current screening scheme has low screening accuracy and needs to be improved. Summary of the Invention

[0004] This disclosure provides a system, apparatus, method, and storage medium for screening disturbance response objects, in order to improve the accuracy of the screening scheme for disturbance response objects.

[0005] According to one aspect of this disclosure, a system for screening disturbance response objects is provided, which may include:

[0006] The processor is configured to:

[0007] For each of at least two pre-defined perturbation conditions, determine the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition, wherein the first expression data is used to express at least one object associated with each cell;

[0008] Obtain the object anomaly probability, where the object anomaly probability is used to characterize the probability that each of at least one object is affected by at least two perturbation conditions;

[0009] Based on the abnormality probability of the object and the first expression data of each cell under the perturbation condition, the second expression data of at least one cell under the perturbation condition is obtained;

[0010] Based on the second expression data corresponding to at least two perturbation conditions, loss calculation is performed to adjust the object anomaly probability to obtain the target anomaly probability, and perturbation response objects are selected from at least one object based on the target anomaly probability.

[0011] According to another aspect of this disclosure, a filtering device for disturbance response objects is provided, configured in a processor, the device including:

[0012] The first expression data determination module is used to determine, for each of the pre-set at least two perturbation conditions, the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition, wherein the first expression data is used to express at least one object associated with each cell;

[0013] The object anomaly probability acquisition module is used to acquire the object anomaly probability, wherein the object anomaly probability is used to characterize the probability that each of at least one object is affected by at least two perturbation conditions.

[0014] The second expression data acquisition module is used to obtain the second expression data of at least one cell under perturbation conditions based on the object's abnormality probability and the first expression data of each cell under perturbation conditions.

[0015] The disturbance response object filtering module is used to perform loss calculation based on the second expression data corresponding to at least two disturbance conditions to adjust the object anomaly probability, obtain the target anomaly probability, and filter the disturbance response object from at least one object based on the target anomaly probability.

[0016] According to another aspect of this disclosure, a method for filtering disturbance response objects is provided, which may include:

[0017] For each of at least two pre-defined perturbation conditions, determine the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition, wherein the first expression data is used to express at least one object associated with each cell;

[0018] Obtain the object anomaly probability, where the object anomaly probability is used to characterize the probability that each of at least one object is affected by at least two perturbation conditions;

[0019] Based on the abnormality probability of the object and the first expression data of each cell under the perturbation condition, the second expression data of at least one cell under the perturbation condition is obtained;

[0020] Based on the second expression data corresponding to at least two perturbation conditions, loss calculation is performed to adjust the object anomaly probability to obtain the target anomaly probability, and perturbation response objects are selected from at least one object based on the target anomaly probability.

[0021] According to another aspect of this disclosure, a computer-readable storage medium is provided that stores computer instructions thereon for causing a processor to execute and implement the method for filtering disturbance response objects provided in any embodiment of this disclosure.

[0022] The technical solution in this embodiment addresses at least two preset perturbation conditions by determining the first expression data of each cell in at least one cell corresponding to the perturbation condition. This first expression data expresses at least one object associated with each cell. An object anomaly probability is obtained, characterizing the probability that each object in the at least one object is affected by at least two perturbation conditions. Based on the object anomaly probability and the first expression data of each cell under the perturbation condition, second expression data of the at least one cell under the perturbation condition is obtained. Then, based on the second expression data corresponding to the at least two perturbation conditions, a loss calculation is performed to adjust the object anomaly probability, resulting in a target anomaly probability. Finally, perturbation-responsive objects are selected from the at least one object based on the target anomaly probability. This technical solution leverages the stability of the first expression data of objects unaffected by perturbation under different perturbation conditions in biology, i.e., the independence between the first expression data and the perturbation conditions. By obtaining the second expression data corresponding to different perturbation conditions through the object anomaly probability and the first expression data, and then learning the object anomaly probability based on this second expression data, the perturbation-responsive object selection process is achieved based on the learned target anomaly probability, thereby improving the selection accuracy.

[0023] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is an execution flowchart of a processor in a disturbance response object screening system provided according to an embodiment of the present disclosure;

[0026] Figure 2 This is an execution flowchart of a processor in a screening system for another disturbance response object provided according to an embodiment of the present disclosure;

[0027] Figure 3 This is an execution flowchart of a processor in a screening system for another disturbance response object provided according to an embodiment of the present disclosure;

[0028] Figure 4 This is a structural block diagram of a disturbance response object screening device provided according to an embodiment of the present disclosure;

[0029] Figure 5 This is a schematic diagram of the structure of an electronic device that implements the disturbance response object screening system in the embodiments of this disclosure. Detailed Implementation

[0030] To enable those skilled in the art to better understand the present disclosure, the technical solutions of the present disclosure will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present disclosure, and not all embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present disclosure.

[0031] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The same applies to "target," "original," etc., and will not be repeated here. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0032] Figure 1 This is a flowchart illustrating the steps configured in a processor within a perturbation response object screening system provided in this embodiment of the present disclosure. This embodiment is applicable to investigating the effects of perturbations on cells, particularly by screening perturbation response objects from at least one object associated with the cell. The object type of the perturbation response object can be a gene, protein, or chromatin accessibility, etc. This step can be performed by the perturbation response object screening device provided in this embodiment of the present disclosure, which can be implemented in software and / or hardware and can be integrated into the processor.

[0033] See Figure 1 The disturbance response object screening system provided in this disclosure embodiment may include a processor, which may be configured to perform the following steps:

[0034] S110. For each of the preset at least two perturbation conditions, determine the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition, wherein the first expression data is used to express at least one object associated with each cell.

[0035] The perturbation conditions can be understood as pre-defined conditions used to disturb the cells. Examples include being stimulated by or not being stimulated by cancerous growth, being stimulated by or not being stimulated by gene editing (such as CRISPR), or being stimulated by or not being stimulated by small molecule drugs. No specific limitation is made here. The number of perturbation conditions can be two or more, which can be set according to the actual situation; no specific limitation is made here.

[0036] For each of the at least two perturbation conditions, the cell corresponding to the perturbation condition can be understood as a cell located under the perturbation condition, and its number can be one, two, or more. Based on this, for each cell in the at least one cell corresponding to the perturbation condition, first expression data of the cell under the perturbation condition is determined. This first expression data can be used to express at least one object associated with all at least one cell located under the perturbation condition, such as at least one gene, at least one protein, or at least one chromatin accessibility. In practical applications, these three can be detected individually or at least two of them can be detected simultaneously. Therefore, combined with subsequent steps, the effect of perturbation on cells can be mutually verified by screening different types of perturbation response objects. Optionally, the first expression data can be understood as the result obtained by detecting (e.g., sequencing) cells located under the perturbation condition, or as the result obtained by further encoding the detected result, etc., without specific limitations here; further optionally, the first expression data can be represented by a count matrix.

[0037] S120. Obtain the object anomaly probability, wherein the object anomaly probability can be used to characterize the probability that each of at least one object is affected by at least two perturbation conditions.

[0038] The object anomaly probability can be used to characterize the probability that each of at least one object is affected by at least two perturbation conditions, or in other words, the probability that each object belongs to a perturbation response object. The perturbation response object can be understood as a differentially expressed object affected by perturbation conditions, or an object abnormally expressed by external perturbations. Specifically, the object anomaly probability may include the anomaly expression probability corresponding to each object. Based on this, for each of the at least one anomaly expression probabilities, the anomaly expression probability can be used to characterize the probability that the object corresponding to it is affected by at least two perturbation conditions, or in other words, the probability that the object belongs to a perturbation response object. In practical applications, optionally, the object anomaly probability M can be represented as M∈[0,1]. d , where d represents the number of at least one object.

[0039] S130. Based on the abnormality probability of the object and the first expression data of each cell under the perturbation condition, obtain the second expression data of at least one cell under the perturbation condition.

[0040] Specifically, based on the object's anomalous probability and the first expression data of each cell under perturbation conditions, second expression data of at least one cell under perturbation conditions is obtained. For example, the interaction result between the object's anomalous probability and the first expression data of each cell under perturbation conditions can be directly used as the second expression data; alternatively, the interaction result can be encoded, and the resulting encoded result can be used as the second expression data; furthermore, the first expression data of each cell under perturbation conditions can be encoded, and the interaction result between the resulting encoded result and the object's anomalous probability can be used as the second expression data; of course, other methods can also be used to obtain the second expression data, without specific limitations here. Thus, it can be seen that the first expression data can be used to characterize the expression data of each cell (i.e., an individual) under perturbation conditions, and the second expression data can be used to characterize the expression data of at least one cell (i.e., the whole body) under perturbation conditions.

[0041] S140. Based on the second expression data corresponding to at least two perturbation conditions, perform loss calculation to adjust the object anomaly probability to obtain the target anomaly probability, and select the perturbation response object from at least one object based on the target anomaly probability.

[0042] Since the above three steps process each of the at least two perturbation conditions separately, second expression data corresponding to each of the at least two perturbation conditions can be obtained. In short, second expression data corresponding to at least two perturbation conditions can be obtained. Based on this, further loss calculation can be performed using the second expression data corresponding to the at least two perturbation conditions to learn the anomalous expression probability of each object, and the object anomalous probability can be adjusted accordingly to obtain the target anomalous probability. Typically, the loss calculation result of the second expression data obtained based on the target anomalous probability satisfies a preset learning termination condition. Further, perturbation response objects are selected from at least one object based on the target anomalous probability.

[0043] It should be noted that a counterfactual problem exists in biological knowledge: for each cell, only its first expression data under a single perturbation condition can be collected, not its first expression data under two or more perturbation conditions. Therefore, it is impossible to screen perturbation response objects by directly comparing the first expression data of the same cell under different perturbation conditions. To address this, the embodiments of this disclosure consider that the first expression data of objects unaffected by perturbation are usually highly similar under different perturbation conditions; that is, the first expression data of these objects under different perturbation conditions is stable, or in other words, the first expression data is independent of the perturbation conditions. Therefore, perturbation response objects can be screened based on the second expression data corresponding to at least two perturbation conditions, i.e., based on this characterization independence. This solves the aforementioned counterfactual problem and improves the screening accuracy of perturbation response objects.

[0044] In light of the application scenarios that may be involved in the embodiments of this disclosure, taking, for example, the object type being a gene, the above technical solution can accurately identify genes that are abnormally expressed due to external perturbation from all genes in single-cell RNA sequencing data containing perturbation tags.

[0045] The technical solution in this embodiment addresses at least two preset perturbation conditions by determining the first expression data of each cell in at least one cell corresponding to the perturbation condition. This first expression data expresses at least one object associated with each cell. An object anomaly probability is obtained, characterizing the probability that each object in the at least one object is affected by at least two perturbation conditions. Based on the object anomaly probability and the first expression data of each cell under the perturbation condition, second expression data of the at least one cell under the perturbation condition is obtained. Then, based on the second expression data corresponding to the at least two perturbation conditions, a loss calculation is performed to adjust the object anomaly probability, resulting in a target anomaly probability. Finally, perturbation-responsive objects are selected from the at least one object based on the target anomaly probability. This technical solution leverages the stability of the first expression data of objects unaffected by perturbation under different perturbation conditions in biology, i.e., the independence between the first expression data and the perturbation conditions. By obtaining the second expression data corresponding to different perturbation conditions through the object anomaly probability and the first expression data, and then learning the object anomaly probability based on this second expression data, the perturbation-responsive object selection process is achieved based on the learned target anomaly probability, thereby improving the selection accuracy.

[0046] An optional technical solution is that the target anomaly probability includes anomaly expression probabilities corresponding to at least one object;

[0047] Based on the target anomaly probability, perturbation response objects are selected from at least one object, including:

[0048] The probabilities of at least one anomalous expression are sorted, and perturbation response objects are selected from at least one object based on the sorting results.

[0049] Each of the at least one anomalous expression probabilities can be used to characterize the probability that the object corresponding to it is affected by at least two perturbation conditions, or in other words, the probability that the object belongs to the perturbation response object. Therefore, the at least one anomalous expression probabilities can be sorted, and perturbation response objects can be selected from the at least one object based on the sorting results. For example, taking sorting from largest to smallest, the objects corresponding to the anomalous expression probabilities ranked in the first preset position or the first preset proportion can be used as perturbation response objects, or the objects corresponding to the anomalous expression probabilities ranked before the target expression probability can be used as perturbation response objects. The target expression probability can be understood as the first anomalous expression probability in the sorting results that is less than the preset expression probability, which can be understood as a pre-set critical probability for selecting perturbation response objects. Of course, when sorting from smallest to largest, the reverse is true, which will not be elaborated further here. The above technical solution achieves the effect of rapid selection of perturbation response objects by sorting the anomalous expression probabilities corresponding to at least one object.

[0050] Based on this, optionally, the number of disturbance response objects is at least two, and the aforementioned processor is further configured to:

[0051] Based on the anomalous expression probabilities corresponding to at least two disturbance response objects, determine the priority of each of the at least two disturbance response objects for the conditional responses to at least two disturbances.

[0052] The above technical solution can quantify the priority of each disturbance response object to the disturbance response, thereby helping to identify the disturbance response object with the greatest response to the disturbance response among at least two disturbance response objects.

[0053] Figure 2 This is a flowchart illustrating the steps configured in a processor within another perturbation response object screening system provided in this disclosure embodiment. This embodiment is an optimization based on the aforementioned technical solutions. Optionally, in this embodiment, the processor is further configured to: determine the normal probability of an object corresponding to the object's anomalous probability, wherein the normal probability of an object may include the probability that each object is unaffected by at least two perturbation conditions; and obtain second expression data of at least one cell under perturbation conditions based on the object's anomalous probability and the first expression data of each cell under perturbation conditions, which may include: for each cell in the at least one cell, obtaining probability data of the cell under perturbation conditions based on the cell's first expression data under perturbation conditions and the object's normal probability; and obtaining second expression data of at least one cell under perturbation conditions based on the probability data of each cell under perturbation conditions. Explanations of terms identical or corresponding to those in the aforementioned embodiments will not be repeated here.

[0054] See Figure 2The processor in this embodiment can be configured to perform the following steps:

[0055] S210. For each of the preset at least two perturbation conditions, determine the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition, wherein the first expression data is used to express at least one object associated with each cell.

[0056] S220. Obtain the object anomaly probability and determine the object normal probability corresponding to the object anomaly probability, wherein the object anomaly probability can be used to characterize the probability that each of at least one object is affected by at least two perturbation conditions, and the object normal probability can be used to characterize the probability that each object is not affected by at least two perturbation conditions.

[0057] Here, the normal probability of an object corresponds to the abnormal probability of an object, and can be used to characterize the probability that each object is not affected by at least two perturbation conditions. For example, in the case of M∈[0,1] d In representing the probability of an object being abnormal, it can be expressed as 1 - M ∈ [0,1]. d This indicates the normal probability of an object.

[0058] S230. For each cell in at least one cell, obtain the probability data of the cell under perturbation conditions based on the cell's first expression data under perturbation conditions and the normal probability of the object.

[0059] Specifically, for each cell in at least one cell, the probability data of the cell under perturbation conditions can be obtained based on the cell's first expression data under perturbation conditions and the normal probability of the object. For example, assuming that X represents the cell's first expression data under perturbation conditions, the above probability data can be expressed as X⊙(1-M), where ⊙ represents the Hadamard product (i.e., dot product), and both X and M are 1xd-dimensional. Of course, probability data can also be obtained through other methods, which are not specifically limited here.

[0060] S240. Based on the probability data of each cell under perturbation conditions, obtain the second expression data of at least one cell under perturbation conditions.

[0061] In this context, having obtained probability data for each cell in at least one cell under perturbation conditions, second expression data for at least one cell under perturbation conditions can be obtained based on this probability data. For example, the average of these probability data can be calculated and used as the second expression data. Alternatively, each probability data point can be encoded separately, and the average of the resulting encodings can be used as the second expression data. That is, the second expression data represents the average value of different cells under perturbation conditions. For example, this can be achieved through... This represents the second set of data, where p i This represents the i-th perturbation condition currently being processed, where i = 1, ..., k, meaning there are a total of k perturbation conditions. θ This indicates an encoder, which encodes probability data, and This means averaging, that is, averaging the encoding results corresponding to each cell in at least one cell. Therefore, after processing k perturbation conditions, we can obtain... Z represents the representation space; of course, the second representation data can also be obtained through other means, which are not specifically limited here.

[0062] S250. Based on the second expression data corresponding to at least two perturbation conditions, perform loss calculation to adjust the object anomaly probability to obtain the target anomaly probability, and select the perturbation response object from at least one object based on the target anomaly probability.

[0063] The technical solution of this disclosure embodiment determines the normal probability of an object, and for each cell in at least one cell, obtains the probability data of the cell under perturbation conditions based on the first expression data of the cell under perturbation conditions and the normal probability of the object. Thus, based on the probability data of each cell under perturbation conditions, a second expression data of at least one cell under perturbation conditions is obtained. This second expression data can better express the object that is not affected by the perturbation.

[0064] An optional technical solution involves performing loss calculations based on second expression data corresponding to at least two perturbation conditions to adjust the object anomaly probability, thereby obtaining the target anomaly probability, including:

[0065] Based on the second expression data corresponding to at least two perturbation conditions, the similarity between the second expression data corresponding to at least two perturbation conditions is obtained;

[0066] Loss is calculated based on similarity, and the abnormal probability of the object is adjusted with the goal of minimizing the loss, thus obtaining the abnormal probability of the target.

[0067] The similarity between the second expression data corresponding to at least two perturbation conditions can be used to characterize the similarity between the second expression data corresponding to any two or more perturbation conditions, or it can be used to characterize the similarity between the second expression data corresponding to certain perturbation conditions among all perturbation conditions, etc., without specific limitations. As explained above, the first expression data of objects unaffected by perturbation under different perturbation conditions are usually highly similar, and the second expression data obtained based on the first expression data and the normal probability of the object can better represent objects unaffected by perturbation. This means that the second expression data corresponding to different perturbation conditions are usually highly similar. Therefore, loss can be calculated based on the similarity between the second expression data corresponding to different perturbation conditions, and the loss can be minimized, or the similarity maximized, as the adjustment objective to adjust the abnormal probability of the object, thereby obtaining a target abnormal probability that can be used to screen out perturbation response objects. In other words, the above technical solution, based on the objective function of independence loss, learns the abnormal expression probability of at least one object. It utilizes the stability of the expression of objects unaffected by perturbation under different perturbation conditions, and can accurately screen out perturbation response objects.

[0068] Based on this, optionally, the similarity between the second expression data corresponding to at least two perturbation conditions is obtained according to the second expression data corresponding to at least two perturbation conditions, including:

[0069] Based on the second expression data corresponding to at least two perturbation conditions, the distance between the second expression data corresponding to any two of the at least two perturbation conditions is obtained; based on the distance, the similarity between the second expression data corresponding to the at least two perturbation conditions is obtained.

[0070] In practical applications, the above-mentioned distance can optionally be represented by the maximum mean discrepancy (MMD) distance. Of course, it can also be represented by other distances, without any specific limitation here.

[0071] Based on this, for example, the objective function (i.e., independence loss) used for loss calculation in the above technical solution is l MMD (M) can be expressed by the following formula:

[0072]

[0073] There are a total of k perturbation conditions. This represents the second expression data corresponding to the i-th perturbation condition. This represents the second expression data corresponding to the j-th perturbation condition. This represents the distance between the two second-expressed data points. τ(·) is the mapping function corresponding to the kernel function. In practical applications, the Gaussian kernel function can be used, but other kernel functions can also be used. Based on this, it can be achieved through... Let l represent the sum of the distances corresponding to any two perturbation conditions, and use this sum as the calculated loss value. Furthermore, using the independence loss mentioned above to learn the anomaly probability of the object, we obtain the value that makes l... MMD (M) The smallest M, here we denote the smallest M as M * M * =argmin M l MMD (M), and thus can be based on M * Filter out the objects that respond to the disturbance.

[0074] Another optional technical solution is that the number of at least one cell includes at least two cells, and the second expression data of at least one cell under perturbation conditions is obtained based on the probability data of each cell under perturbation conditions, including: calculating the average value of the probability data of each cell under perturbation conditions, and using the average value as the second expression data of at least one cell under perturbation conditions, that is, the second expression data is represented by the average value, which can better represent the second expression data of at least one cell as a whole under perturbation conditions.

[0075] Another optional technical solution involves, after obtaining the probability data of cells under perturbation conditions, configuring the processor to: encode the probability data of cells under perturbation conditions to obtain first characterization data of cells under perturbation conditions. For example, the first characterization data can be represented as f θ (X⊙(1-M))|P=p i , where p i This represents the i-th perturbation condition;

[0076] Obtaining second expression data for at least one cell under perturbation conditions based on the probability data of each cell under perturbation conditions may include: obtaining second characterization data for at least one cell under perturbation conditions based on the first characterization data of each cell under perturbation conditions. For example, the second characterization data may be represented as...

[0077] The loss calculation is performed based on the second representation data corresponding to at least two perturbation conditions to adjust the object anomaly probability. This includes: calculating the loss based on the second characterization data corresponding to at least two perturbation conditions to adjust the object anomaly probability. That is, the second characterization data is directly used as the second representation data here.

[0078] The above technical solution, considering that the encoding result, i.e., the representation data, can better represent at least one object, adjusts the object's anomaly probability by performing loss calculations using the representation data, thereby improving the adjustment effect of the object's anomaly probability. Based on this, and considering the application scenarios that may be involved in the embodiments of this disclosure, for example, taking a gene as an object type, the above technical solution inputs all gene expression features when extracting cell features, thereby reducing feature loss caused by screening for highly variable genes.

[0079] Figure 3 This is a flowchart illustrating the steps configured in a processor within another perturbation response object screening system provided in this embodiment. This embodiment is an optimization based on the aforementioned technical solutions. Optionally, in this embodiment, the processor is further configured to: input the first expression data corresponding to at least two cells corresponding to at least two perturbation conditions into a trained autoencoder model for data reconstruction, to obtain reconstructed expression data corresponding to at least two cells; update the first expression data corresponding to at least two cells based on the reconstructed expression data; and determine the first expression data of each cell in at least one cell corresponding to a perturbation condition under the perturbation condition, including: determining the first expression data of each cell in at least one cell corresponding to a perturbation condition under the perturbation condition from the first expression data corresponding to at least two cells. Explanations of terms identical or corresponding to those in the aforementioned embodiments are not repeated here.

[0080] See Figure 3 The method in this embodiment may specifically include the following steps:

[0081] S310. For at least two cells corresponding to at least two preset perturbation conditions and for each of the at least two perturbation conditions, obtain the first expression data of each cell in at least one cell corresponding to the perturbation condition in the at least two cells under the perturbation condition, so as to obtain the first expression data corresponding to the at least two cells respectively, wherein the first expression data is used to express at least one object associated with each cell.

[0082] Here, "at least two cells" can be understood as cells corresponding to at least two perturbation conditions. For example, here we take at least two perturbation conditions including A, B and C as an example. Some cells in the at least two cells are located under A, some cells are located under B, and another part of the cells are located under C.

[0083] For each of the at least two perturbation conditions, the first expression data of each cell in at least one of the at least two cells corresponding to the perturbation condition (i.e. located under the perturbation condition) is obtained under the perturbation condition, thereby obtaining the first expression data corresponding to at least two cells respectively, that is, the first expression data corresponding to each cell in at least two cells respectively.

[0084] S320. Input the first expression data corresponding to at least two cells into the trained autoencoder model for data reconstruction to obtain the reconstructed expression data corresponding to at least two cells, and update the first expression data corresponding to at least two cells based on the reconstructed expression data corresponding to at least two cells.

[0085] The auto-encoder (AE) model can be understood as a pre-trained artificial neural network (ANN) used to reconstruct data. It should be noted that in practical applications, the detected first expression data may contain missing data. For example, due to the limitations of sequencing technology, the gene expression matrix obtained from single-cell RNA sequencing often has more than 90% missing data. Such severe data loss can easily affect the screening accuracy of perturbation response genes. Therefore, to solve the aforementioned data loss problem, this embodiment of the disclosure inputs the first expression data corresponding to at least two cells into the auto-encoder model, thereby reconstructing and filling in the missing data in these first expression data to obtain their respective reconstructed expression data. Subsequent steps are then implemented based on the reconstructed expression data.

[0086] For example, an autoencoder model can consist of two parts: an encoder and a decoder. The encoder can be represented as f. θ ∶χ→z, where This represents the input sample space, where d is the dimension of the input sample, such as the total number of genes. Let represent the representation space, h be the dimension of the representation, and θ be the model parameters of the encoder; the decoder can be represented as g. φ :Z→Z, φ are the model parameters of the decoder. In practical applications, f is optional. θ With g φ Both can use a two-layer multilayer perceptron (MLP). For the input data sample (i.e., the first representation data) x∈Z, the output of the autoencoder model can be expressed as:

[0087] S330. For each of the at least two perturbation conditions, determine the first expression data of each cell in at least one cell corresponding to the perturbation condition from the first expression data corresponding to the at least two cells respectively.

[0088] S340. Obtain the object anomaly probability, wherein the object anomaly probability can be used to characterize the probability that each of at least one object is affected by at least two perturbation conditions.

[0089] S350. Based on the abnormality probability of the object and the first expression data of each cell under the perturbation condition, obtain the second expression data of at least one cell under the perturbation condition.

[0090] S360. Based on the second expression data corresponding to at least two perturbation conditions, perform loss calculation to adjust the object anomaly probability to obtain the target anomaly probability, and select the perturbation response object from at least one object based on the target anomaly probability.

[0091] The technical solution of this disclosure embodiment reconstructs the missing data in the first expression data through an autoencoder model, thereby filtering perturbation response objects based on the obtained reconstructed expression data, thus solving the data missing problem and further improving the filtering accuracy of perturbation response objects.

[0092] In an alternative technical solution, the aforementioned processor is further configured to:

[0093] The first expression data corresponding to at least two cells is input into the original model to be trained to obtain the third expression data corresponding to at least two cells. For example, in conjunction with the above example, the third expression data can be obtained through... To express;

[0094] For each cell in at least two cells, determine the first measured data detected in the first expression data corresponding to the cell and the third measured data corresponding to the first measured data in the third expression data corresponding to the cell, and calculate the loss value of the cell based on the first measured data and the third measured data;

[0095] Based on the loss values ​​corresponding to at least two cells, the model parameters of the original model are adjusted to train an autoencoder model.

[0096] Specifically, for each cell in at least two cell groups, the first measured data for that cell can be understood as the data in the first expression data for that cell that has been detected and is not missing, and the third measured data can be understood as the data in the third expression data for that cell that corresponds to the first measured data. Loss is calculated based on the first and third measured data to obtain the loss value for that cell.

[0097] For example, the above loss calculation process can be implemented using the missing value-sensitive loss function described below, thereby obtaining the loss value of the cell.

[0098]

[0099] Where x represents the first expression data corresponding to that cell. This represents the third expression data corresponding to this cell, x. i This represents the data used to represent the i-th object in the first set of data. This represents the data used to represent the i-th object in the third set of data, where d represents the number of at least one object. Based on this, It can represent an indicator, when x i When the value is greater than 0, it indicates that the data of the i-th object within the cell has been detected. The value is 1 for all cases and 0 for the rest. In other words, the above loss function implements loss constraints for a single cell, and only applies loss constraints to the data detected within the single cell (i.e., the first and third measured data), without applying loss constraints to the undetected data.

[0100] Furthermore, based on the loss values ​​corresponding to at least two cells, the model parameters of the original model are adjusted to train the autoencoder model. For example, loss is calculated based on the loss values ​​corresponding to at least two cells (e.g., their sum or average is calculated), and the model parameters are adjusted with loss minimization as the adjustment objective, thereby training the autoencoder model.

[0101] For example, continuing with the above example, the training process of an autoencoder can be represented as follows:

[0102]

[0103] Where N represents the number of at least two cells, x c This indicates the first expression data of the c-th cell in at least two cells. This represents the third expression data of the c-th cell. This represents the loss value of the c-th cell. This represents the sum of the loss values ​​of N cells. Based on this, with... Minimize the target value, adjust θ and φ, and thus train the autoencoder model.

[0104] It should be noted that, typically, some objects may be detected in one cell and others in another. For example, assuming d=3 and N=2, the first and third objects are detected in the first cell, and the first and second objects are detected in the second cell. This means that although the second object is not constrained in the loss function of the first cell, it is constrained in the loss function of the second cell. Therefore, combined with the above formula for loss constraint on all cells, it can constrain the vast majority of objects overall, meaning that the vast majority of objects can be detected in a cell. This allows the trained autoencoder model to effectively reconstruct the missing data in the first expression data, ensuring the data reconstruction effect of the autoencoder model.

[0105] Based on this, optionally, determining the first measured data detected in the first expression data corresponding to the cell may include: taking data with values ​​greater than 0 in the first expression data corresponding to the cell as the first measured data detected in the first expression data corresponding to the cell, thereby ensuring the rapid and accurate determination of the first measured data.

[0106] Another optional technical solution involves obtaining second expression data for at least one cell under perturbation conditions based on the object anomaly probability and the first expression data of each cell under perturbation conditions. This can include: encoding the object anomaly probability and the first expression data of each cell under perturbation conditions using an encoder in an autoencoder model, and using the resulting third characterization data as the second expression data of at least one cell under perturbation conditions. In other words, the encoder in the autoencoder model trained to achieve the data reconstruction function has a good match with the first expression data, and therefore can also be used to solve counterfactual problems, thereby ensuring the screening accuracy of the perturbation response objects selected in this way.

[0107] To demonstrate the effectiveness of the above technical solutions, specific test results are provided below for illustrative purposes. For example, this study uses genes as the object type and tests were conducted on a Jurkat T-cell dataset. The perturbation condition for the cells in this dataset is whether or not they have been treated with anti-CD3 / anti-CD28 antibodies. The experimental goal is to screen for T-cell receptor activation-related pathways and genes for genomic enrichment analysis. Specifically,

[0108] Experimental data include gene expression matrices Where 12000 represents the number of cells and 10368 represents the number of genes, and the perturbation conditions (0: no perturbation; 1: perturbation) for cells are known. The structure of the autoencoder model used in the experiment is as follows: both the encoder and decoder are two-layer MLPs, with 1024 hidden layers and a representation dimension of 256. For all data X, the model parameters θ and φ of the autoencoder model are first trained using the following formula:

[0109]

[0110] Among them, X c Let X represent the gene expression matrix corresponding to the c-th cell. Then, introduce the gene aberration probability M∈[0,1]. d The final target anomaly probability M is obtained by optimization using the following formula. * :

[0111] M * =argmin M l MMD (M)

[0112] Then, for M * The abnormal expression probabilities of different genes were ranked to obtain the 2000 genes with the highest abnormal expression probabilities. Then, genome enrichment analysis was used to calculate the overlap with genes in known pathways and the combined score. The calculation process for the combined score is as follows:

[0113] First, Fisher's exact test is used to calculate the enrichment of the input genes, and the mean rank and standard deviation from the expected rank are calculated for each metabolic or signaling pathway in each gene set to obtain the p-value. Then, using the expected rank and its variance lookup table, the z-score is calculated for that rank. The composite score is obtained by multiplying the p-value and the z-score.

[0114] c = log(p) * Z

[0115] Where c is the composite score, p is the p-value obtained from Fisher's exact test, and Z is the z-score obtained from the table lookup.

[0116] The results of the MELD algorithm are compared with the test results, as shown in the table below:

[0117] Table 1. Combined Score Results

[0118] MELD 4.5 12.8 12.0 9.7 2.1 0.59 Ours 16.7 74.8 70.6 53.4 58.6 15.1

[0119] Table 2. Pathway enrichment overlap results

[0120] MELD 24 / 89 19 / 53 27 / 82 21 / 64 9 / 35 9 / 46 Ours 19 / 89 19 / 53 26 / 82 20 / 64 13 / 35 11 / 46

[0121] It should be noted that the MELD method screened 3,900 genes, many of which were false positives. In contrast, the test results in this study, with about half the number of genes screened by MELD, achieved a very close overlap and a significantly higher overall score, demonstrating a significant advantage. This also illustrates the effectiveness and efficiency of the aforementioned technical solutions.

[0122] This disclosure provides a method for screening perturbation response objects. This method is applicable to investigating the effects of perturbations on cells, particularly by screening perturbation response objects from at least one object associated with the cell. The object type of the perturbation response object can be a gene, protein, or chromatin accessibility, etc. This method can be executed by the perturbation response object screening device provided in this disclosure, which can be implemented in software and / or hardware and can be integrated into the processor described in any of the above embodiments.

[0123] The method of this disclosure specifically includes the following steps:

[0124] S410. For each of the preset at least two perturbation conditions, determine the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition, wherein the first expression data is used to express at least one object associated with each cell.

[0125] S420. Obtain the object anomaly probability, wherein the object anomaly probability can be used to characterize the probability that each of at least one object is affected by at least two perturbation conditions.

[0126] S430. Based on the abnormality probability of the object and the first expression data of each cell under the perturbation condition, obtain the second expression data of at least one cell under the perturbation condition.

[0127] S440. Based on the second expression data corresponding to at least two perturbation conditions, perform loss calculation to adjust the object anomaly probability to obtain the target anomaly probability, and select the perturbation response object from at least one object based on the target anomaly probability.

[0128] The technical solution in this embodiment addresses at least two preset perturbation conditions by determining the first expression data of each cell in at least one cell corresponding to the perturbation condition. This first expression data expresses at least one object associated with each cell. An object anomaly probability is obtained, characterizing the probability that each object in the at least one object is affected by at least two perturbation conditions. Based on the object anomaly probability and the first expression data of each cell under the perturbation condition, second expression data of the at least one cell under the perturbation condition is obtained. Then, based on the second expression data corresponding to the at least two perturbation conditions, a loss calculation is performed to adjust the object anomaly probability, resulting in a target anomaly probability. Finally, perturbation-responsive objects are selected from the at least one object based on the target anomaly probability. This technical solution leverages the stability of the first expression data of objects unaffected by perturbation under different perturbation conditions in biology, i.e., the independence between the first expression data and the perturbation conditions. By obtaining the second expression data corresponding to different perturbation conditions through the object anomaly probability and the first expression data, and then learning the object anomaly probability based on this second expression data, the perturbation-responsive object selection process is achieved based on the learned target anomaly probability, thereby improving the selection accuracy.

[0129] Figure 4 This is a structural block diagram of a disturbance response object screening device provided in an embodiment of this disclosure. This device can be used to execute the disturbance response object screening method provided in any of the above embodiments. This device and the disturbance response object screening methods of the above embodiments belong to the same inventive concept. Details not described in detail in the embodiments of the disturbance response object screening device can be found in the embodiments of the disturbance response object screening methods described above. See also... Figure 4 The device, configured within a processor, may specifically include: a first expression data determination module 510, an object anomaly probability acquisition module 520, a second expression data acquisition module 530, and a disturbance response object screening module 540.

[0130] The first expression data determination module 510 is used to determine, for each of the preset at least two perturbation conditions, the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition, wherein the first expression data is used to express at least one object associated with each cell;

[0131] The object anomaly probability acquisition module 520 is used to acquire the object anomaly probability, wherein the object anomaly probability is used to characterize the probability that each of at least one object is affected by at least two perturbation conditions.

[0132] The second expression data acquisition module 530 is used to obtain the second expression data of at least one cell under perturbation conditions based on the object abnormality probability and the first expression data of each cell under perturbation conditions.

[0133] The disturbance response object screening module 540 is used to perform loss calculation based on the second expression data corresponding to at least two disturbance conditions to adjust the object anomaly probability, obtain the target anomaly probability, and screen out the disturbance response object from at least one object based on the target anomaly probability.

[0134] Optionally, the above-mentioned screening device further includes:

[0135] The normal probability determination module is used to determine the normal probability of an object corresponding to the abnormal probability of an object. The normal probability of an object is used to characterize the probability that each object is not affected by at least two perturbation conditions.

[0136] The second expression data acquisition module 530 includes:

[0137] The probability data acquisition submodule is used to obtain the probability data of a cell under perturbation conditions for each cell in at least one cell, based on the cell’s first expression data under perturbation conditions and the normal probability of the object.

[0138] The second expression data acquisition submodule is used to obtain the second expression data of at least one cell under perturbation conditions based on the probability data of each cell under perturbation conditions.

[0139] Based on this, optionally, the disturbance response object filtering module 540 includes:

[0140] The similarity acquisition submodule is used to obtain the similarity between the second expression data corresponding to at least two perturbation conditions based on the second expression data corresponding to at least two perturbation conditions respectively;

[0141] The target anomaly probability acquisition submodule is used to calculate the loss based on similarity and adjust the object anomaly probability with the goal of minimizing the loss, thus obtaining the target anomaly probability.

[0142] Based on this, optional sub-modules are derived from similarity, specifically used for:

[0143] Based on the second expression data corresponding to at least two perturbation conditions, the distance between the second expression data corresponding to any two of the at least two perturbation conditions is obtained.

[0144] Based on the distance between them, the similarity between the second expression data corresponding to at least two perturbation conditions is obtained.

[0145] Alternatively, at least two cells are present, and the second expression data is used to obtain a submodule specifically for:

[0146] Calculate the average of the probability data for each cell under perturbation conditions, and use the average as the second expression data for at least one cell under perturbation conditions.

[0147] Alternatively, the above-mentioned screening device may further include:

[0148] The first characterization data acquisition module is used to encode the probability data of cells under perturbation conditions after obtaining the probability data of cells under perturbation conditions, so as to obtain the first characterization data of cells under perturbation conditions.

[0149] The second data expression submodule is specifically used for:

[0150] Based on the first characterization data of each cell under perturbation conditions, obtain the second characterization data of at least one cell under perturbation conditions;

[0151] Disturbance response object filtering module 540 includes:

[0152] The object anomaly probability adjustment submodule is used to perform loss calculations based on the second characterization data corresponding to at least two perturbation conditions to adjust the object anomaly probability.

[0153] Optionally, the above-mentioned screening device further includes:

[0154] The module for reconstructing expression data is used to input the first expression data corresponding to at least two cells corresponding to at least two perturbation conditions into a trained autoencoder model for data reconstruction, thereby obtaining reconstructed expression data corresponding to at least two cells.

[0155] The first expression data update module is used to update the first expression data corresponding to at least two cells based on the reconstructed expression data corresponding to at least two cells respectively;

[0156] The first expression data determination module 510 includes:

[0157] The first expression data determination submodule is used to determine, from the first expression data corresponding to at least two cells respectively, the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition.

[0158] In addition, optionally, the above-mentioned screening device further includes:

[0159] The third expression data acquisition module is used to input the first expression data corresponding to at least two cells into the original model to be trained, and obtain the third expression data corresponding to at least two cells.

[0160] The loss value acquisition module is used to determine, for each cell in at least two cells, the first measured data detected in the first expression data corresponding to the cell and the third measured data corresponding to the first measured data in the third expression data corresponding to the cell, and to perform loss calculation based on the first measured data and the third measured data to obtain the loss value of the cell.

[0161] The autoencoder model acquisition module is used to adjust the model parameters of the original model based on the loss values ​​corresponding to at least two cells, so as to train the autoencoder model.

[0162] Based on this, optional modules for obtaining the loss value include:

[0163] The first measured data acquisition submodule is used to take data with values ​​greater than 0 in the first expression data corresponding to the cell as the first measured data that has been detected in the first expression data corresponding to the cell.

[0164] Alternatively, the second expression data obtaining module 530 is specifically used for:

[0165] Based on the encoder in the autoencoder model, the abnormal probability of the object and the first expression data of each cell under perturbation conditions are encoded, and the resulting third characterization data is used as the second expression data of at least one cell under perturbation conditions.

[0166] Optionally, the target anomaly probability may include anomaly expression probabilities corresponding to at least one object, and the perturbation response object filtering module 540 includes:

[0167] The perturbation response object filtering submodule is used to sort the probabilities of at least one anomalous expression and filter out perturbation response objects from at least one object based on the sorting results.

[0168] Based on this, optionally, the number of disturbance response objects is at least two, and the above-mentioned screening device further includes:

[0169] The priority determination module is used to determine the priority of each of the at least two disturbance response objects for the at least two disturbance condition responses based on the abnormal expression probabilities corresponding to the at least two disturbance response objects respectively.

[0170] Optionally, at least one object may be of the type including genes, chromatin accessibility, or proteins.

[0171] The disturbance response object screening device provided in this embodiment of the present disclosure, for each of at least two preset disturbance conditions, determines the first expression data of each cell in at least one cell corresponding to the disturbance condition under the disturbance condition through a first expression data determination module. The first expression data is used to express at least one object associated with each cell. The object anomaly probability acquisition module acquires the object anomaly probability, which is used to characterize the probability that each object in at least one object is affected by at least two disturbance conditions. The second expression data acquisition module obtains the second expression data of at least one cell under the disturbance condition based on the object anomaly probability and the first expression data of each cell under the disturbance condition. Then, the disturbance response object screening module performs loss calculation to adjust the object anomaly probability based on the second expression data corresponding to at least two disturbance conditions to obtain a target anomaly probability, and screens out disturbance response objects from at least one object based on the target anomaly probability. The aforementioned device utilizes the stability of the first expression data of an object unaffected by perturbation under different perturbation conditions in biology, i.e., the first expression data is independent of the perturbation conditions. By obtaining the object's anomaly probability and the first expression data, the device obtains the second expression data corresponding to different perturbation conditions. Then, based on these second expression data, the device learns the object's anomaly probability. Finally, based on the learned target anomaly probability, the device performs the screening process for perturbation response objects, thereby improving the screening accuracy.

[0172] The disturbance response object screening device provided in this disclosure can execute the disturbance response object screening method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects of the method execution.

[0173] It is worth noting that in the embodiments of the above-mentioned disturbance response object screening device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the protection scope of this disclosure.

[0174] The aforementioned system for screening disturbance response objects can be implemented in an electronic device designed to represent various forms of digital computers, such as laptops, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframes, sequencers, gene sequencing systems, large-scale population genomics platforms, PCR instruments, spatiotemporal omics systems, laboratory automation systems, sample preparation equipment, dispensing equipment, library production equipment, pipetting equipment, magnetic bead detection equipment, nucleic acid purification equipment, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0175] As used herein, the term "sequencing instrument" generally refers to the instrument used to determine the sequence of the genetic material of a sample. Sequencing instruments can function in a variety of ways and based on a variety of technologies, including sequencing using labeled or unlabeled nucleotides via primer extension, such as sequencing-while-ligating or pyrosequencing, for example using Sanger dideoxy sequencing, nanopore sequencing, or any of the "NexGen" sequencing methods in the art (e.g., using MGI sequencing platforms, ROCHE 454 sequencing platforms, ILLUMINA™ SOLEXA™ sequencing platforms, LIFE TECHNOLOGIES / APPLIED BIOSYSTEMS' SOLID™ sequencing platforms, PACIFIC BIOSCIENCES' SMRT™ sequencing platforms, POLLONATOR Polony sequencing platforms, COMPLETE GENOMICS sequencing platforms, INTELLIGENT BIOSYSTEMS' sequencing platforms, HELICOS sequencing platforms, or any other sequencing instruments and systems in the art).

[0176] Quantitative real-time PCR (QPCR) is a method that uses fluorescent chemicals to determine the total amount of product after each polymerase chain reaction (PCR) cycle in a DNA amplification reaction. It is a method for quantitative analysis of specific DNA sequences in a sample using internal or external controls.

[0177] Quantitative PCR (qPCR) is a technique built upon immunology, biochemistry, and microscopy. Based on the principle of antigen-antibody reactions, it involves labeling known antigens or antibodies with fluorescent groups, then using this fluorescent antibody (or antigen) as a probe to detect the corresponding antigen (or antibody) within cells or tissues. Fluorescence microscopy allows visualization of the cells or tissues where fluorescence is present, thus determining the nature and location of the antigen or antibody, and enabling quantitative techniques (such as flow cytometry) to measure its concentration. However, it employs fiber optic transmission technology for detection, which suffers from light attenuation, has a long transmission path, and a complex structure.

[0178] For example, such as Figure 5 The illustrated electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer programs stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0179] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0180] Processor 11 is a computing unit and can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), graphics processing units (GPUs), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 can perform the various methods and processes described above, such as methods for screening disturbance response objects.

[0181] In some embodiments, the method for screening disturbance response objects may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the method for screening disturbance response objects described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the method for screening disturbance response objects by any other suitable means (e.g., by means of firmware).

[0182] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0183] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0184] In the context of this disclosure, a computer-readable storage medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium can be a machine-readable signal medium or a machine-readable storage medium. A computer-readable storage medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more lines, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0185] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0186] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0187] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0188] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this disclosure can be achieved, and this is not limited herein.

[0189] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A system for screening disturbance response objects, characterized in that, include: The processor is configured to: For each of at least two preset perturbation conditions, determine the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition, wherein the first expression data is used to express at least one object associated with each cell; Obtain the object anomaly probability, wherein the object anomaly probability is used to characterize the probability that each of the at least one object is affected by the at least two perturbation conditions; Based on the abnormality probability of the object and the first expression data of each cell under the perturbation condition, the second expression data of the at least one cell under the perturbation condition is obtained; Based on the second expression data corresponding to the at least two perturbation conditions, loss calculation is performed to adjust the object anomaly probability to obtain the target anomaly probability, and perturbation response objects are selected from the at least one object based on the target anomaly probability.

2. The system according to claim 1, characterized in that, The processor is also configured to: Determine the normal probability of an object corresponding to the abnormal probability of the object, wherein the normal probability of the object is used to characterize the probability that each object is not affected by the at least two perturbation conditions; The step of obtaining second expression data of at least one cell under the perturbation condition based on the abnormality probability of the object and the first expression data of each cell under the perturbation condition includes: For each cell in the at least one cell, probability data of the cell under the perturbation condition is obtained based on the first expression data of the cell under the perturbation condition and the normal probability of the object; Based on the probability data of each cell under the perturbation condition, second expression data of the at least one cell under the perturbation condition are obtained.

3. The system according to claim 2, characterized in that, The step of performing loss calculation based on the second expression data corresponding to the at least two perturbation conditions to adjust the object anomaly probability and obtain the target anomaly probability includes: Based on the second expression data corresponding to the at least two perturbation conditions, the similarity between the second expression data corresponding to the at least two perturbation conditions is obtained; Loss is calculated based on the similarity, and the abnormal probability of the object is adjusted with the goal of minimizing the loss, to obtain the abnormal probability of the target.

4. The system according to claim 3, characterized in that, The step of obtaining the similarity between the second expression data corresponding to the at least two perturbation conditions based on the second expression data corresponding to the at least two perturbation conditions includes: Based on the second expression data corresponding to the at least two perturbation conditions, the distance between the second expression data corresponding to any two of the at least two perturbation conditions is obtained; Based on the distance between them, the similarity between the second expression data corresponding to the at least two perturbation conditions is obtained.

5. The system according to claim 2, characterized in that, The number of the at least one cell includes at least two, and the second expression data of the at least one cell under the perturbation condition is obtained based on the probability data of each cell under the perturbation condition, including: Calculate the average value of the probability data of each cell under the perturbation condition, and use the average value as the second expression data of the at least one cell under the perturbation condition.

6. The system according to claim 2, characterized in that, After obtaining the probability data of the cell under the perturbation condition, the processor is further configured to: The probability data of the cell under the perturbation condition are encoded to obtain the first characterization data of the cell under the perturbation condition; The step of obtaining second expression data of the at least one cell under the perturbation condition based on the probability data of each cell under the perturbation condition includes: Based on the first characterization data of each cell under the perturbation condition, the second characterization data of the at least one cell under the perturbation condition is obtained; The step of performing loss calculation to adjust the anomaly probability of the object based on the second expression data corresponding to the at least two perturbation conditions includes: Based on the second characterization data corresponding to the at least two perturbation conditions, loss calculation is performed to adjust the abnormal probability of the object.

7. The system according to claim 1, characterized in that, The processor is also configured to: For at least two cells corresponding to the at least two perturbation conditions, the first expression data corresponding to the at least two cells are input into the trained autoencoder model for data reconstruction to obtain the reconstructed expression data corresponding to the at least two cells. Based on the reconstructed expression data corresponding to the at least two cells respectively, the first expression data corresponding to the at least two cells respectively are updated; The determination of the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition includes: From the first expression data corresponding to the at least two cells respectively, determine the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition.

8. The system according to claim 7, characterized in that, The processor is also configured to: The first expression data corresponding to each of the at least two cells is input into the original model to be trained to obtain the third expression data corresponding to each of the at least two cells. For each of the at least two cells, a first measured data detected in the first expression data corresponding to the cell and a third measured data corresponding to the first measured data in the third expression data corresponding to the cell are determined, and a loss calculation is performed based on the first measured data and the third measured data to obtain the loss value of the cell; Based on the loss values ​​corresponding to the at least two cells, the model parameters of the original model are adjusted to train the autoencoder model.

9. The system according to claim 8, characterized in that, The determination of the first measured data detected in the first expression data corresponding to the cell includes: Data with values ​​greater than 0 in the first expression data corresponding to the cell are taken as the first detected data in the first expression data corresponding to the cell.

10. The system according to claim 7, characterized in that, The step of obtaining second expression data of at least one cell under the perturbation condition based on the abnormality probability of the object and the first expression data of each cell under the perturbation condition includes: Based on the encoder in the autoencoder model, the abnormal probability of the object and the first expression data of each cell under the perturbation condition are encoded, and the obtained third characterization data is used as the second expression data of the at least one cell under the perturbation condition.

11. The system according to claim 1, characterized in that, The target anomaly probability includes the anomaly expression probability corresponding to each of the at least one object; The step of filtering perturbation response objects from the at least one object based on the target anomaly probability includes: The probabilities of the at least one abnormal expression are sorted, and perturbation response objects are selected from the at least one object based on the sorting results.

12. The system according to claim 11, characterized in that, The number of disturbance response objects is at least two, and the processor is further configured to: Based on the abnormal expression probabilities corresponding to the at least two disturbance response objects, determine the priority of each of the at least two disturbance response objects for the at least two disturbance conditional responses.

13. The system according to any one of claims 1-12, characterized in that, The type of at least one object includes genes, chromatin accessibility, or proteins.

14. A device for screening disturbance response objects, characterized in that, The device, configured in a processor, includes: The first expression data determination module is used to determine, for each of at least two preset perturbation conditions, the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition, wherein the first expression data is used to express at least one object associated with each cell; An object anomaly probability acquisition module is used to acquire object anomaly probabilities, wherein the object anomaly probability is used to characterize the probability that each of the at least one object is affected by the at least two perturbation conditions. The second expression data obtaining module is used to obtain the second expression data of the at least one cell under the perturbation condition based on the abnormality probability of the object and the first expression data of each cell under the perturbation condition; The disturbance response object filtering module is used to perform loss calculation based on the second expression data corresponding to the at least two disturbance conditions to adjust the object anomaly probability, obtain the target anomaly probability, and filter out disturbance response objects from the at least one object based on the target anomaly probability.

15. A method for screening disturbance response objects, characterized in that, When applied to a processor, the method includes: For each of at least two preset perturbation conditions, determine the first expression data of each cell in at least one cell corresponding to the perturbation condition under the perturbation condition, wherein the first expression data is used to express at least one object associated with each cell; Obtain the object anomaly probability, wherein the object anomaly probability is used to characterize the probability that each of the at least one object is affected by the at least two perturbation conditions; Based on the abnormality probability of the object and the first expression data of each cell under the perturbation condition, the second expression data of the at least one cell under the perturbation condition is obtained; Based on the second expression data corresponding to the at least two perturbation conditions, loss calculation is performed to adjust the object anomaly probability to obtain the target anomaly probability, and perturbation response objects are selected from the at least one object based on the target anomaly probability.

16. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the method for filtering disturbance response objects as described in claim 15.

Images