Treatment response prediction method, device, equipment and medium for combined treatment of acute myelogenous leukemia based on VEN / AZA

Abstract

The invention discloses a treatment response prediction method, device and equipment for combined treatment of acute myelogenous leukemia based on VEN / AZA and a medium, and relates to the technical field of prognosis evaluation of treatment, and the method comprises the following steps: obtaining target transcriptome data of a to-be-predicted acute myelogenous leukemia patient; the target transcriptome data is expression quantity information of a preset number of specific genes on a transcriptional level, wherein the expression quantity information is obtained after transcriptome sequencing is carried out on the acute myelogenous leukemia patient to be predicted; inputting the target transcriptome data into a preset treatment response prediction model, so as to calculate a response probability value generated after VEN / AZA combined treatment of the to-be-predicted acute myelogenous leukemia patient through the preset treatment response prediction model and based on the target transcriptome data, thereby obtaining a corresponding target response probability; and outputting a treatment reaction result corresponding to the VEN / AZA combined treatment scheme of the to-be-predicted acute myelogenous leukemia patient based on the target response probability.

Description

Technical Field

[0001] The present invention relates to the technical field of prognosis evaluation of treatment, and in particular to a method, device, equipment and medium for predicting treatment response based on VEN / AZA combined treatment of acute myeloid leukemia. Background Art

[0002] Acute myeloid leukemia (AML) is a highly heterogeneous and aggressive hematological malignancy. The combination of Venetoclax (a selective BCL-2 inhibitor) and Azacitidine (a demethylating drug) (VEN / AZA regimen) has become the standard treatment for newly diagnosed AML patients who are not suitable for high-intensity chemotherapy. Although this treatment regimen has good efficacy and tolerability, many patients still cannot obtain long-term clinical benefits due to the biological heterogeneity and complex clinical manifestations of AML. Existing risk stratification systems, such as the European LeukemiaNet (ELN) risk stratification system, are mainly developed based on young AML patients receiving high-intensity induction chemotherapy, which limits its application in elderly AML patients, especially those receiving Venetoclax-based treatment. To fill this gap, several prognostic models for Venetoclax therapy have been proposed internationally in recent years. For example, a four-gene model including TP53, FLT3-ITD, NRAS and KRAS mutations, which divides AML patients into high-risk (TP53 mutation), intermediate-risk (FLT3-ITD, NRAS and KRAS mutations) and low-risk (other mutations) risk groups, is used to predict the prognosis of patients treated with Venetoclax + Azacitidine. Similarly, a prognostic model based on gene mutations divides patients into high-risk (TP53, FLT3-ITD, NRAS and KRAS mutations), intermediate-risk (other mutations) and low-risk (NPM1 and IDH2 mutations) risk groups; patients with TP53, FLT3-ITD and RUNX1 mutations are classified as high-risk, patients with NPM1, IDH2 and DDX41 mutations are classified as low-risk, and patients with other mutations are classified as intermediate-risk. Although these models can predict patient outcomes to a certain extent, the latest research points out that when these models are applied to real-world AML patients treated with venetoclax with azacitidine or low-dose cytarabine, the predictive effect is still poor. Summary of the invention

[0003] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for predicting the therapeutic response of VEN / AZA combined treatment for acute myeloid leukemia, which can accurately predict the therapeutic response of acute myeloid leukemia patients of all ages before VEN / AZA combined treatment is applied. The specific scheme is as follows:

[0004] In a first aspect, the present application discloses a method for predicting treatment response of acute myeloid leukemia based on VEN / AZA combined treatment, which is applied to a computer device and comprises:

[0005] Obtaining target transcriptome data of the patient to be predicted for acute myeloid leukemia; the target transcriptome data is the expression amount information of a preset number of specific genes at the transcriptional level obtained after transcriptome sequencing of the patient to be predicted for acute myeloid leukemia; the specific genes are genes that have a sensitive reaction or a drug resistance reaction with VEN drug or AZA drug;

[0006] Inputting the target transcriptome data into a preset treatment response prediction model, so as to calculate the response probability value of the acute myeloid leukemia patient to be predicted after VEN / AZA combined treatment based on the preset treatment response prediction model and the target transcriptome data, so as to obtain a corresponding target response probability;

[0007] The treatment response result corresponding to the VEN / AZA combined treatment regimen of the acute myeloid leukemia patient to be predicted is output based on the target response probability.

[0008] Optionally, after obtaining the target transcriptome data of the patient with acute myeloid leukemia to be predicted, the method further includes:

[0009] Performing log2 normalization on the target transcriptome data to obtain normalized transcriptome data;

[0010] Performing feature extraction processing on the standardized transcriptome data to obtain target data features including gene expression patterns and relationships between relative expression levels;

[0011] Accordingly, inputting the target transcriptome data into a preset treatment response prediction model includes:

[0012] The target data features are input into a preset treatment response prediction model.

[0013] Optionally, before inputting the target transcriptome data into a preset treatment response prediction model, the method further includes:

[0014] The gene information that is resistant or sensitive to VEN or AZA drugs is screened from the transcriptome data of patients with acute myeloid leukemia who received VEN / AZA combination therapy to obtain the corresponding transcriptome data training set;

[0015] The initial treatment response prediction model is trained according to the in vitro drug response data training set containing patient grouping data carrying drug resistance labels or sensitive labels and the transcriptome data training set, so that the initial treatment response prediction model learns the association between gene information, in vitro drug response data and drug resistance labels or sensitive labels that characterize the response to treatment, and updates the model parameters to obtain a preset treatment response prediction model.

[0016] Optionally, the method uses the transcriptome data of patients receiving VEN / AZA combined treatment for acute myeloid leukemia to screen out gene information that has drug resistance or sensitivity to VEN drugs or AZA drugs to obtain a corresponding transcriptome data training set, including:

[0017] The transcriptome data of sensitive and resistant patients with acute myeloid leukemia who received VEN / AZA combination therapy were compared, and the gene expression information associated with the efficacy of VEN or AZA drugs was screened out through a preset data analysis method to obtain the corresponding transcriptome data training set.

[0018] Optionally, before training the initial treatment response prediction model based on the in vitro drug response data training set containing patient grouping data carrying drug resistance labels or sensitive labels and the transcriptome data training set, the method further includes:

[0019] Unsupervised clustering was performed on patients with acute myeloid leukemia who received VEN / AZA combination therapy to obtain patient grouping data under corresponding drug resistance labels and patient grouping data under sensitive labels, and an in vitro drug response data training set was constructed based on the patient grouping data under the drug resistance label and the patient grouping data under the sensitive label.

[0020] Optionally, before obtaining the preset treatment response prediction model, the method further includes:

[0021] A five-fold cross-validation was performed on the process treatment response prediction model with updated model parameters to select the process treatment response prediction model with the best performance in the validation process as the preset treatment response prediction model.

[0022] Optionally, the outputting of the treatment response result corresponding to the VEN / AZA combined treatment regimen for the acute myeloid leukemia patient to be predicted based on the target response probability includes:

[0023] If the target response probability is greater than a preset response probability threshold, then determining that the treatment response effect of the VEN / AZA combined treatment regimen of the acute myeloid leukemia patient to be predicted corresponding to the target response probability is a sensitive response result;

[0024] If the target response probability is less than or equal to a preset response probability threshold, it is determined that the therapeutic response effect of the VEN / AZA combined treatment regimen for the acute myeloid leukemia patient to be predicted corresponding to the target response probability is a drug resistance reaction result.

[0025] In a second aspect, the present application discloses a therapeutic response prediction device based on VEN / AZA combined treatment of acute myeloid leukemia, comprising:

[0026] A data acquisition module is used to acquire target transcriptome data of a patient with acute myeloid leukemia to be predicted; the target transcriptome data is the expression amount information of a preset number of specific genes at the transcriptional level obtained after transcriptome sequencing of the patient with acute myeloid leukemia to be predicted; the specific genes are genes that have sensitive reactions or drug resistance reactions with VEN drugs or AZA drugs;

[0027] A probability prediction module, used for inputting the target transcriptome data into a preset treatment response prediction model, so as to calculate the response probability value of the acute myeloid leukemia patient to be predicted after VEN / AZA combined treatment based on the preset treatment response prediction model and the target transcriptome data, so as to obtain a corresponding target response probability;

[0028] A result feedback module is used to output the treatment response result corresponding to the VEN / AZA combined treatment regimen of the acute myeloid leukemia patient to be predicted based on the target response probability.

[0029] In a third aspect, the present application discloses an electronic device, comprising:

[0030] Memory, used to store computer programs;

[0031] A processor is used to execute the computer program to implement the steps of the aforementioned disclosed method for predicting the therapeutic response of acute myeloid leukemia based on the VEN / AZA combination therapy.

[0032] In a fourth aspect, the present application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the steps of the aforementioned disclosed method for predicting treatment response of acute myeloid leukemia based on VEN / AZA combined treatment are implemented.

[0033] It can be seen that the present application discloses a method for predicting the treatment response of acute myeloid leukemia based on VEN / AZA combined treatment, which is applied to a computer device, including: obtaining target transcriptome data of a patient with acute myeloid leukemia to be predicted; the target transcriptome data is the expression amount information of a preset number of specific genes at the transcriptional level obtained after transcriptome sequencing of the patient with acute myeloid leukemia to be predicted; the specific gene is a gene that has a sensitive reaction or a drug resistance reaction with VEN drugs or AZA drugs; the target transcriptome data is input into a preset treatment response prediction model, so as to calculate the response probability value generated after VEN / AZA combined treatment of the patient with acute myeloid leukemia to be predicted based on the preset treatment response prediction model and the target transcriptome data, so as to obtain the corresponding target response probability; based on the target response probability, the treatment response result corresponding to the VEN / AZA combined treatment scheme of the patient with acute myeloid leukemia to be predicted is output. It can be seen that by analyzing the transcriptome data of specific genes that have a sensitive reaction or a drug resistance reaction with VEN drugs or AZA drugs, the model can accurately calculate the patient's response probability value to VEN / AZA combined treatment. A more accurate estimate of the treatment effect on the patient can be made before treatment, avoiding blind selection of treatment plans and improving the scientificity and accuracy of treatment decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0035] Figure 1 A flow chart of a method for predicting treatment response based on VEN / AZA combined treatment of acute myeloid leukemia disclosed in the present application;

[0036] Figure 2 A flow chart of a method for training and validating a preset treatment response prediction model disclosed in this application;

[0037] Figure 3 This is a schematic diagram of the structure of a treatment response prediction device based on VEN / AZA combined treatment of acute myeloid leukemia disclosed in this application;

[0038] Figure 4 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0040] Acute myeloid leukemia is a highly heterogeneous and aggressive hematological malignancy. The combination of Venetoclax and Azacitidine (VEN / AZA regimen) has become the standard treatment for newly diagnosed AML patients who are not suitable for high-intensity chemotherapy. Although this treatment regimen has good efficacy and tolerability, many patients still cannot obtain long-term clinical benefits due to the biological heterogeneity and complex clinical manifestations of AML. Existing risk stratification systems, such as the European Leukemia Network Risk Stratification System, are mainly developed based on young AML patients receiving high-intensity induction chemotherapy, which limits its application in elderly AML patients, especially those receiving Venetoclax-based treatment. To fill this gap, several prognostic models for Venetoclax therapy have been proposed internationally in recent years. For example, a four-gene model containing TP53, FLT3-ITD, NRAS and KRAS mutations, which divides AML patients into high-risk, intermediate-risk and low-risk groups, is used to predict the prognosis of patients receiving Venetoclax + Azacitidine treatment. Similarly, a prognostic model based on gene mutations divides patients into high-risk, intermediate-risk and low-risk groups; patients with TP53, FLT3-ITD and RUNX1 mutations are classified as high-risk, patients with NPM1, IDH2 and DDX41 mutations are classified as low-risk, and patients with other mutations are classified as intermediate-risk. Although these models can predict the prognosis of patients to a certain extent, the latest research points out that when these models are applied to real-world AML patients treated with Venetoclax and Azacitidine or low-dose cytarabine, the predictive effect is still poor.

[0041] To this end, the present invention provides a treatment response prediction scheme based on VEN / AZA combined treatment of acute myeloid leukemia, which can accurately predict the treatment response after VEN / AZA combined treatment for acute myeloid leukemia patients of all ages before applying VEN / AZA combined treatment.

[0042] Reference Figure 1 As shown, the embodiment of the present invention discloses a method for predicting treatment response of acute myeloid leukemia based on VEN / AZA combined treatment, which is applied to a computer device and includes:

[0043] Step S11: Obtain target transcriptome data of the patient to be predicted for acute myeloid leukemia; the target transcriptome data is the expression level information of a preset number of specific genes at the transcriptional level obtained after transcriptome sequencing of the patient to be predicted for acute myeloid leukemia; the specific genes are genes that have sensitive reactions or resistance reactions to VEN drugs or AZA drugs.

[0044] In this embodiment, in order to accurately predict whether an acute myeloid leukemia patient is suitable for the VEN / AZA combination regimen, the patient is used as an acute myeloid leukemia patient to be predicted, and the gene expression values ​​of 8 specific genes of the acute myeloid leukemia patient to be predicted are obtained as target transcriptome data, wherein the 8 specific genes include: PDZK1IP1, AFAP1, MANEAL, SASH1, ROGDI, CEBPE, SLC9A3, PINK1. It should be noted that these 8 specific genes are strictly screened and verified, and they are closely related to the patient's response to VEN / AZA combined treatment in the transcriptome data. During the training process, the expression levels of these genes showed obvious differences in sensitive AML and resistant AML, and were significantly correlated with the patient's clinical treatment response (such as achieving complete remission CR, complete remission CRi with incomplete hematological recovery, etc.). Therefore, they can be used as effective indicators to predict the patient's response to treatment. These 8 genes are involved in a variety of biological processes related to VEN / AZA combined treatment, such as cell differentiation, immune response, metabolic pathways, etc. Changes in their expression levels can comprehensively reflect the mechanism of action and therapeutic effect of the drug in the patient.

[0045] In this embodiment, after obtaining the target transcriptome data of the patient with acute myeloid leukemia to be predicted, it also includes: performing log2 normalization on the target transcriptome data to obtain standardized transcriptome data; performing feature extraction on the standardized transcriptome data to obtain target data features including the pattern of gene expression and the relationship between relative expression levels. It can be understood that the computer device of the present invention can be a Venetoclax-Azacitidine efficacy prediction system developed using the Shiny R package, which is convenient for doctors to operate. Therefore, after the Venetoclax-Azacitidine efficacy prediction system receives the target transcriptome data, it performs normalization on the target transcriptome data, that is, performs log2 normalization on the gene expression value. Specifically, the gene expression values ​​of 8 specific genes are usually expressed in the form of raw counts, reads per kilobase per million mapped reads (RPKM), fragments per kilobase per million mapped reads (FPKM), Transcripts per kilobase of exon model per million mapped reads (TPM), etc., which are the expression amount information of specific genes at the transcription level. Therefore, there is a problem that the original gene expression values ​​have a large range of differences, so the range of gene expression values ​​needs to be compressed and adjusted to make it more in line with the input requirements of the model. After log2 conversion, the data can be more concentrated in a relatively small range, which is convenient for the model to process and analyze. Moreover, in some cases, gene expression values ​​may show a skewed distribution, and log2 conversion can make them closer to a normal distribution, which helps to improve the performance and stability of the model. Finally, the gene expression values ​​after log2 conversion can better reflect the changes in the relative expression of genes and can more intuitively reflect the degree of change in gene expression.

[0046] Specifically, the log2 standard process is: perform logarithmic transformation calculation on the gene expression value, where the calculation formula is to calculate the log2 value: For the expression value of each specific gene, use the logarithmic transformation formula to calculate, the calculation formula is log2(expression value+1). Adding 1 here is to avoid the situation where the logarithmic calculation is meaningless when the gene expression value is 0. For example, if the expression value of a gene is 100, then the value after log2 transformation is log2(100+1)=log2(101)≈6.658.

[0047] Step S12: Input the target transcriptome data into a preset treatment response prediction model so that the response probability value of the acute myeloid leukemia patient to be predicted after VEN / AZA combined treatment is calculated based on the preset treatment response prediction model and the target transcriptome data to obtain a corresponding target response probability.

[0048] In this embodiment, the target data features are input into a preset treatment response prediction model. It is understandable that the Venetoclax-Azacitidine efficacy prediction system calls the preset treatment response prediction model to predict the effect information after combined treatment based on the target data features obtained by the standardized processing. Specifically, the response probability value of the target data features is predicted and calculated by the preset treatment response prediction model to output the corresponding target response probability.

[0049] Specifically, the preset treatment response prediction model is the trained RF8 model, which is a classifier based on machine learning algorithms such as random forests. It consists of multiple decision trees, each of which learns different features and decision rules during the training process. When the target data features are input, the model automatically extracts key features from these target data features. Key features include combinations of gene expression values, relative size relationships, and change trends. For example, a combination of high expression of certain genes and low expression of other genes may be associated with a specific treatment response. In the random forest, each decision tree predicts the patient's treatment response based on the extracted features. The decision tree makes decisions based on the gene expression features step by step according to its internal node and branch structure, and finally gives a prediction result of whether the patient is a responder or a non-responder. Since the RF8 model consists of multiple decision trees, the final prediction result is obtained by integrating the predictions of all decision trees. Voting is usually used, that is, the prediction results of the majority of decision trees are counted as the final prediction. If most decision trees predict that the patient is a responder, the RF8 model will output a higher response probability value; conversely, if most decision trees predict that the patient is a non-responder, the response probability value will be lower. The RF8 (Random Forest 8 model) model outputs a response probability value between 0 and 1. This probability value reflects the likelihood that the patient will respond to Venetoclax-Azacitidine treatment. For example, a probability value of 0.8 means that the patient has an 80% chance of responding well to the treatment.

[0050] Step S13: outputting the treatment response result corresponding to the VEN / AZA combined treatment regimen for the acute myeloid leukemia patient to be predicted based on the target response probability.

[0051] In this embodiment, if the target response probability is greater than the preset response probability threshold, the therapeutic response effect of the VEN / AZA combined treatment regimen of the acute myeloid leukemia patient to be predicted corresponding to the target response probability is determined to be a sensitive reaction result; if the target response probability is less than or equal to the preset response probability threshold, the therapeutic response effect of the VEN / AZA combined treatment regimen of the acute myeloid leukemia patient to be predicted corresponding to the target response probability is determined to be a drug resistance reaction result. It can be understood that the preset response probability threshold is first set according to the actual application scenario. Due to the high expectation for the treatment effect, the preset response probability threshold is set to 0.5 in the present invention. In this way, when the target response probability is greater than 0.5, it is determined that if the treatment effect of the VEN / AZA combined treatment regimen is a sensitive reaction result, the combined treatment regimen is given priority. When the target response probability is less than or equal to 0.5, it is determined that if the treatment effect of the VEN / AZA combined treatment regimen is a drug resistance reaction result, other treatment regimens are given priority.

[0052] It can be seen that the present application discloses a method for predicting the treatment response of acute myeloid leukemia based on VEN / AZA combined treatment, which is applied to a computer device, including: obtaining target transcriptome data of a patient with acute myeloid leukemia to be predicted; the target transcriptome data is the expression amount information of a preset number of specific genes at the transcriptional level obtained after transcriptome sequencing of the patient with acute myeloid leukemia to be predicted; the specific gene is a gene that has a sensitive reaction or a drug resistance reaction with VEN drugs or AZA drugs; the target transcriptome data is input into a preset treatment response prediction model, so as to calculate the response probability value generated after VEN / AZA combined treatment of the patient with acute myeloid leukemia to be predicted based on the preset treatment response prediction model and the target transcriptome data, so as to obtain the corresponding target response probability; based on the target response probability, the treatment response result corresponding to the VEN / AZA combined treatment scheme of the patient with acute myeloid leukemia to be predicted is output. It can be seen that by analyzing the transcriptome data of specific genes that have a sensitive reaction or a drug resistance reaction with VEN drugs or AZA drugs, the model can accurately calculate the patient's response probability value to VEN / AZA combined treatment. A more accurate estimate of the treatment effect on the patient can be made before treatment, avoiding blind selection of treatment plans and improving the scientificity and accuracy of treatment decisions.

[0053] Reference Figure 2 As shown, the embodiment of the present invention discloses a specific method for training and verifying a preset treatment response prediction model. Compared with the previous embodiment, this embodiment further illustrates and optimizes the technical solution. Specifically:

[0054] Step S21: Gene information showing drug resistance or sensitivity to VEN or AZA drugs is screened out from transcriptome data of patients receiving VEN / AZA combined treatment for acute myeloid leukemia to obtain a corresponding transcriptome data training set.

[0055] In this embodiment, the transcriptome data of sensitive patients and resistant patients receiving VEN / AZA combined treatment for acute myeloid leukemia are compared, and the gene expression information associated with the efficacy of VEN drugs or AZA drugs is screened out by a preset data analysis method to obtain a corresponding transcriptome data training set. It can be understood that in order to systematically identify biomarkers associated with Venetoclax and Azacitidine response, the in vitro pharmacological profile data (in vitro drug response data) and RNA sequencing data (transcriptome data) of AML patients receiving Venetoclax-Azacitidine combined treatment in the FPMTB data set are first integrated.

[0056] Among them, CIBERSORTx software was used to perform deconvolution analysis on the transcriptome data of AML patients. Sensitive AML was enriched with "Primed LSPC", while drug-resistant AML was enriched with "Mono-like blasts". In this way, gene expression information associated with the efficacy of VEN or AZA drugs was screened out through data analysis, and the screened gene expression information was integrated to obtain the corresponding transcriptome data training set.

[0057] Step S22: Train the initial treatment response prediction model based on the in vitro drug response data training set containing patient grouping data carrying drug resistance labels or sensitive labels and the transcriptome data training set, so that the initial treatment response prediction model learns the relationship between gene information, in vitro drug response data and drug resistance labels or sensitive labels that characterize the treatment response, and updates the model parameters to obtain a preset treatment response prediction model.

[0058] In this embodiment, before training the initial treatment response prediction model according to the in vitro drug response data training set containing patient grouping data carrying a drug resistance label or a sensitive label and the transcriptome data training set, it also includes: performing unsupervised clustering on patients receiving VEN / AZA combined treatment for acute myeloid leukemia to obtain the corresponding patient grouping data under the drug resistance label and the patient grouping data under the sensitive label, and constructing an in vitro drug response data training set according to the patient grouping data under the drug resistance label and the patient grouping data under the sensitive label. It can be understood that the drug sensitivity data of Venetoclax and Azacitidine can be unsupervised clustered using the pheatmap R package, which can divide the patients into two major groups, each showing different sensitivity characteristics. Among them, the patients in Cluster-1 showed resistance to both Venetoclax and Azacitidine, and were therefore classified as resistant AML (Resistant-AMLs). In contrast, patients in Cluster-2 and Cluster-3 showed significant sensitivity to Venetoclax and Azacitidine, respectively, and were therefore classified as sensitive AML (Sensitive-AMLs). By comparing RNA-seq (transcriptome data) of sensitive AML with resistant AML, 1626 differentially expressed genes (DEGs) were identified using the DESeq2 R package. Among these differentially expressed genes, MSI2 (a primitive hematopoietic stem cell marker) was significantly upregulated in sensitive AML, while monocytic differentiation markers (such as CD14, MAFB, and S100A9) were significantly downregulated. Gene set enrichment analysis further revealed that sensitive AML was enriched in leukemic stem cell features, while resistant AML was mainly enriched in monocytic differentiation pathways, indicating that monocytic differentiation is associated with primary resistance to Venetoclax-Azacitidine therapy, while primitive blasts tend to be more sensitive to this treatment regimen.

[0059] It should be noted that the above process of selecting and constructing the data training set emphasized the potential role of cell differentiation status in predicting the response to VEN / AZA treatment. In addition, using the Metascape online database, functional enrichment analysis of differentially expressed genes showed that these genes were associated with innate immune response pathways, lysosomal function, and metabolic pathways, suggesting that they may play a role in regulating the efficacy of Venetoclax-Azacitidine. Overall, these data provide rich predictive biomarkers associated with the efficacy of Venetoclax-Azacitidine in AML patients. In order to further translate the correlation between the expression of predictive biomarkers and the individual response of patients to Venetoclax-Azacitidine therapy into clinical practice, a cohort of 110 newly diagnosed AML patients treated with Venetoclax-Azacitidine was first collected, and high-throughput targeted sequencing and RNA sequencing were performed (RJAML cohort 1). Of these patients, 68 (61.8%) achieved complete remission (CR), and another 9 (8.2%) achieved complete remission with incomplete hematologic recovery (CRi), with an overall CR / CRi rate of 70% (responders). 2.7% of patients achieved morphologic leukemia-free status, while 7.3% achieved partial remission (PR), and 20% had refractory disease (non-responders). As expected, survival analysis using the survival R package found that responders (i.e., patients who achieved CR / CRi) showed significantly improved event-free survival (EFS) and overall survival (OS) compared with non-responders, further emphasizing the key role of achieving CR / CRi in predicting long-term survival outcomes.

[0060] By analyzing RNA-seq data (transcriptome data) stratified by the in vivo response of patients to Venetoclax-Azacitidine treatment, 1,116 differentially expressed genes (DEGs) were discovered using the DESeq2 R package. Intersecting with the 1,626 DEGs obtained in the previous step, 70 genes were found to be consistently downregulated in responders (sensitive AML), which we labeled as Venetoclax-Azacitidine resistance genes. In contrast, 14 genes were consistently upregulated in responders and were labeled as Venetoclax-Azacitidine sensitive genes. To further validate the role of these genes in regulating Venetoclax / Azacitidine sensitivity, CRISPR-Cas9 screening data for Venetoclax and Azacitidine were utilized. Using the resistance gene signature, GSEA analysis based on CRISPR knockout scores showed that the resistance signature was negatively enriched in both drug screens, indicating that knocking out these genes can enhance drug sensitivity. In general, these results indicate that the sensitivity of AML patients to Venetoclax-Azacitidine therapy can be regulated by changing the expression of Venetoclax-Azacitidine response genes. And 84 high-confidence Venetoclax-Azacitidine response genes were screened out. In this way, transcriptome data and in vitro drug screening data are integrated. Transcriptome data provides rich information on gene expression, which can reflect the molecular state and function of cells; in vitro drug screening data directly reflects the sensitivity of patients to drugs. By integrating these two data types, the RF8 model can more comprehensively capture features related to treatment effects, thereby improving prediction accuracy.

[0061] In this embodiment, before obtaining the preset treatment response prediction model, it also includes: performing a five-fold cross-validation on the process treatment response prediction model with updated model parameters to select the process treatment response prediction model with the best performance in the verification process as the preset treatment response prediction model. It can be understood that in order to build a robust prediction model based on 84 Venetoclax-Azacitidine response gene expression profiles, in RJAML cohort 1, the binary variables of patients (responders vs. non-responders) were analyzed, and the prediction performance of ten machine learning algorithms was evaluated using the mlr3 R package system. Each model was subjected to 1,00 five-fold cross-validation, which is equivalent to a total of 5,00 random training-validation groupings, ensuring unbiased and reliable performance estimates. The best performing model was selected for final training and testing on the validation data set. Based on the area under the curve (AUC) and accuracy (ACC) evaluation, the random forest (RF) model showed the best performance among all classifiers. Therefore, RF was selected as the final model, and the Boruta feature selection algorithm was used to identify the most relevant feature subset for predicting Venetoclax-Azacitidine responders. Finally, eight key features were identified, and an integrated RF classifier, called RF8, was developed based on 70% of the training set. Using this model, the contribution of each feature to the difference in response between patients was quantified. At the individual level, the RF8 model can be used to score each patient based on the probability of achieving CR / CRi, with higher scores indicating a greater likelihood of response. In this way, the Boruta feature selection algorithm was used to identify the most relevant feature subsets for predicting Venetoclax-Azacitidine responders, and finally the relevant key features were determined. These key features are selected from a large number of data features. They contain the information most closely related to the treatment response and can effectively improve the predictive accuracy of the model. In addition, the RF8 model uses advanced machine learning algorithms such as random forests, which have strong learning and generalization capabilities, and can learn complex patterns and laws from a large amount of data, thereby improving the accuracy of predictions.

[0062] The RF8 model demonstrated excellent predictive performance, with excellent performance in both the test set (AUC=0.85) and the entire dataset (AUC=0.98). In addition, the continuous probability values ​​generated by the RF8 model were significantly correlated with the patient's response, and patients who achieved CR / CRi had significantly higher RF8 scores. To classify patients as responders or non-responders, a default probability threshold of 0.5 was applied, and patients greater than this threshold were predicted to be responders. Using the binary score, RF8 predicted that the odds ratios for Venetoclax-Azacitidine response in patients with high RF8 (responders) versus low RF8 (non-responders) were 6.7 and 128.5 in the test set and the entire dataset, respectively.

[0063] To validate the performance of RF8, RNA-seq data of 32 diagnostic samples from an independent cohort (RJAML cohort 2) were analyzed. In this data, the RF8 model had an AUC of 0.83, demonstrating its strong predictive ability for clinical response to Venetoclax-Azacitidine. In addition, the RF8 model effectively distinguished resistant from sensitive patients, and higher prediction probabilities were associated with increased sensitivity to Venetoclax / Azacitidine based on evaluation of in vitro drug screening in the FPMTB dataset. For further validation, multi-omics data from the BeatAML cohort were integrated. The RF8 model was applied to calculate the response probability score for each patient, and the results showed that the RF8 prediction model performed well in distinguishing sensitive from resistant patients, with an AUC value of 0.76. In summary, these results indicate that the RF8 gene expression model is able to capture the core transcriptome programs associated with predicting Venetoclax-Azacitidine response and has significant predictive performance. In this way, the RF8 model was validated in multiple patient cohorts, including RJAML Cohort 1 and RJAML Cohort 2, as well as integrating multi-omics data from the BeatAML cohort. These validation processes demonstrated the versatility and accuracy of the model in different patient populations and clinical settings, and its wide applicability. The RF8 model is able to make personalized predictions based on the individual characteristics of the patient, providing each patient with a unique response probability value. This personalized prediction capability enables the model to be applicable to different types of AML patients, and can provide doctors with valuable predictive information regardless of factors such as the patient's age, gender, and severity of the disease.

[0064] Next, we explored the relationship between RF8 scores and clinical outcomes in patients treated with Venetoclax-Azacitidine. Notably, the RF8 features showed a clear trend: as the RF8 score increased, the probability of a patient achieving complete remission (CR) or complete remission with incomplete hematologic recovery (CRi) gradually increased from 0% to 100%. This monotonic relationship enables clinicians to estimate the probability of a patient's response to Venetoclax-Azacitidine treatment based on the eight features included in the RF8 model. Importantly, the RF8 score was able to effectively distinguish patients with a high probability of response (probability > 80%) from those with a lower probability of benefit (probability < 30%) using the optimized cutoff values ​​of 0.67 and 0.37, respectively.

[0065] In addition, higher RF8 scores were consistently associated with better event-free survival (EFS) and overall survival (OS). Based on their RF8 scores, patients were divided into three equal groups: 0–34%, 34–68%, and 68–100%. This three-equal classification method demonstrated the strong discriminatory power of the RF8 model, and when the lowest tertile of RF8 scores (0–34%) was compared with the highest tertile (68–100%), the hazard ratios (HRs) for EFS and OS reached 6.9 and 6.7, respectively.

[0066] To benchmark the performance of the RF8 model against ELN2022 and other established prognostic systems for patients treated with Venetoclax-Azacitidine, the Brier score, which quantifies the prediction error by comparing the observed and predicted survival probabilities, was used. The RF8 model had a lower Brier score, indicating that its predictive accuracy for EFS and OS was superior to that of other prognostic systems. In addition, the C-index (a predictive accuracy metric ranging from 0 to 1, with 0.5 indicating random performance) was calculated. Notably, the RF8 model demonstrated a significantly higher C-index in EFS and OS predictions, highlighting its superior prognostic value.

[0067] The predictive and prognostic capabilities of the RF8 model were further validated in RJAML cohort 2, with the CR / CRi rate steadily increasing and the survival rate significantly improving as the RF8 score increased. Collectively, these findings suggest that the machine learning-based RF8 model can provide highly accurate predictions for treatment response, EFS, and OS before venetoclax-azacitidine treatment.

[0068] It can be seen that when applying the RF8 model for prediction, only the expression data of the patient's eight key genes need to be input. The expression data of these genes can be obtained through conventional technical means such as transcriptome sequencing, which is relatively simple to operate and does not require complex detection methods and a large amount of data input. The patient's gene expression data can be input through simple operations, and the system will automatically perform standardized processing and call the RF8 model for prediction, and output the patient's response probability value. This convenient prediction system makes the RF8 model easy to promote and apply in clinical practice. The continuous probability values ​​generated by the trained RF8 model are significantly correlated with the patient's treatment response rate and survival rate. This provides valuable insights for clinicians, helping them to evaluate the patient's response possibility to treatment based on the predicted probability value and adjust the treatment plan in an individualized manner. By setting an optimized cutoff value, patients with high and low response possibilities can be effectively distinguished. Doctors can classify patients as responders or non-responders based on the cutoff value, providing a clear basis for treatment decisions.

[0069] Reference Figure 3 As shown, the present invention also discloses a treatment response prediction device based on VEN / AZA combined treatment of acute myeloid leukemia, comprising:

[0070] The data acquisition module 11 is used to obtain target transcriptome data of the patient to be predicted for acute myeloid leukemia; the target transcriptome data is the expression amount information of a preset number of specific genes at the transcriptional level obtained after transcriptome sequencing of the patient to be predicted for acute myeloid leukemia; the specific genes are genes that have sensitive reactions or drug resistance reactions with VEN drugs or AZA drugs;

[0071] A probability prediction module 12 is used to input the target transcriptome data into a preset treatment response prediction model, so as to calculate the response probability value of the acute myeloid leukemia patient to be predicted after VEN / AZA combined treatment based on the preset treatment response prediction model and the target transcriptome data, so as to obtain a corresponding target response probability;

[0072] The result feedback module 13 is used to output the treatment response result corresponding to the VEN / AZA combined treatment regimen of the acute myeloid leukemia patient to be predicted based on the target response probability.

[0073] It can be seen that the present application discloses obtaining target transcriptome data of patients with acute myeloid leukemia to be predicted; the target transcriptome data is the expression information of a preset number of specific genes at the transcriptional level obtained after transcriptome sequencing of patients with acute myeloid leukemia to be predicted; the specific gene is a gene with a sensitive reaction or a drug resistance reaction to VEN drugs or AZA drugs; the target transcriptome data is input into a preset treatment response prediction model, so that the response probability value generated after VEN / AZA combined treatment of the patient with acute myeloid leukemia to be predicted is calculated based on the preset treatment response prediction model and the target transcriptome data to obtain the corresponding target response probability; the treatment response result corresponding to the VEN / AZA combined treatment scheme of the patient with acute myeloid leukemia to be predicted is output based on the target response probability. It can be seen that by analyzing the transcriptome data of specific genes with sensitive reactions or drug resistance reactions to VEN drugs or AZA drugs, the model can accurately calculate the response probability value of the patient to VEN / AZA combined treatment. A more accurate estimate of the treatment effect of the patient can be made before treatment, avoiding blind selection of treatment schemes and improving the scientificity and accuracy of treatment decisions.

[0074] Furthermore, the present application also discloses an electronic device. Figure 4 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram cannot be regarded as any limitation on the scope of use of the present application.

[0075] Figure 4 A schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, and the computer program is loaded and executed by the processor 21 to implement the relevant steps in the method for predicting the treatment response of acute myeloid leukemia based on the combined treatment of VEN / AZA with any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0076] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0077] Among them, the processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 21 can be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 21 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 21 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 21 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0078] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0079] Among them, the operating system 221 is used to manage and control the hardware devices and computer programs 222 on the electronic device 20, so as to realize the operation and processing of the massive data 223 in the memory 22 by the processor 21, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the treatment response prediction method based on VEN / AZA combined treatment of acute myeloid leukemia performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks. In addition to data transmitted from an external device received by the electronic device, the data 223 can also include data collected by its own input and output interface 25, etc.

[0080] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the aforementioned disclosed method for predicting the therapeutic response of acute myeloid leukemia based on the combined treatment of VEN / AZA is implemented. The specific steps of the method can be referred to the corresponding contents disclosed in the aforementioned embodiments, and will not be repeated here.

[0081] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0082] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application. The steps of the method or algorithm described in conjunction with the embodiments disclosed herein can be implemented directly with hardware, a software module executed by a processor, or a combination of the two. The software module can be placed in a random access memory RAM (Random Access Memory), memory, read-only memory ROM (Read Only Memory), electrically programmable EPROM (Electrically Programmable Read Only Memory), electrically erasable programmable EEPROM (ElectricErasable Programmable Read Only Memory), register, hard disk, removable disk, CD-ROM (CompactDisc-Read Only Memory), or any other form of storage medium known in the technical field.

[0083] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0084] The scheme provided by the present invention is introduced in detail above. Specific examples are used in this article to illustrate the principle and implementation mode of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation mode and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for predicting treatment response of acute myeloid leukemia based on VEN / AZA combined treatment, characterized in that: Applicable to computer devices, including: Obtaining target transcriptome data of the patient to be predicted for acute myeloid leukemia; the target transcriptome data is the expression amount information of a preset number of specific genes at the transcriptional level obtained after transcriptome sequencing of the patient to be predicted for acute myeloid leukemia; the specific genes are genes that have a sensitive reaction or a drug resistance reaction with VEN drug or AZA drug; Inputting the target transcriptome data into a preset treatment response prediction model, so as to calculate the response probability value of the acute myeloid leukemia patient to be predicted after VEN / AZA combined treatment based on the preset treatment response prediction model and the target transcriptome data, so as to obtain a corresponding target response probability; The treatment response result corresponding to the VEN / AZA combined treatment regimen of the acute myeloid leukemia patient to be predicted is output based on the target response probability.

2. The method for predicting treatment response of acute myeloid leukemia based on VEN / AZA combined treatment according to claim 1, characterized in that: After obtaining the target transcriptome data of the patient with acute myeloid leukemia to be predicted, the method further includes: Performing log2 normalization on the target transcriptome data to obtain normalized transcriptome data; Performing feature extraction processing on the standardized transcriptome data to obtain target data features including gene expression patterns and relationships between relative expression levels; Accordingly, inputting the target transcriptome data into a preset treatment response prediction model includes: The target data features are input into a preset treatment response prediction model.

3. The method for predicting treatment response of acute myeloid leukemia based on VEN / AZA combined treatment according to claim 1, characterized in that: Before inputting the target transcriptome data into a preset treatment response prediction model, the method further includes: The gene information that is resistant or sensitive to VEN or AZA drugs is screened from the transcriptome data of patients with acute myeloid leukemia who received VEN / AZA combination therapy to obtain the corresponding transcriptome data training set; The initial treatment response prediction model is trained according to the in vitro drug response data training set containing patient grouping data carrying drug resistance labels or sensitive labels and the transcriptome data training set, so that the initial treatment response prediction model learns the association between gene information, in vitro drug response data and drug resistance labels or sensitive labels that characterize the response to treatment, and updates the model parameters to obtain a preset treatment response prediction model.

4. The method for predicting treatment response of acute myeloid leukemia based on VEN / AZA combined treatment according to claim 3, characterized in that: The method uses the transcriptome data of patients receiving VEN / AZA combined treatment for acute myeloid leukemia to screen out gene information that has drug resistance or sensitivity to VEN drugs or AZA drugs to obtain a corresponding transcriptome data training set, including: The transcriptome data of sensitive and resistant patients with acute myeloid leukemia who received VEN / AZA combination therapy were compared, and the gene expression information associated with the efficacy of VEN or AZA drugs was screened out through a preset data analysis method to obtain the corresponding transcriptome data training set.

5. The method for predicting treatment response of acute myeloid leukemia based on VEN / AZA combined treatment according to claim 3, characterized in that: Before training the initial treatment response prediction model based on the in vitro drug response data training set containing patient grouping data carrying drug resistance labels or sensitive labels and the transcriptome data training set, the method further includes: Unsupervised clustering was performed on patients with acute myeloid leukemia who received VEN / AZA combination therapy to obtain patient grouping data under corresponding drug resistance labels and patient grouping data under sensitive labels, and an in vitro drug response data training set was constructed based on the patient grouping data under the drug resistance label and the patient grouping data under the sensitive label.

6. The method for predicting treatment response of acute myeloid leukemia based on VEN / AZA combined treatment according to claim 3, characterized in that: Before obtaining the preset treatment response prediction model, the method further includes: A five-fold cross-validation was performed on the process treatment response prediction model with updated model parameters to select the process treatment response prediction model with the best performance in the validation process as the preset treatment response prediction model.

7. The method for predicting treatment response of acute myeloid leukemia based on VEN / AZA combined treatment according to claim 1, characterized in that: The outputting of the treatment response result corresponding to the VEN / AZA combined treatment regimen of the acute myeloid leukemia patient to be predicted based on the target response probability includes: If the target response probability is greater than a preset response probability threshold, then determining that the treatment response effect of the VEN / AZA combined treatment regimen of the acute myeloid leukemia patient to be predicted corresponding to the target response probability is a sensitive response result; If the target response probability is less than or equal to a preset response probability threshold, it is determined that the therapeutic response effect of the VEN / AZA combined treatment regimen for the acute myeloid leukemia patient to be predicted corresponding to the target response probability is a drug resistance reaction result.

8. A therapeutic response prediction device based on VEN / AZA combined treatment of acute myeloid leukemia, characterized in that: include: A data acquisition module, used to acquire target transcriptome data of patients with acute myeloid leukemia to be predicted; The target transcriptome data is the expression information of a preset number of specific genes at the transcriptional level obtained after transcriptome sequencing of the patient to be predicted to have acute myeloid leukemia; the specific genes are genes that have sensitive reactions or drug resistance reactions to VEN drugs or AZA drugs; A probability prediction module, used for inputting the target transcriptome data into a preset treatment response prediction model, so as to calculate the response probability value of the acute myeloid leukemia patient to be predicted after VEN / AZA combined treatment based on the preset treatment response prediction model and the target transcriptome data, so as to obtain a corresponding target response probability; A result feedback module is used to output the treatment response result corresponding to the VEN / AZA combined treatment regimen of the acute myeloid leukemia patient to be predicted based on the target response probability.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor is used to execute the computer program to implement the steps of the method for predicting the treatment response of acute myeloid leukemia based on VEN / AZA combined treatment as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store computer programs; wherein, when the computer program is executed by a processor, the steps of the method for predicting treatment response of acute myeloid leukemia based on VEN / AZA combined treatment are implemented as described in any one of claims 1 to 7.

Images