Dasatinib drug sensitivity gene marker combination and drug sensitivity prediction system

Abstract

The invention discloses a dasatinib drug sensitivity gene marker combination and a drug sensitivity prediction system. A signal interaction network is constructed and key regulatory factors are identified by comprehensively analyzing a gene expression profile, a gene marker combination remarkably related to the drug sensitivity of the dasatinib is obtained, and a prediction model can be trained through the gene markers and used for accurately predicting the drug sensitivity of the dasatinib, so that a reliable basis is provided for personalized treatment, and the drug sensitivity of the dasatinib is accurately predicted. And reference is provided for clinical treatment decisions.

Description

Technical Field

[0001] The present invention belongs to the technical field of drug sensitivity prediction, and in particular relates to a dasatinib drug sensitivity gene marker combination and a drug sensitivity prediction system. Background Art

[0002] At present, the methods of drug sensitivity detection mainly include in vitro cell experiments, PCR mutation detection, RNA sequencing (RNA-Seq) analysis, single cell sequencing analysis (scRNA-seq), proteomics analysis, etc. These methods provide a basis for drug selection to a certain extent, but they have obvious shortcomings in detection range, sensitivity and accuracy. For example:

[0003] Although in vitro cell experiments are simple to operate and low-cost, they cannot fully reflect the complex in vivo environment, and the representativeness of cell lines is limited; PCR mutation detection can detect gene mutations associated with drug sensitivity (such as BCR-ABL), but it can only detect known genes or specific types of mutations, and cannot discover new mutations and cannot fully understand the molecular characteristics of tumors; single-cell sequencing analysis (scRNA-seq) can analyze heterogeneity at the single-cell level and can discover rare cell populations, but it is expensive, the data analysis is complex, and the data noise is also large.

[0004] Transcriptome sequencing (RNA-Seq) is a high-throughput sequencing technology that can reveal transcriptional changes of cells under different conditions by comprehensively detecting gene expression profiles. In recent years, the application of RNA-Seq in tumor drug sensitivity detection has gradually increased. Its cost and technical requirements are lower than single-cell sequencing analysis. However, the amount of raw data generated by RNA-Seq is large, and the data analysis process is complex. It requires professional bioinformatics tools and technologies, and the analysis cost and time are high, which limits its widespread application.

[0005] Proteomic analysis provides direct functional information on drug sensitivity by detecting protein expression and modification in cells. However, due to the complexity of proteomic technology and the need for high-precision equipment, its application in clinical testing is still limited.

[0006] Dasatinib is an oral drug that targets BCR-ABL kinase and Src family kinases and is widely used in the treatment of chronic myeloid leukemia (CML) and acute lymphoblastic leukemia (ALL). It prevents the growth and division of cancer cells by inhibiting the activity of these kinases. Although dasatinib has shown significant efficacy in the treatment of certain types of leukemia, there are significant differences in the response of different patients to the drug. Therefore, determining the patient's drug sensitivity to dasatinib is crucial to optimizing treatment options, improving efficacy and reducing side effects. Summary of the invention

[0007] The purpose of the present invention is to propose a dasatinib drug sensitivity gene marker combination and a drug sensitivity prediction system, which is used to accurately predict dasatinib drug sensitivity by comprehensively analyzing gene expression profiles and related molecular markers, thereby providing a reliable basis for personalized treatment and a reference for clinical treatment decisions.

[0008] In view of this, the scheme of the present invention is as follows:

[0009] The first aspect of the present invention proposes the use of a combination of gene markers in the preparation of a dasatinib drug sensitivity prediction product; the combined gene markers include LCK, TGFBR1, TEC, PDGFRB, PAG1, IL4, CD28, PTCRA, PDCD1, TNK2, CD3G, MYLK, and SYK.

[0010] In a second aspect of the present invention, a dasatinib drug sensitivity detection kit is provided, comprising detection reagents for gene markers, wherein the gene markers include LCK, TGFBR1, TEC, PDGFRB, PAG1, IL4, CD28, PTCRA, PDCD1, TNK2, CD3G, MYLK, and SYK.

[0011] Furthermore, the kit is selected from a transcriptome sequencing library construction kit and a PCR amplification kit.

[0012] The third aspect of the present invention provides a dasatinib drug sensitivity prediction system, comprising:

[0013] A gene expression acquisition unit is used to acquire the expression level of the target gene of the sample to be tested; the target genes include LCK, TGFBR1, TEC, PDGFRB, PAG1, IL4, CD28, PTCRA, PDCD1, TNK2, CD3G, MYLK, and SYK;

[0014] The drug sensitivity prediction unit inputs the expression level of each target gene into a drug sensitivity prediction model to obtain a prediction result; the drug sensitivity prediction model uses a multi-layer perceptron algorithm, selects the target gene as a feature and is trained based on a patient transcriptome data sample.

[0015] Furthermore, the target gene is screened by the following method:

[0016] Constructing a gene expression matrix based on the transcriptome data sample library, including a first matrix based on drug-resistant samples and a second matrix based on drug-inresistant samples;

[0017] The significant gene pairs are selected according to the mutual information values ​​of each group of genes in the first matrix and the second matrix, and the first signal interaction network and the second signal interaction network are constructed respectively with each gene as a node;

[0018] The differential gene network was constructed by network fusion based on the first signal interaction network and the second signal interaction network, and genes with significant differences were screened as target genes.

[0019] Preferably, the differential gene network analyzes and identifies genes with significant differences based on the betweenness centrality value and the eigenvector centrality value, respectively, and the two groups of differential genes obtained are merged and deduplicated to obtain the target gene.

[0020] Furthermore, the drug sensitivity prediction model is trained by taking the target gene expression level in the sample to generate a target gene expression matrix, and dividing the target gene expression matrix into a training set, a validation set and a test set.

[0021] Preferably, the model is cross-validated using a test set.

[0022] Furthermore, the expression level of the target gene in the sample to be tested is obtained by performing transcriptome sequencing or single-cell sequencing and quantitative analysis on a clinical sample of a patient, and the clinical sample is a blood sample.

[0023] Preferably, the quantitative analysis process of the target gene expression is to remove low-quality data in the sequencing data, align it to the reference genome, and quantify the gene expression.

[0024] Furthermore, the drug sensitivity prediction unit is used to predict the dose of dasatinib corresponding to the cell death rate, and to determine whether the sample to be tested is sensitive to dasatinib according to the sensitivity standard.

[0025] Compared with the prior art, the beneficial effects of the present invention include but are not limited to:

[0026] The gene marker combination of the present invention is obtained by comprehensively analyzing gene expression profiles to construct a signal interaction network and identify key regulatory factors, and therefore has a significant correlation with dasatinib drug sensitivity, and can be used to prepare a product for predicting dasatinib drug sensitivity, providing a reference for clinical treatment decisions.

[0027] The prediction system of the present invention is based on the identified key regulatory factors, uses a multi-layer perceptron algorithm to build a prediction model, predicts drug sensitivity, and evaluates model performance, which can achieve efficient and accurate prediction of dasatinib drug sensitivity; and has significant application value in clinical practice for optimizing treatment plans, improving efficacy, and reducing side effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0029] Figure 1 The figure is a flow chart for constructing the drug sensitivity prediction model of the present invention.

[0030] Figure 2 The present invention is based on the dasatinib drug sensitivity-related gene expression matrix.

[0031] Figure 3 Schematic diagram of the gene signal interaction network constructed in the present invention.

[0032] Figure 4 Schematic diagram of the differential gene signal interaction network constructed in the present invention.

[0033] Figure 5 This is the performance evaluation result of the model trained by the present invention. DETAILED DESCRIPTION

[0034] The following provides definitions of some terms used in this specification. Unless otherwise specified, all terms used herein are generally understood by those skilled in the art to which this solution belongs.

[0035] In the following example, we have a group of RNA-seq gene expression data of patients who are sensitive and insensitive to dasatinib, as well as the corresponding dasatinib LC50 results. Therefore, taking this as an example, we construct a drug resistance prediction model. The specific process is as follows: Figure 1 As shown in the figure, the model construction process includes the following steps:

[0036] 1. Data preparation and preprocessing

[0037] The patient's RNA-seq results are aligned to the reference gene, and the expression level (read count) of each gene is calculated using a file (GTF) containing gene location information. Patient samples are divided into resistant samples and non-resistant samples based on clinical diagnosis. For each group of samples, the expression levels of all patient genes are merged together, with rows representing genes and columns representing patients. The data are normalized using a standardization method. Low-expression genes are removed, and high-expression genes are retained to form an expression matrix. The expression matrix is ​​in the form of Figure 2 shown.

[0038] 2. Signal interaction network construction

[0039] In this step, we aim to construct a signal interaction network between genes by calculating the mutual information of gene expression data. The specific steps are as follows:

[0040] 1) Mutual Information Calculation

[0041] For the gene expression matrix X of the sample, where the rows represent genes and the columns represent samples, we first calculate the mutual information MI(i, j) between the expression levels of any two genes i and j in all samples. The mutual information MI(i, j) measures the statistical dependence between gene i and gene j, and the formula is as follows:

[0042]

[0043] Among them, p(x i , x j ) is the joint probability distribution of genes i and j in the sample, p(x i ) and p(y i ) are their marginal probability distributions respectively.

[0044] 2) Mutual information matrix construction

[0045] Calculate the mutual information values ​​of all gene pairs (i, j) to obtain a symmetric mutual information matrix M, where M[i][j] = MI(i, j). This matrix reflects the relationship between gene pairs.

[0046] 3) Threshold screening

[0047] A threshold θ is set to filter out insignificant gene pairs. Gene pairs (i, j) with mutual information values ​​greater than the threshold θ are retained. To ensure the statistical significance of the results, a statistical test, such as a hypothesis test or a permutation test, is performed on each mutual information value MI(i, j), and significant gene pairs are retained.

[0048] 4) Signal network construction

[0049] Each gene is taken as a node in the network, and the selected gene pairs are retained as edges. The weight of the edge is set to the corresponding mutual information value MI(i,j). Finally, a signal interaction network G based on mutual information is generated, which reflects the regulatory relationship between genes (such as Figure 3 Here, two signal interaction networks were obtained: a signal interaction network based on drug-resistant samples and a signal interaction network based on drug-intolerant samples.

[0050] 3. Network Convergence

[0051] In this step, based on the signal interaction network of drug-resistant and non-resistant samples, network fusion is performed and a differential gene network is constructed. The specific steps are as follows:

[0052] 1) Constructing a separate co-expression network

[0053] According to the above steps, the signal interaction networks G are constructed for resistant and non-resistant samples respectively. res and G non-res In both networks, nodes represent genes and edges represent the mutual information between gene pairs.

[0054] 2) Calculation of differentially expressed genes

[0055] Compare the corresponding edges in the two networks and calculate the difference in mutual information between each pair of genes in resistant and non-resistant samples:

[0056] ΔMI(i,j)=MI res (i,j)-MI non-res (i,j)

[0057] If ΔMI(i, j) deviates significantly from zero, it indicates that there is a significant difference in the regulatory relationship between the gene pair in resistant and non-resistant samples.

[0058] 3) Construction of differentially expressed gene network

[0059] The above gene pairs with significant differences were screened out to construct the differential gene signal interaction network G diff , where the weight of the edge is the mutual information difference ΔMI(i, j). This difference network reveals the changes in gene regulatory relationships under different drug sensitivity backgrounds. Figure 4 shown.

[0060] 4. Identification of key regulatory factors

[0061] Betweenness centrality and eigenvector centrality analysis methods were used to identify genes with the greatest differences in the fused network.

[0062] The calculation method is as follows:

[0063] 1) Betweenness Centrality

[0064] Among them, σ st is the number of shortest paths from node s to node t, σ st (v) is the number of shortest paths passing through node v.

[0065] The larger the betweenness centrality value, the more important the role of the gene in the network. This example selects the top 10 genes with the betweenness centrality values.

[0066] 2) Eigenvector Centrality

[0067] Among them, λ is the eigenvalue, CE is the eigenvector.

[0068] The size of the eigenvector centrality value also reflects the importance of a gene in the network. This embodiment selects the top 10 genes with the eigenvector centrality values.

[0069] The genes obtained by the above two methods were merged and deduplicated to obtain the key regulatory factors of dasatinib resistance, including LCK, TGFBR1, TEC, PDGFRB, PAG1, IL4, CD28, PTCRA, PDCD1, TNK2, CD3G, MYLK, and SYK.

[0070] 5. Construction of drug sensitivity prediction model

[0071] A prediction model was constructed using a multi-layer perceptron algorithm, key regulatory factors were selected as features, the model was trained, and the performance was evaluated.

[0072] The specific process is as follows:

[0073] 1) Select the expression levels of key regulatory factors from the expression matrix to form a new expression matrix, of which 70% of the patient data is used as the training set, 15% of the patient data is used as the validation set, and 15% of the patient data is used as the test set;

[0074] 2) Model initialization: Set the hidden layer to 100 neurons, the maximum number of optimizer iterations to 10,000, and the activation function to ReLU. The ReLU formula is f(x) = max(0, x), which sets all negative input values ​​to 0 and keeps the positive input values ​​unchanged.

[0075] 3) Model training

[0076] a. Forward propagation: The model processes the output of each layer through the weighted sum of each neuron and the activation function (ReLU) through the input data, and propagates forward in the network. The output of each layer becomes the input of the next layer. The data reaches the output layer and generates a prediction result.

[0077] b. Loss calculation: Use the loss function to calculate the difference between the model’s output and the true label.

[0078] c. Backpropagation: Calculate the gradient of the loss with respect to the model parameters (weights and biases) using an optimization algorithm (Stochastic Gradient Descent SGD) and use these gradients to update the parameters.

[0079] d. Iteration: Repeat the forward propagation and backpropagation process until the specified number of iterations is completed, or the loss no longer decreases significantly, at which point the model has converged.

[0080] 4) Model evaluation: Use the reserved test set to evaluate the performance of the model. This example uses cross-validation. The test set is divided into 5 parts. During the cross-validation process, each part is used as the test set in turn, and the rest is used as the training set. Therefore, the model will be trained and evaluated 5 times, each time with a different training set and test set. The model performance evaluation results are shown in Figure 1. Figure 5 shown.

[0081] After training, we derived the importance of each gene for model prediction, and the results are shown in Table 1.

[0082] Table 1: Final weight values ​​of genes used in the model

[0083]

[0084]

[0085] In Table 1, a positive weight value indicates that the gene has a positive effect on drug sensitivity prediction, while a negative weight value indicates that the gene has a negative effect on drug sensitivity prediction. The larger the absolute value of the weight, the greater the impact of the gene on the model.

[0086] 6. Verification of the model’s accuracy in predicting dasatinib resistance

[0087] 1) Collect clinical samples and perform high-throughput transcriptome sequencing (RNA-seq).

[0088] 2) Data preprocessing: remove low-quality data, align to the reference genome, quantify gene expression, and generate a gene expression matrix; the results are shown in Table 2.

[0089] Table 2: Gene expression matrix of clinical samples

[0090]

[0091]

[0092] 3) Input the gene expression matrix data of clinical samples into the dasatinib resistance prediction model;

[0093] 4) Obtain the drug resistance prediction results of each sample.

[0094] The sample cells described in Table 2 were subjected to a dasatinib drug sensitivity test, and the results are shown in Table 3. The dose of dasatinib required for 50% cell death was counted, and the dose greater than 100 nM was a drug-sensitive sample, and the dose less than 100 nM was a drug-insensitive sample.

[0095] Table 3:

[0096]

[0097] After the above steps, the final dasatinib drug sensitivity prediction model is obtained, which can be used to predict dasatinib drug sensitivity based on the patient's RNA-seq gene expression data as input. And the prediction model has been verified to have high accuracy. Therefore, the treatment strategy can be adjusted according to the prediction effect to improve the treatment effect and reduce unnecessary side effects.

[0098] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. Application of a combination of gene markers in the preparation of a product for predicting drug sensitivity of dasatinib; the combination of gene markers includes LCK, TGFBR1, TEC, PDGFRB, PAG1, IL4, CD28, PTCRA, PDCD1, TNK2, CD3G, MYLK, and SYK.

2. Dasatinib drug sensitivity test kit, It is characterized in that It includes detection reagents for gene markers, and the gene markers include LCK, TGFBR1, TEC, PDGFRB, PAG1, IL4, CD28, PTCRA, PDCD1, TNK2, CD3G, MYLK, and SYK.

3. The detection kit according to claim 2, It is characterized in that The kit is selected from a transcriptome sequencing library construction kit and a PCR amplification kit.

4. Dasatinib drug sensitivity prediction system, It is characterized in that include: A gene expression quantity acquisition unit is used to obtain the expression quantity of the target gene of the sample to be tested; Target genes include LCK, TGFBR1, TEC, PDGFRB, PAG1, IL4, CD28, PTCRA, PDCD1, TNK2, CD3G, MYLK, and SYK; The drug sensitivity prediction unit inputs the expression level of each target gene into a drug sensitivity prediction model to obtain a prediction result; the drug sensitivity prediction model uses a multi-layer perceptron algorithm, selects the target gene as a feature and is trained based on a patient transcriptome data sample.

5. The prediction system according to claim 4, It is characterized in that The target gene was screened by the following method: Constructing a gene expression matrix based on the transcriptome data sample library, including a first matrix based on drug-resistant samples and a second matrix based on drug-inresistant samples; The significant gene pairs are selected according to the mutual information values ​​of each group of genes in the first matrix and the second matrix, and the first signal interaction network and the second signal interaction network are constructed respectively with each gene as a node; The differential gene network was constructed by network fusion based on the first signal interaction network and the second signal interaction network, and genes with significant differences were screened as target genes.

6. The prediction system according to claim 5, It is characterized in that The differential gene network analyzes and identifies genes with significant differences based on the betweenness centrality value and the eigenvector centrality value, respectively, and the two groups of differential genes obtained are merged and deduplicated to obtain the target gene.

7. The prediction system according to claim 4, It is characterized in that The drug sensitivity prediction model is trained by taking the target gene expression amount in the sample to generate a target gene expression matrix, and dividing the target gene expression matrix into a training set, a validation set and a test set.

8. The prediction system according to claim 4, It is characterized in that The expression level of the target gene in the sample to be tested is obtained by performing transcriptome sequencing or single-cell sequencing and quantitative analysis on a clinical sample of a patient, and the clinical sample is a blood sample.

9. The prediction system according to claim 8, It is characterized in that The quantitative analysis process of the target gene expression is to remove low-quality data in the sequencing data, align it to the reference genome, and quantify the gene expression.

10. The prediction system according to claim 4, It is characterized in that The drug sensitivity prediction unit is used to predict the dose of dasatinib corresponding to the cell death rate, and to determine whether the sample to be tested is sensitive to dasatinib according to the sensitivity standard.

Images