Biomarker screening methods based on health background constraints and perturbation causality
By constructing a health covariance matrix and a disease evolution direction vector, and combining single-gene virtual perturbation and high-dimensional embedding representation, the system displacement vector and gene driver score are calculated. This solves the problem of accurately identifying key driver genes in the process of inflammation transforming into cancer in existing technologies, and achieves accurate gene screening.
Patent Information
- Application Number
- CN202611301735.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-26
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies struggle to accurately identify key driver genes in the process of inflammation transforming into cancer, and existing methods are inadequate in terms of data dependence and causal characterization, leading to unstable screening results.
By constructing a health covariance matrix and a disease evolution direction vector, and combining single-gene virtual perturbation and high-dimensional embedding representation, the system displacement vector and gene drive score are calculated to screen out key driver genes.
This enabled accurate screening of key driver genes in the transformation of inflammation into cancer, improving the biological interpretability and stability of the screening results.
Smart Images

Figure CN123120165A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of bioinformatics and artificial intelligence, and in particular to a method for screening biomarkers based on health background constraints and perturbation causality. Background Technology
[0002] The transformation of inflammation into cancer is a complex biological process involving multiple stages and multi-gene synergistic regulation. Essentially, it represents a systematic remodeling of gene expression regulatory networks under long-term pathological stimulation. Therefore, accurately identifying key driver genes in this process is crucial for elucidating disease mechanisms, early diagnosis, and targeted intervention.
[0003] Existing biomarker screening methods mainly include differential expression analysis, correlation analysis, and feature selection based on machine learning. However, these methods generally suffer from the following limitations: differential expression analysis typically screens based solely on statistical differences between disease and healthy groups, making it difficult to characterize the intrinsic fluctuation structure of genes under normal physiological conditions and effectively distinguish between pathological driving signals and physiological background noise; correlation analysis can only characterize the statistical association between gene expression and disease phenotypes, failing to further reveal the causal driving effect of gene changes on the overall gene regulatory network, thus making it difficult to effectively identify truly key regulatory genes; machine learning-based screening methods typically rely on the model's fitting ability on specific datasets, and their results are greatly affected by data distribution and sample bias, lacking explicit modeling of disease evolution direction, resulting in poor stability of screening results across different datasets.
[0004] Therefore, existing technologies generally lack an accurate and reliable screening mechanism for key driver genes in the process of inflammation transforming into cancer. Summary of the Invention
[0005] This invention provides a biomarker screening method based on health background constraints and perturbation causality to achieve accurate screening of key driver genes in the process of inflammation transforming into cancer.
[0006] In a first aspect, embodiments of the present invention provide a biomarker screening method based on health background constraints and perturbation causality, the method comprising: Identify healthy samples, inflammatory samples, and cancer samples; determine the health covariance matrix based on the healthy samples; and determine the disease evolution direction vector based on the inflammatory and cancer samples. In the inflammatory samples, the base sample is identified, and the perturbation sample is identified after virtual perturbation of the base sample with the corresponding gene of the biomarker. Based on the high-dimensional embedding representation of the perturbation sample and the high-dimensional embedding representation of the base sample, the system shift vector of the corresponding gene of the biomarker is determined. Based on the system displacement vector and the disease evolution direction vector, the effective perturbation component along the disease evolution direction is determined, and the effective perturbation component is weighted and scored according to the health covariance matrix to obtain the gene drive score of the corresponding gene of the biomarker. Based on the gene drive scores corresponding to each biomarker, the set of target biomarker genes driving inflammation-cancer transformation is determined among the corresponding genes of each biomarker.
[0007] Secondly, embodiments of the present invention also provide a biomarker screening device based on health background constraints and perturbation causality, the device comprising: The module for determining the health covariance matrix and the disease evolution direction vector is used to identify healthy samples, inflammatory samples, and cancer samples, and to determine the health covariance matrix based on the healthy samples and the disease evolution direction vector based on the inflammatory and cancer samples. The system displacement vector determination module is used to determine the base sample in the inflammatory sample and the perturbation sample after virtual perturbation of the base sample with the corresponding gene of the biomarker. Based on the high-dimensional embedding representation of the perturbation sample and the high-dimensional embedding representation of the base sample, the system displacement vector of the corresponding gene of the biomarker is determined. The gene drive score determination module is used to determine the effective perturbation component along the disease evolution direction based on the system displacement vector and the disease evolution direction vector, and to weight the effective perturbation component according to the health covariance matrix to obtain the gene drive score of the gene corresponding to the biomarker. The target biomarker corresponding gene set determination module is used to determine the target biomarker corresponding gene set driving inflammation-cancer transformation from each biomarker corresponding gene based on the gene drive score of each biomarker corresponding gene.
[0008] Thirdly, embodiments of the present invention also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the biomarker screening method based on health background constraints and perturbation causality as described in any of the embodiments of the present invention.
[0009] Fourthly, embodiments of the present invention also provide a storage medium for storing computer-executable instructions, which, when executed by a computer processor, are used to perform the biomarker screening method based on health background constraints and perturbation causality as described in any of the embodiments of the present invention.
[0010] The technical solution of this invention establishes a health covariance matrix based on healthy samples to characterize the statistical distribution structure of gene expression under normal physiological conditions; constructs disease evolution direction vectors based on samples from different disease stages to represent the overall change direction in the process of inflammation transforming into cancer; performs virtual perturbation on the genes corresponding to single biomarkers in the basic samples determined in the inflammation samples to obtain the changes in high-dimensional embedding representations before and after perturbation, calculates the system displacement vector to simulate the impact of gene expression changes on the global regulatory system, and achieves an approximate characterization of potential causal driving relationships; projects the system displacement vector onto the disease evolution direction and performs weighted processing in conjunction with the health covariance matrix to uniformly quantify and score the perturbation response of each gene; finally, it screens the set of genes corresponding to biomarkers based on the gene driving score. This invention, by integrating health background constraints, disease evolution direction constraints, single-gene causal perturbation analysis, and unified quantitative scoring of gene perturbation responses, performs multi-level decoupling and quantitative characterization of pathological driving signals in the gene regulation system, and achieves accurate screening of key driving genes in the process of inflammation transforming into cancer.
[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description
[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0013] Figure 1 This is a flowchart of a biomarker screening method based on health background constraints and perturbation causality provided in Embodiment 1 of the present invention; Figure 2 This is a flowchart of another biomarker screening method based on health background constraints and perturbation causality provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the structure of a biomarker screening device based on health background constraints and perturbation causality provided in Embodiment 3 of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device provided in Embodiment 4 of the present invention. Detailed Implementation
[0014] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0015] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention 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 the invention described herein can be implemented in orders other than those illustrated or described herein. 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 device 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 devices. In the embodiments of this application, certain software, components, models, and other existing industry solutions may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solutions of this application, and do not imply that the applicant has already used or necessarily used such solutions.
[0016] The acquisition, transmission, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.
[0017] Example 1 Figure 1 This document provides a flowchart of a biomarker screening method based on health background constraints and perturbation causality, as described in Embodiment 1 of the present invention. This embodiment is applicable to screening key driver genes in the process of inflammation transforming into cancer. The method can be executed by a biomarker screening device based on health background constraints and perturbation causality. This device can be implemented in hardware and / or software and can be configured in a server or electronic device.
[0018] like Figure 1 As shown, the method includes: S110. Identify healthy samples, inflammatory samples, and cancer samples, and determine the health covariance matrix based on the healthy samples, and determine the disease evolution direction vector based on the inflammatory and cancer samples.
[0019] A healthy sample refers to a transcriptome sample in a healthy state. A transcriptome sample is a biological sample from which total RNA is extracted from biological tissues or cells, reflecting the transcriptional level or expression level of all genes under specific spatiotemporal and treatment conditions. Correspondingly, an inflammation sample refers to a transcriptome sample from the inflammatory stage, and a cancer sample refers to a transcriptome sample from the cancer stage.
[0020] The health covariance matrix is used to characterize the natural fluctuations in gene expression systems, the stability of gene associations, and the distribution structure of the healthy background under healthy physiological conditions. The disease evolution direction vector is used to represent the directional characteristics of disease evolution from an inflammatory state to a cancerous state.
[0021] Specifically, S110 can include: S11. Using a pre-trained transcriptome model, healthy samples, inflammatory samples, and cancer samples are processed respectively to obtain high-dimensional embedding representations corresponding to healthy samples, inflammatory samples, and cancer samples. S12. Based on the high-dimensional embedding representation of each healthy sample, calculate the mean vector of the healthy sample; based on the high-dimensional embedding representation of each inflammatory sample, calculate the mean vector of the inflammatory sample; and based on the high-dimensional embedding representation of each cancer sample, calculate the mean vector of the cancer sample. S13. Calculate the health covariance matrix based on the mean vector of the health samples and the high-dimensional embedding representation of each health sample; S14. Calculate the difference between the mean vector of the inflammation sample and the mean vector of the cancer sample, and normalize the difference by the magnitude of the difference to obtain the disease evolution direction vector.
[0022] The transcriptome model is a model obtained through self-supervised training of a deep neural network model using a large amount of training data from healthy transcriptomes as well as transcriptomes from inflammatory and cancer stages. For example, the deep neural network model can be built based on a Transformer architecture, including encoder structures, decoder structures, encoder-decoder structures, or combinations thereof, to learn the global regulatory structure of the gene expression system and the evolutionary patterns of diseases in healthy and disease states. The self-supervised training process of the deep neural network model can include: inputting the gene expression vectors or transcriptome expression matrices corresponding to the training data into the deep neural network model, performing feature reconstruction, expression prediction, mask recovery, or equivalent self-supervised task training, and the model output being the corresponding high-dimensional embedding representation, expression prediction result, or latent state representation. The trained transcriptome model can establish a high-dimensional semantic representation space of the gene expression system and can learn the statistical associations, regulatory dependencies, and global expression context structure between gene expressions.
[0023] In this embodiment, transcriptome samples in a healthy state are obtained. , take health samples Input to transcriptome model In China, transcriptome models Output the high-dimensional embedding representation of the healthy sample. Similarly, take the inflammation sample Input to transcriptome model In China, transcriptome models Output the high-dimensional embedding representation of the inflammation sample ; cancer samples Input to transcriptome model In China, transcriptome models Output the high-dimensional embedding representation of the cancer sample. .
[0024] Based on the high-dimensional embedding representation of each healthy sample, the mean vector of the healthy samples is calculated, which can be expressed by the following formula: . This represents the number of healthy samples. Using the same calculation method described above, the mean vector of the inflammatory samples is calculated. and the mean vector of cancer samples .
[0025] Based on the mean vector of the healthy samples and the high-dimensional embedding representation of each healthy sample, the health covariance matrix is calculated. Specifically, the health covariance matrix can be expressed by the following formula: .
[0026] Calculate the difference between the mean vector of inflammation samples and the mean vector of cancer samples, and normalize the difference using its magnitude to obtain the disease evolution direction vector. Specifically, the disease evolution direction vector can be calculated using the following formula: , This represents the magnitude of the difference between the mean vector of the inflammation sample and the mean vector of the cancer sample.
[0027] In this embodiment, a unified coordinate system is provided for subsequent gene-driven analysis by unifying the health background constraint and the disease evolution axis. A high-dimensional embedding representation of healthy samples is constructed, and a health covariance matrix is established to characterize the statistical distribution structure of gene expression under normal physiological conditions. This allows the gene screening process to explicitly model the statistical fluctuation structure of genes under normal physiological conditions, effectively reducing the interference of background noise on the results. Based on the high-dimensional embedding representation of samples from different disease stages, a disease evolution direction vector is constructed to characterize the overall direction of change during the transformation from inflammation to cancer. This ensures that the gene screening process maintains directional consistency with the dynamic evolutionary path from inflammation to cancer, improving the biological interpretability of subsequent gene screening results.
[0028] S120. Identify the base sample in the inflammatory sample and determine the perturbation sample after virtual perturbation of the biomarker corresponding gene in the base sample. Based on the high-dimensional embedding representation of the perturbation sample and the high-dimensional embedding representation of the base sample, determine the system shift vector of the biomarker corresponding gene.
[0029] In this context, the baseline samples are representative transcriptome samples from the inflammatory phase. In this embodiment, selecting representative baseline samples from the inflammatory samples as the targets of gene perturbation improves the representativeness and stability of single-gene virtual perturbation analysis and avoids bias in perturbation response estimation caused by extreme samples.
[0030] Biomarkers are objectively measurable biological indicators. Genes can serve as biomarkers through sequence variations or transcriptional expression levels. Transcriptome-based biomarkers use gene expression values to distinguish different physiological or disease states. In this embodiment, a single biomarker's corresponding gene is used as the object for virtual perturbation.
[0031] Virtual gene perturbation refers to simulating changes in gene function to extrapolate the effects of global transcriptome and cellular state changes. In this embodiment, a single virtual perturbation is performed on the gene corresponding to each biomarker. During any perturbation process, only the expression value of the gene corresponding to one biomarker is changed, while the expression values of the genes corresponding to other biomarkers remain unchanged. Expression values are the core observational data of transcriptome samples, measuring the on / off state and activity level of genes.
[0032] In this embodiment, an independent perturbation scenario for genes is constructed by controlling a single variable to avoid coupling interference introduced by synchronous changes in multiple genes, thereby achieving an independent quantitative assessment of the impact of a single gene on the global gene expression regulation system.
[0033] Compared to the baseline sample, only the expression values of the genes corresponding to the target biomarkers changed in the perturbation sample, while the expression values of the other genes remained unchanged.
[0034] The system shift vector is used to characterize the changes in the high-dimensional embedding representation of a sample before and after a single gene perturbation. The system shift vector can characterize the intensity and direction of the influence of a single gene perturbation on the global transcriptional regulatory state.
[0035] Specifically, S120 can include: S21. Select basic samples from the inflammation samples. The selection method includes at least one of the following: random selection, selecting the sample whose expression value is closest to the mean vector or median vector, or selecting the sample whose expression value is closest to the center of the embedding space of the high-dimensional embedding representation or low-dimensional embedding representation of the inflammation sample. S22. Perform virtual perturbation on the basic sample for the gene corresponding to a single biomarker to obtain a perturbed sample; S23. Determine the high-dimensional embedding representation of the perturbation sample and the high-dimensional embedding representation of the base sample, and use the difference between the high-dimensional embedding representation of the perturbation sample and the high-dimensional embedding representation of the base sample as the system shift vector of the gene corresponding to the biomarker.
[0036] The selection of the basic sample can be done in several ways: random selection, where a single sample is randomly selected from the inflammatory samples; statistical center selection, where the sample whose expression value is closest to the mean vector or median vector is selected from the inflammatory samples; or embedding space center selection, where the inflammatory samples are mapped to a high-dimensional or low-dimensional embedding representation space, and the sample closest to the center of the overall embedding space is selected as the basic sample to characterize the typical molecular state of the inflammatory stage.
[0037] Taking the selection of the center of the embedding space as an example, the specific explanation is as follows: the inflammation sample is mapped to a high-dimensional embedding representation space. , Calculate the embedding space center vector of the inflammatory sample: The basic sample can be represented by the following formula: , This represents the set of inflammation samples. Specifically, it selects the inflammation sample in the high-dimensional embedding representation space that has the closest distance to the center vector of the overall embedding space as the base sample.
[0038] In this embodiment, by using a basic sample selection method based on embedded spatial center, the distribution bias caused by random selection can be effectively avoided, and the selected basic sample can be used to characterize the typical molecular state of the disease stage, thereby improving the stability and consistency of subsequent analysis of the global regulatory response to single gene perturbation.
[0039] Specifically, in inflammation samples In the middle, select the basic sample , recorded as G represents the total number of genes. A virtual perturbation is applied to gene g corresponding to a single biomarker to obtain a perturbed sample. : Among them, perturbation samples Compared to the base sample Only the expression value of gene g corresponding to a single biomarker was modified, and the expression value of gene g was replaced with the expression value of gene g from the most representative sample in the cancer sample. (The most representative sample was selected in the same way as the baseline sample.)
[0040] The basic samples are input into the transcriptome model to obtain the high-dimensional embedding representation z output by the model, where... Furthermore, by inputting the perturbed samples into the transcriptome model, a high-dimensional embedding representation of the model output is obtained. , .
[0041] System shift vector of genes corresponding to biomarkers It can be expressed by the following formula: The system shift vector can be used to characterize the intensity and direction of the influence of single-gene perturbations on the global transcriptional regulatory state.
[0042] In this embodiment, by performing a virtual perturbation on a single gene in the basic sample, the changes in the high-dimensional embedding representation of the system before and after the perturbation are obtained, and the system displacement vector is calculated to simulate the impact of gene expression changes on the global regulatory system, thereby achieving an approximate characterization of the potential causal driving relationship between genes and disease changes.
[0043] S130. Based on the system displacement vector and the disease evolution direction vector, determine the effective perturbation component along the disease evolution direction, and weight the effective perturbation component according to the health covariance matrix to obtain the gene drive score of the gene corresponding to the biomarker.
[0044] In this embodiment, the system displacement vector is projected onto the disease evolution direction vector to obtain the effective perturbation component along the disease evolution direction, which characterizes the effective response strength of a single-gene perturbation in the disease evolution direction. The effective perturbation component allows us to measure whether the single-gene perturbation drives the system state to shift along the path from inflammation to cancer.
[0045] Understandably, since the system displacement vector may simultaneously contain disease-related changes, irrelevant noise, and state fluctuations in other directions, projecting it onto the disease evolution direction can extract effective perturbation components consistent with the disease evolution progress and filter out changes unrelated to the disease evolution direction, thereby improving the directionality and biological interpretability of driver gene screening.
[0046] In this embodiment, the effective perturbation components are weighted and scored based on the health covariance matrix, which can then normalize and correct the statistical fluctuation differences in different dimensions of the embedded feature space. Specifically, when the perturbation of the gene corresponding to the biomarker has a large shift in the direction corresponding to the low variance dimension in the healthy state, the corresponding scoring weight factor is increased; when the perturbation of the gene corresponding to the biomarker is mainly distributed in the high variance dimension in the healthy state, the corresponding scoring weight factor is decreased, so as to suppress the healthy background noise structure and enhance the stable biological signal.
[0047] Gene drive scoring is a quantitative evaluation of the driving contribution of genes in the transformation of inflammation into cancer under healthy background constraints. Screening key driver genes through gene drive scoring improves the accuracy and reliability of key driver gene screening.
[0048] Specifically, S130 may include: S31. Transpose the system displacement vector and perform a vector dot product with the disease evolution direction vector to obtain the effective perturbation component along the disease evolution direction. S32. Calculate the Mahalanobis distance based on the health covariance matrix and the system displacement vector, and use the Mahalanobis distance as the weight of the effective perturbation component to calculate the gene drive score of the gene corresponding to the biomarker.
[0049] After transposing the system displacement vector, perform a vector dot product with the disease evolution direction vector to obtain the effective perturbation component along the disease evolution direction, which can be expressed by the following formula: ,in, Indicates the effective perturbation component. Let be the system displacement vector. This represents the direction vector of disease evolution.
[0050] The Mahalanobis distance, calculated from the health covariance matrix and the system displacement vector, can be expressed by the following formula: , Represents Mahalanobis distance, For the health covariance matrix, Let be the system displacement vector.
[0051] In this embodiment, Mahalanobis distance is used as the weight of the effective perturbation component, which enables scale correction of the health fluctuation characteristics of different dimensions of the embedded space.
[0052] The gene drive score of the gene corresponding to the biomarker can be expressed by the following formula: ,in, Gene drive score representing the gene corresponding to the biomarker.
[0053] S140. Based on the gene drive score corresponding to each biomarker gene, determine the set of target biomarker genes driving inflammation-cancer transformation from each biomarker gene set.
[0054] Since the gene drive score measures the driving contribution of genes in the direction of disease evolution, which drives inflammation to cancer transformation under the constraint of a healthy background, the genes corresponding to key biomarkers driving inflammation-cancer transformation can be screened by calculating the gene drive score of each biomarker corresponding to the total number of genes G, and then ranking the gene drive scores of each biomarker corresponding to the genes.
[0055] In an optional embodiment, a gene drive score threshold can be set, and genes corresponding to biomarkers with gene drive scores greater than or equal to the gene drive score threshold can be added to the target biomarker corresponding gene set.
[0056] In another optional embodiment, the genes corresponding to the biomarkers can be sorted in descending order of gene drive score, and a preset number of biomarker-corresponding genes can be selected and added to the target biomarker-corresponding gene set. The preset number can be a fixed value or a value calculated based on the total number of genes and a preset ratio.
[0057] In yet another alternative embodiment, S140 may include: S41. Sort the gene drive scores corresponding to each biomarker in descending order to obtain the gene drive score sequence. S42. Identify the inflection point of the gene-driven scoring sequence through the second-order difference structure, and determine the set of target biomarkers corresponding to the genes that drive the transformation of inflammation and cancer based on the inflection point position and the corresponding genes of each biomarker before the inflection point position.
[0058] In this embodiment, the gene drive scoring sequence can be represented by S(i), and the inflection point of the gene drive scoring sequence is identified through a second-order difference structure. Specifically, it can be represented by the following formula: The inflection point is defined as: Where i represents the gene drive score sequence The position index in the middle, This represents the curvature change threshold, used to determine whether significant structural changes have occurred in the gene-driven scoring sequence. This indicates that the intensity of the local curvature change has exceeded the preset threshold for the first time. The location is used to determine the inflection point between the strong driving signal region and the background noise region. The inflection point is the location where a significant curvature change first appears in the gene-driven scoring sequence, and significance is determined by the absolute value of the second-order difference exceeding an adaptive threshold. The inflection point location and the corresponding genes of each biomarker before the inflection point location are all added to the target biomarker corresponding gene set: .
[0059] It should be noted that this embodiment only uses the identification of inflection points in gene-driven scoring sequences using a second-order difference structure as an example to illustrate the process of screening genes in gene-driven scoring sequences. Inflection points can also be determined using first-order rate of change or second-order change characteristics. Specifically, based on the distribution structure analysis results of gene-driven scoring sequences, statistical modeling is performed on the overall distribution characteristics of the gene-driven scoring sequences to determine whether the gene-driven scores conform to a preset distribution pattern. The preset distribution patterns include normal distribution characteristics, right-skewed long-tailed distribution characteristics, or highly concentrated distribution characteristics. Based on the inflection point location and the distribution characteristics of gene-driven scores, a preset screening strategy is used to determine the gene set corresponding to the target biomarker. The preset screening strategy includes a screening strategy based on inflection point truncation, a screening strategy based on percentile thresholds, or a screening strategy based on statistical thresholds. The statistical threshold screening strategy includes screening based on threshold ranges constructed from the mean and standard deviation.
[0060] The technical solution of this invention establishes a health covariance matrix based on healthy samples to characterize the statistical distribution structure of gene expression under normal physiological conditions; constructs disease evolution direction vectors based on samples from different disease stages to characterize the overall change direction in the process of inflammation transforming into cancer; performs virtual perturbation on the genes corresponding to single biomarkers in the basic samples determined in the inflammation samples to obtain the changes in high-dimensional embedding representations before and after perturbation, calculates the system displacement vector to simulate the impact of gene expression changes on the global regulatory system, and achieves an approximate characterization of potential causal driving relationships; projects the system displacement vector onto the disease evolution direction and performs weighted processing in conjunction with the health covariance matrix to uniformly quantify and score the perturbation response of each gene; finally, it screens the set of genes corresponding to biomarkers based on the gene driving score. This invention, by integrating health background constraints, disease evolution direction constraints, single-gene causal perturbation analysis, and unified quantitative scoring of gene perturbation responses, performs multi-level decoupling and quantitative characterization of pathological driving signals in the gene regulation system, realizing the transformation from causal perturbation response to structured driving gene screening, and achieving accurate screening of key driving genes in the process of inflammation transforming into cancer.
[0061] Example 2 Figure 2 This is a flowchart of another biomarker screening method based on health background constraints and perturbation causality provided in Embodiment 2 of the present invention. Based on the above embodiments, this embodiment further adds a secondary gene screening process based on pseudo-time analysis of disease progression.
[0062] like Figure 2 As shown, the method includes: S210. Identify healthy samples, inflammatory samples, and cancer samples, and determine the health covariance matrix based on the healthy samples, and determine the disease evolution direction vector based on the inflammatory and cancer samples.
[0063] S220. Identify the base sample in the inflammatory sample and determine the perturbation sample after virtual perturbation of the base sample with the corresponding gene of the biomarker. Based on the high-dimensional embedding representation of the perturbation sample and the high-dimensional embedding representation of the base sample, determine the system shift vector of the corresponding gene of the biomarker.
[0064] S230. Based on the system displacement vector and the disease evolution direction vector, determine the effective perturbation component along the disease evolution direction, and weight the effective perturbation component according to the health covariance matrix to obtain the gene drive score of the gene corresponding to the biomarker.
[0065] S240. Sort the gene drive scores corresponding to the genes of each biomarker in descending order to obtain the gene drive score sequence.
[0066] S250. Identify the inflection point of gene-driven scoring sequences through second-order difference structures.
[0067] S260. The genes corresponding to each biomarker before the inflection point are used as candidate genes. Based on the candidate genes, inflammation samples, and cancer samples, a candidate gene expression matrix is constructed.
[0068] The above process has been described in the above embodiments, and will not be repeated here. Based on the content described in the above embodiments, this embodiment adds a secondary gene screening process that combines statistical correlation based on disease progression pseudo-time analysis, thereby realizing the consistency verification between genes and disease trajectory evolution, so as to further screen the genes corresponding to the final biomarkers.
[0069] The candidate gene expression matrix can be represented by the following formula: Where M is the sum of the number of inflammation samples and the number of cancer samples, and N represents the number of candidate genes.
[0070] S270. Perform low-dimensional embedding representation on the candidate gene expression matrix to determine the pseudo-time values of disease progression for inflammatory and cancer samples in low-dimensional space.
[0071] The candidate gene expression matrix can be represented by a low-dimensional embedding using the following formula: ,in, This represents an arbitrary manifold learning or dimensionality reduction mapping method used to characterize the global and local structural neighborhood relationships between samples.
[0072] The pseudotime value of disease progression represents the relative progression order of a sample in the process of inflammation evolving into cancer. Taking the sample in the inflammation stage as the starting reference, the pseudotime value of disease progression shows a monotonic trend as the disease evolves into the cancer stage. The smaller the pseudotime value of disease progression, the closer it is to the early stage of inflammation, and the larger it is to the late stage of cancer.
[0073] In this embodiment, a low-dimensional embedding representation is used to represent the candidate gene expression matrix, constructing a disease progression trajectory space. Modeling is performed based on samples from different disease stages, and pseudo-time values of disease progression are calculated through diffusion processes or equivalent probability propagation mechanisms. By using trajectory modeling and pseudo-time analysis, static expression data is extended into a continuous evolutionary process, thereby enhancing the ability to characterize the dynamic structure of disease progression. The pseudo-time information of disease progression is used to verify the consistency of candidate gene expression dynamics, ultimately achieving precise screening of the driving biomarkers.
[0074] Specifically, S270 may include: performing low-dimensional embedding representation on the candidate gene expression matrix, determining the sample similarity matrix in the low-dimensional space, and determining the transition probability matrix based on the sample similarity matrix; determining the initial diffusion starting point in the inflammatory sample, diffusing the transition probability matrix, and mapping the diffusion process to a continuous scalar to obtain the pseudo-time value of disease progression for the inflammatory sample and the cancer sample; wherein, the pseudo-time value of disease progression represents the relative progression order of the sample in the process of evolution from inflammation to cancer.
[0075] The sample similarity matrix W is determined in a low-dimensional space, and the elements of the sample similarity matrix W can be represented by the following formula: ,in, This represents the similarity weight between sample i and sample j. This represents the low-dimensional embedding representation of sample i. This represents the low-dimensional embedding representation of sample j. This is the scale parameter.
[0076] The transition probability matrix, determined based on the sample similarity matrix, can be expressed by the following formula: , P is the transition probability matrix, specifically the transition probability matrix of a Markov random walk.
[0077] This is understandable, because the pseudotime value of disease progression represents the relative progression order of a sample in the process of inflammation evolving into cancer; therefore, the initial point of spread needs to be determined from the inflammatory sample. Specifically, the initial point of spread can be determined using the following formula: , This indicates that the initial diffusion origin is in the inflammatory sample. The position index in the middle, , Indicates sample The coordinate values on the first principal trajectory direction in the low-dimensional embedding space (e.g., the first dimension of UMAP (Uniform Manifold Approximation and Projection)). In this embodiment, the initial state of disease evolution is defined by the initial diffusion origin.
[0078] The transition probability matrix is diffused, and the diffusion process is mapped to a continuous scalar to obtain the pseudo-time values of disease progression for inflammation and cancer samples. Specifically, the diffusion process is defined as follows: t is the diffusion step number. This is the state transition probability matrix after t propagations. The diffusion process is mapped to a continuous scalar: ,in, , representing the i-th sample The pseudotime value of disease progression This represents the quasi-time mapping function, which can be implemented based on diffusion distance, mainstream direction projection, or sorting function.
[0079] In this embodiment, the diffusion process or equivalent probability propagation process of disease progression is defined based on the sample similarity matrix to characterize the state transition relationship between samples; based on the diffusion process or probability propagation process, the samples are continuously scalarized and mapped to obtain the pseudo-time value of disease progression; based on the pseudo-time value of disease progression, the sample distribution in the low-dimensional embedding space is continuously function-mapped or color-mapped to construct the disease progression trajectory space, which is used to characterize the continuous evolution structure of the inflammation-cancer transformation process.
[0080] S280. Determine the expression values of candidate genes in inflammation and cancer samples, and calculate the statistical correlation between candidate gene expression values and pseudotime values of disease progression.
[0081] The total number of inflammation samples and cancer samples is M, and the expression values of candidate gene g in the inflammation samples and cancer samples are denoted as expression vectors. The pseudotime values of disease progression for each sample are denoted as time vectors. .
[0082] In this embodiment, the linear correlation between the expression vector corresponding to the candidate gene expression value and the vector corresponding to the pseudo-time value of disease progression can be quantitatively analyzed using the Pearson correlation coefficient.
[0083] Specifically, it can be expressed by the following formula: ,in, Indicates statistical correlation, used to show the degree of synchronous change in gene expression as disease progresses. It is the covariance between the expression vector corresponding to the candidate gene expression value and the vector corresponding to the pseudo-time value of disease progression. This represents the standard deviation of the expression vector corresponding to the candidate gene expression value. The standard deviation of the vector corresponding to the pseudotime values of disease progression.
[0084] Understandably, this embodiment uses the Pearson correlation coefficient as an example to specifically explain the calculation method of statistical correlation. The Pearson correlation coefficient can be replaced by any equivalent linear or nonlinear correlation measure, and this embodiment does not limit it.
[0085] S290. Based on statistical correlation, determine the set of genes corresponding to the target biomarkers driving inflammation-cancer transformation among the candidate genes.
[0086] Furthermore, considering that a single statistical calculation may be affected by disturbances or data distribution bias, this embodiment further adopts repeated random subsampling or cross-validation to evaluate the robustness of statistical correlation. The statistical correlations obtained from multiple calculations are averaged and fused to obtain more stable statistical correlation results.
[0087] After obtaining the statistical correlation of each candidate gene, we can also analyze the statistical correlation of each candidate gene. Perform a significance test to obtain the significance level. The significance test process will not be described in detail in this embodiment. Meanwhile, to avoid false positives caused by multiple hypothesis testing, the Benjamini–Hochberg method is used to determine the significance level. Corrections are made to obtain the false detection rate. It is used to measure the reliability of statistical correlation results.
[0088] Based on statistical correlation, the set of genes corresponding to the target biomarkers driving inflammation-cancer transformation is determined from each candidate gene pool. Specifically, the candidate genes are sorted in descending order of their absolute statistical correlation values. From the sorted candidate genes, those with a statistical correlation greater than a first correlation threshold and a false discovery rate less than a first false discovery rate threshold are added to the set of upregulated genes significantly positively correlated with inflammation-cancer transformation. Simultaneously, from the sorted candidate genes, those with a negative statistical correlation but an absolute value greater than a second correlation threshold and a false discovery rate less than a second false discovery rate threshold are added to the set of downregulated genes significantly negatively correlated with inflammation-cancer transformation. The upregulated gene set and the downregulated gene set constitute the set of genes corresponding to the target biomarkers.
[0089] The first relevance threshold and the second relevance threshold can be the same or different, and are used to filter out non-significant genes; for example, they can be set to 0.35. Similarly, the first false discovery rate threshold and the second false discovery rate threshold can be the same or different; for example, they can be set to 0.05.
[0090] Furthermore, after determining the sets of upregulated and downregulated genes, a functional enrichment analysis and protein interaction network analysis model was constructed based on the sets of upregulated and downregulated genes to identify the core hub genes as the final set of biomarkers driving inflammation-cancer transformation.
[0091] In this embodiment, by combining statistical correlation analysis and significance testing, secondary verification and noise reduction of candidate genes are achieved, thereby improving the stability and reliability of gene screening results.
[0092] The technical solution of this embodiment, by integrating health background constraint modeling, disease evolution direction constraint, single-gene causal perturbation analysis, and dynamic trajectory and pseudo-time modeling, can perform multi-level screening and verification of key driver genes in the process of inflammation transforming into cancer under a unified model. It achieves multi-level decoupling and quantitative characterization of pathological driver signals in the gene regulation system, effectively distinguishes between correlation and causal driving in gene expression changes, significantly reduces the false positive rate in the screening process, and improves the stability, interpretability, and biological consistency of biomarker identification. It can be used in application scenarios such as disease mechanism analysis, early diagnostic biomarker mining, and targeted therapy target development.
[0093] Example 3 Figure 3 This is a schematic diagram of a biomarker screening device based on health background constraints and perturbation causality provided in Embodiment 3 of the present invention. Figure 3 As shown, the device includes: The health covariance matrix and disease evolution direction vector determination module 310 is used to determine healthy samples, inflammatory samples and cancer samples, and to determine the health covariance matrix based on the healthy samples and the disease evolution direction vector based on the inflammatory samples and cancer samples. The system displacement vector determination module 320 is used to determine the base sample in the inflammation sample and the perturbation sample after virtual perturbation processing of the base sample with the corresponding gene of the biomarker. Based on the high-dimensional embedding representation of the perturbation sample and the high-dimensional embedding representation of the base sample, the system displacement vector of the corresponding gene of the biomarker is determined. The gene drive score determination module 330 is used to determine the effective perturbation component along the disease evolution direction based on the system displacement vector and the disease evolution direction vector, and to perform weighted scoring on the effective perturbation component based on the health covariance matrix to obtain the gene drive score of the gene corresponding to the biomarker. The target biomarker corresponding gene set determination module 340 is used to determine the target biomarker corresponding gene set driving inflammation-cancer transformation from each biomarker corresponding gene based on the gene drive score corresponding to each biomarker corresponding gene.
[0094] The technical solution of this invention establishes a health covariance matrix based on healthy samples to characterize the statistical distribution structure of gene expression under normal physiological conditions; constructs disease evolution direction vectors based on samples from different disease stages to represent the overall change direction in the process of inflammation transforming into cancer; performs virtual perturbation on the genes corresponding to single biomarkers in the basic samples determined in the inflammation samples to obtain the changes in high-dimensional embedding representations before and after perturbation, calculates the system displacement vector to simulate the impact of gene expression changes on the global regulatory system, and achieves an approximate characterization of potential causal driving relationships; projects the system displacement vector onto the disease evolution direction and performs weighted processing in conjunction with the health covariance matrix to uniformly quantify and score the perturbation response of each gene; finally, it screens the set of genes corresponding to biomarkers based on the gene driving score. This invention, by integrating health background constraints, disease evolution direction constraints, single-gene causal perturbation analysis, and unified quantitative scoring of gene perturbation responses, performs multi-level decoupling and quantitative characterization of pathological driving signals in the gene regulation system, and achieves accurate screening of key driving genes in the process of inflammation transforming into cancer.
[0095] Based on the above embodiments, optionally, the health covariance matrix and disease evolution direction vector determination module 310 includes: The high-dimensional embedding representation unit is used to process healthy samples, inflammatory samples, and cancer samples respectively through a pre-trained transcriptome model to obtain the high-dimensional embedding representations corresponding to the healthy samples, inflammatory samples, and cancer samples. The mean vector determination unit is used to calculate the mean vector of healthy samples based on the high-dimensional embedding representation of each healthy sample, to calculate the mean vector of inflammatory samples based on the high-dimensional embedding representation of each inflammatory sample, and to calculate the mean vector of cancer samples based on the high-dimensional embedding representation of each cancer sample. The health covariance matrix determination unit is used to calculate the health covariance matrix based on the mean vector of the health samples and the high-dimensional embedding representation of each health sample. The disease evolution direction vector determination unit is used to calculate the difference between the mean vector of inflammation samples and the mean vector of cancer samples, and normalize the difference by the magnitude of the difference to obtain the disease evolution direction vector.
[0096] Based on the above embodiments, optionally, the system displacement vector determination module 320 includes: The basic sample selection unit selects basic samples from the inflammation samples. The selection method includes at least one of the following: random selection, selecting the sample whose expression value is closest to the mean vector or median vector, or selecting the sample whose expression value is closest to the center of the embedding space of the high-dimensional embedding representation or low-dimensional embedding representation of the inflammation sample. The perturbation sample determination unit is used to perform virtual perturbation processing on the basic sample for the gene corresponding to a single biomarker to obtain the perturbation sample; The system displacement vector determination unit is used to determine the high-dimensional embedding representation of the perturbed sample and the high-dimensional embedding representation of the base sample, and uses the difference between the high-dimensional embedding representation of the perturbed sample and the high-dimensional embedding representation of the base sample as the system displacement vector of the gene corresponding to the biomarker.
[0097] Based on the above embodiments, optionally, the gene drive scoring determination module 330 includes: The effective disturbance component determination unit is used to transpose the system displacement vector and perform a vector dot product with the disease evolution direction vector to obtain the effective disturbance component along the disease evolution direction. The gene drive score determination unit is used to calculate the Mahalanobis distance based on the health covariance matrix and the system displacement vector, and to use the Mahalanobis distance as the weight of the effective perturbation component to calculate the gene drive score of the gene corresponding to the biomarker.
[0098] Based on the above embodiments, optionally, the target biomarker corresponding gene set determination module 340 includes: The gene drive score sorting unit is used to sort the gene drive scores corresponding to each biomarker in descending order to obtain the gene drive score sequence. The target biomarker corresponding gene set determination unit is used to identify the inflection point of the gene-driven scoring sequence through a second-order difference structure, and determine the target biomarker corresponding gene set driving inflammation-cancer transformation based on the inflection point position and the corresponding genes of each biomarker before the inflection point position.
[0099] Based on the above embodiments, optionally, the target biomarker corresponding gene set determination unit is specifically used for: The genes corresponding to each biomarker at the inflection point and before the inflection point are used as candidate genes. Based on the candidate genes, as well as inflammation and cancer samples, a candidate gene expression matrix is constructed. The candidate gene expression matrix is represented by low-dimensional embedding to determine the pseudo-time values of disease progression in inflammatory and cancer samples in low-dimensional space; Determine the expression values of candidate genes in inflammation and cancer samples, and calculate the statistical correlation between candidate gene expression values and pseudotime values of disease progression; Based on statistical correlation, the set of genes corresponding to the target biomarkers driving inflammation-cancer transformation was determined among the candidate genes.
[0100] Based on the above embodiments, optionally, the target biomarker corresponding gene set determination unit is specifically used for: The candidate gene expression matrix is represented by a low-dimensional embedding, the sample similarity matrix is determined in the low-dimensional space, and the transition probability matrix is determined based on the sample similarity matrix. In inflammatory samples, the initial diffusion starting point is determined, the transition probability matrix is diffused, and the diffusion process is mapped to a continuous scalar to obtain the pseudo-time values of disease progression for inflammatory and cancer samples. Among them, the pseudotime value of disease progression represents the relative progression order of the sample in the process of inflammation evolving into cancer.
[0101] The biomarker screening device based on health background constraints and perturbation causality provided in the embodiments of the present invention can execute the biomarker screening method based on health background constraints and perturbation causality provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.
[0102] Example 4 Figure 4 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, 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 (e.g., helmets, glasses, watches, etc.), 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 invention described and / or claimed herein.
[0103] like Figure 4 As shown, the 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 program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can 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.
[0104] 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.
[0105] Processor 11 can be a variety of 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), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as biomarker screening methods based on health background constraints and perturbation causality.
[0106] In some embodiments, the biomarker screening method based on health background constraints and perturbation causality can 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 can 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 biomarker screening method based on health background constraints and perturbation causality described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the biomarker screening method based on health background constraints and perturbation causality by any other suitable means (e.g., by means of firmware).
[0107] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments 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.
[0108] Computer programs used to implement the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to the processor of a general-purpose computer, a special-purpose computer, or other programmable biomarker screening device based on health background constraints and perturbation causality, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer programs can 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.
[0109] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.
[0110] 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).
[0111] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or 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.
[0112] 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.
[0113] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.
[0114] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. 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 invention should be included within the scope of protection of this invention.
Claims
1. A biomarker screening method based on health background constraints and perturbation causality, characterized in that, include: Identify healthy samples, inflammatory samples, and cancer samples; determine the health covariance matrix based on the healthy samples; and determine the disease evolution direction vector based on the inflammatory and cancer samples. In the inflammatory samples, the base sample is identified, and the perturbation sample is identified after virtual perturbation of the base sample with the corresponding gene of the biomarker. Based on the high-dimensional embedding representation of the perturbation sample and the high-dimensional embedding representation of the base sample, the system shift vector of the corresponding gene of the biomarker is determined. Based on the system displacement vector and the disease evolution direction vector, the effective perturbation component along the disease evolution direction is determined, and the effective perturbation component is weighted and scored according to the health covariance matrix to obtain the gene drive score of the corresponding gene of the biomarker. Based on the gene drive scores corresponding to each biomarker, the set of target biomarker genes driving inflammation-cancer transformation is determined among the corresponding genes of each biomarker.
2. The method according to claim 1, characterized in that, Identify healthy, inflammatory, and cancer samples. Calculate the health covariance matrix based on the healthy samples and the disease evolution direction vector based on the inflammatory and cancer samples, including: By using a pre-trained transcriptome model, healthy samples, inflammatory samples, and cancer samples are processed respectively to obtain high-dimensional embedding representations corresponding to healthy samples, inflammatory samples, and cancer samples. The mean vector of healthy samples is calculated based on the high-dimensional embedding representation of each healthy sample; the mean vector of inflammatory samples is calculated based on the high-dimensional embedding representation of each inflammatory sample; and the mean vector of cancer samples is calculated based on the high-dimensional embedding representation of each cancer sample. The health covariance matrix is calculated based on the mean vector of the healthy samples and the high-dimensional embedding representation of each healthy sample. The difference between the mean vector of inflammation samples and the mean vector of cancer samples is calculated, and the difference is normalized by the magnitude of the difference to obtain the disease evolution direction vector.
3. The method according to claim 1, characterized in that, In inflammatory samples, a baseline sample is identified, and a perturbation sample is determined after virtual perturbation of the baseline sample with the corresponding biomarker gene. Based on the high-dimensional embedding representations of the perturbation sample and the baseline sample, the system shift vector of the corresponding biomarker gene is determined, including: The basic sample is selected from the inflammation sample. The selection method includes at least one of the following: random selection, selecting the sample whose expression value is closest to the mean vector or median vector, or selecting the sample whose expression value is closest to the center of the embedding space of the high-dimensional embedding representation or low-dimensional embedding representation of the inflammation sample. The base sample is subjected to virtual perturbation of the gene corresponding to a single biomarker to obtain the perturbed sample; The high-dimensional embedding representations of the perturbed samples and the basic samples are determined, and the difference between the high-dimensional embedding representations of the perturbed samples and the basic samples is used as the system shift vector of the corresponding gene of the biomarker.
4. The method according to claim 1, characterized in that, Based on the system displacement vector and the disease evolution direction vector, the effective perturbation components along the disease evolution direction are determined. These effective perturbation components are then weighted and scored according to the health covariance matrix to obtain the gene drive score of the corresponding gene for the biomarker, including: After transposing the system displacement vector, perform a vector dot product with the disease evolution direction vector to obtain the effective perturbation component along the disease evolution direction. Mahalanobis distance is calculated based on the health covariance matrix and the system displacement vector, and the Mahalanobis distance is used as the weight of the effective perturbation component to calculate the gene drive score of the gene corresponding to the biomarker.
5. The method according to claim 1, characterized in that, Based on the gene drive scores corresponding to each biomarker, the set of target biomarker genes driving inflammation-cancer transformation was determined from the genes corresponding to each biomarker, including: The gene drive scores corresponding to each biomarker were sorted in descending order to obtain the gene drive score sequence. The inflection point of the gene-driven scoring sequence was identified by using a second-order difference structure, and the set of target biomarkers corresponding to the genes driving the transformation of inflammation and cancer was determined based on the position of the inflection point and the corresponding genes of each biomarker before the inflection point.
6. The method according to claim 5, characterized in that, Based on the inflection point and the corresponding genes of each biomarker before the inflection point, the set of target biomarkers and corresponding genes driving inflammation-cancer transformation was determined, including: The genes corresponding to each biomarker at the inflection point and before the inflection point are used as candidate genes. Based on the candidate genes, as well as inflammation and cancer samples, a candidate gene expression matrix is constructed. The candidate gene expression matrix is represented by low-dimensional embedding to determine the pseudo-time values of disease progression in inflammatory and cancer samples in low-dimensional space; Determine the expression values of candidate genes in inflammation and cancer samples, and calculate the statistical correlation between candidate gene expression values and pseudotime values of disease progression; Based on statistical correlation, the set of genes corresponding to the target biomarkers driving inflammation-cancer transformation was determined among the candidate genes.
7. The method according to claim 6, characterized in that, Low-dimensional embedding representation of candidate gene expression matrices is used to determine pseudo-time values of disease progression in inflammatory and cancer samples in low-dimensional space, including: The candidate gene expression matrix is represented by a low-dimensional embedding, the sample similarity matrix is determined in the low-dimensional space, and the transition probability matrix is determined based on the sample similarity matrix. In inflammatory samples, the initial diffusion starting point is determined, the transition probability matrix is diffused, and the diffusion process is mapped to a continuous scalar to obtain the pseudo-time values of disease progression for inflammatory and cancer samples. Among them, the pseudotime value of disease progression represents the relative progression order of the sample in the process of inflammation evolving into cancer.
8. A biomarker screening device based on health background constraints and perturbation causality, characterized in that, include: The module for determining the health covariance matrix and the disease evolution direction vector is used to identify healthy samples, inflammatory samples, and cancer samples, and to determine the health covariance matrix based on the healthy samples and the disease evolution direction vector based on the inflammatory and cancer samples. The system displacement vector determination module is used to determine the base sample in the inflammatory sample and the perturbation sample after virtual perturbation of the base sample with the corresponding gene of the biomarker. Based on the high-dimensional embedding representation of the perturbation sample and the high-dimensional embedding representation of the base sample, the system displacement vector of the corresponding gene of the biomarker is determined. The gene drive score determination module is used to determine the effective perturbation component along the disease evolution direction based on the system displacement vector and the disease evolution direction vector, and to weight the effective perturbation component according to the health covariance matrix to obtain the gene drive score of the gene corresponding to the biomarker. The target biomarker corresponding gene set determination module is used to determine the target biomarker corresponding gene set driving inflammation-cancer transformation from each biomarker corresponding gene based on the gene drive score of each biomarker corresponding gene.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the biomarker screening method based on health background constraints and perturbation causality as described in any one of claims 1-7.
10. A storage medium for storing computer-executable instructions, characterized in that, The computer-executable instructions, when executed by a computer processor, are used to perform the biomarker screening method based on health background constraints and perturbation causality as described in any one of claims 1-7.
