Anti-PD-1 prognosis prediction method, system, device and medium for malignant melanoma
The prognostic prediction model constructed using a differential gene-weighted-machine learning triple filtering method solves the problem of lack of specificity and accuracy in existing melanoma anti-PD-1 treatment prediction models, and achieves high-precision prediction of the efficacy of anti-PD-1 treatment.
Patent Information
- Application Number
- CN202511039906.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-28
- Publication Date
- 2025-11-14
AI Technical Summary
Existing clinical prognostic prediction models for anti-PD-1 therapy in melanoma lack specificity and have low prediction accuracy, making it impossible to accurately predict treatment outcomes.
By acquiring pre-PD-1 sequencing data from patients with malignant melanoma, an optimal prognostic prediction model was constructed using a differential gene-weighted-machine learning triple filtering method. This included multiple rounds of filtering of characteristic genes of CXCL13+CD8T cells and training with machine learning algorithms to select the optimal prediction model.
It improves the predictive accuracy of anti-PD-1 therapy for malignant melanoma, meets the specificity and accuracy requirements of prediction, and provides more precise prediction of treatment efficacy.
Smart Images
Figure CN120954701A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of biomedicine, and in particular to a method, system, device and medium for predicting the prognosis of malignant melanoma using anti-PD-1 therapy. Background Technology
[0002] Malignant melanoma is a malignant tumor originating from melanocytes, primarily occurring in the skin, but also in mucous membranes, the uvea of the eye, and other sites. The treatment outcome of malignant melanoma is related to various factors, including tumor stage, thickness, location, and individual patient differences. Prognostic prediction, after a definitive diagnosis, is a method of comprehensively assessing the future progression of the disease, treatment response, and survival time based on various disease characteristics (such as pathological type, stage, gene expression, etc.), the patient's physical condition (age, underlying diseases, etc.), and other relevant factors. It is a crucial aspect of medical research and clinical practice. Clinical prognostic prediction models are typically based on RNA-seq sequencing to quantify the risk of disease progression after medication. For example, CN115631857A discloses a prognostic assessment method for thyroid cancer based on a CD8+ T cell immune-related gene model, including: screening for CD8+ T cell immune-related characteristic genes based on mRNA information from multiple thyroid cancer samples and multiple normal thyroid samples; analyzing the characteristic gene information and clinical data related to CD8+ T cells to construct a prognostic model of CD8+ T cell immune-related genes; and predicting the prognosis based on the prognostic model of CD8+ T cell immune-related genes. This method, based on a CD8+ T cell immune-related gene model, can accurately determine the prognosis of thyroid cancer patients. However, existing clinical prognostic prediction models for melanoma anti-PD-1 therapy mostly use models that predict broad immunotherapy, lacking specificity for predicting anti-PD-1 therapy; furthermore, previous studies often used features that were not rigorously filtered, resulting in low model prediction accuracy. Summary of the Invention
[0003] The purpose of this invention is to overcome the defects of the prior art and provide a method, system, device and medium for predicting the prognosis of malignant melanoma using anti-PD-1.
[0004] The objective of this invention can be achieved through the following technical solutions:
[0005] According to a first aspect of the present invention, an anti-PD-1 prognostic prediction method for malignant melanoma is provided, the method comprising the following steps:
[0006] Obtain pre-treatment sequencing data and corresponding anti-PD-1 treatment results of patients with malignant melanoma, including disease progression and no disease progression.
[0007] Calculate the genes upregulated in patients with no disease progression in the public ICI cohort compared to patients with disease progression, and perform the first round of filtering for subgroup marker genes;
[0008] For CXCL13+CD8T cells that have undergone the first round of filtering, their first weighted expression score in the target cell population is calculated, and the difference between the first weighted expression score and the second weighted expression score of other cell types in the microenvironment in the target cell population is calculated to obtain the weighted difference. The genes are then filtered in the second round based on the weighted difference.
[0009] Based on multiple machine learning algorithms, a third round of filtering is performed on the gene features that have undergone the second round of filtering to construct a dataset;
[0010] The constructed dataset was used to train and test anti-PD-1 prognostic prediction models based on different machine learning models to obtain the optimal prognostic prediction model. The anti-PD-1 prognostic prediction model takes the screened gene features as input and the anti-PD-1 treatment result as output.
[0011] The optimal prognostic prediction model is used to predict treatment outcomes.
[0012] As a preferred technical solution, the weighted expression score is the product of the subpopulation ratio, expression level, and positive cell rate within the subpopulation.
[0013] As a preferred technical solution, the second round of gene filtering based on weighted difference specifically involves selecting genes with positive weighted difference values as potential biomarkers and filtering out genes with weighted difference values of 0 or negative.
[0014] As a preferred technical solution, the third round of filtering of gene features after the second round of filtering based on multiple machine learning algorithms specifically involves: obtaining the first feature genome without filtering, obtaining the second feature genome based on random forest filtering, obtaining the third feature genome based on support vector machine filtering, obtaining the fourth feature genome based on elastic network filtering, and obtaining the fifth feature genome based on logistic regression filtering.
[0015] According to a second aspect of the present invention, an anti-PD-1 prognostic prediction system for malignant melanoma is provided, comprising:
[0016] Data acquisition module: Acquires pre-treatment sequencing data and corresponding anti-PD-1 treatment results of patients with malignant melanoma, including disease progression and no disease progression;
[0017] The triple-filtering module performs the following steps: First, it calculates the upregulated genes in non-progressing patients compared to those in progressive patients in the public ICI cohort, performing a first round of filtering on subpopulation marker genes. For CXCL13+CD8T cells that have passed the first round of filtering, it calculates their first weighted expression score in the target cell population and the difference between this first weighted expression score and the second weighted expression scores of other cell types in the microenvironment within the target cell population, obtaining a weighted difference. Based on this weighted difference, it performs a second round of filtering on genes. Finally, based on multiple machine learning algorithms, it performs a third round of filtering on the gene features that have passed the second round of filtering, constructing the dataset.
[0018] Model training module: The constructed dataset is used to train and test the anti-PD-1 prognostic prediction models built based on different machine learning models to obtain the optimal prognostic prediction model. The anti-PD-1 prognostic prediction model takes the screened gene features as input and the anti-PD-1 treatment results as output.
[0019] Prediction module: Utilizes the optimal prognosis prediction model to predict treatment outcomes.
[0020] As a preferred technical solution, the weighted expression score is the product of the subpopulation ratio, expression level, and positive cell rate within the subpopulation.
[0021] As a preferred technical solution, the second round of gene filtering based on weighted difference specifically involves selecting genes with positive weighted difference values as potential biomarkers and filtering out genes with weighted difference values of 0 or negative.
[0022] As a preferred technical solution, the third round of filtering of gene features after the second round of filtering based on multiple machine learning algorithms specifically involves: obtaining the first feature genome without filtering, obtaining the second feature genome based on random forest filtering, obtaining the third feature genome based on support vector machine filtering, obtaining the fourth feature genome based on elastic network filtering, and obtaining the fifth feature genome based on logistic regression filtering.
[0023] According to a third aspect of the present invention, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described thereon.
[0024] According to a fourth aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method described thereon.
[0025] Compared with the prior art, the present invention has the following beneficial effects:
[0026] This invention is based on tumor-invasive effector CXCL13+CD8T cells confirmed by omics and basic experimental research. Through differential gene-weighted-machine learning triple filtering and multiple training, the obtained gene features can accurately express PD-1 specificity, thereby training an optimal prognostic prediction model for predicting the efficacy of anti-PD1 therapy for malignant melanoma, improving the accuracy of the prediction model and meeting the specificity of predicting anti-PD-1 therapy. Attached Figure Description
[0027] Figure 1 This is a flowchart of the method of the present invention;
[0028] Figure 2 This is a schematic diagram of the model training process in one embodiment.
[0029] Figure 3 This is a schematic diagram of a third round of screening in machine learning in one embodiment, where (a) represents a random forest model, (b) represents an elastic network model, (c) represents an SVM, and (d) represents logistic regression.
[0030] Figure 4 This is a histogram of the average AUC of the model on the validation and test sets in one embodiment.
[0031] Figure 5 This is a schematic diagram illustrating the stability of the model's AUC on the validation and test sets in one embodiment.
[0032] Figure 6 The AUC of seven models trained after random forest selection in one embodiment is shown on the validation set.
[0033] Figure 7 This represents the AUC of seven models trained using random forest filtering in one embodiment on the test set.
[0034] Figure 8 This is a schematic diagram of survival analysis of model-based prognostic prediction results in one embodiment. Detailed Implementation
[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.
[0036] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.
[0037] This embodiment provides a method for predicting the prognosis of malignant melanoma using anti-PD-1 inhibitors, such as... Figure 1 As shown, the method includes the following steps:
[0038] S1: Obtain pre-treatment sequencing data and corresponding anti-PD-1 treatment results for patients with malignant melanoma, including disease progression and no disease progression.
[0039] This embodiment obtained publicly available RNA-seq data from 244 melanoma patients as model filtering and training data. The sequencing data of these 244 patients were all measured before PD-1 treatment, and the efficacy after treatment (progressive PD / non-progressive NPD) was recorded.
[0040] This embodiment first identifies and confirms, based on bioinformatics analysis of public databases, that tumor-specific CXCL13+CD8T cells are the main effector cells of anti-tumor immunity. Further, through Western blotting, confocal microscopy, immunohistochemistry, and validation using other public databases, it is demonstrated that CXCL13+CD8T promotes anti-tumor immunity by facilitating the formation of tertiary lymphoid structures. Furthermore, using existing transcriptome data, it is verified that CXCL13+CD8T is a good biomarker for predicting the efficacy of anti-PD-1 therapy for melanoma. Based on these conclusions, subsequent feature filtering and screening can better identify gene characteristics suitable for anti-PD-1 therapy in melanoma.
[0041] To more accurately obtain the feature genes of CXCL13+CD8T, a differential gene-weighted-machine learning triple filtering scheme is used to filter the feature genes of CXCL13+CD8T, as shown in steps S2 to S4. This process is as follows: Figure 2 As shown in I, II, and III.
[0042] S2, calculate the genes upregulated in patients with no disease progression in the public ICI cohort compared to patients with disease progression, and perform a first round of filtering on the subgroup marker genes to ensure that upregulation of any gene in the above gene list at the RNA level is beneficial for controlling disease progression.
[0043] In one embodiment, the process includes the following steps: First, the gene expression data is standardized to remove low-expression genes and batch effects. Second, statistical tests (such as t-tests, Wilcoxon rank-sum tests, etc.) are used to compare gene expression differences between the disease-free and disease-progressing groups. Screening criteria are set to identify genes that are significantly upregulated in the disease-free group. Alternatively, genes can be ranked or scored based on statistical measures of the difference analysis (such as t-values or p-values), and genes with higher scores can be selected as the filtered gene list.
[0044] S3. For the CXCL13+CD8T cells that have passed the first round of filtering, calculate their first weighted expression score in the target cell population, and calculate the difference between the first weighted expression score and the second weighted expression score of other cell types in the microenvironment in the target cell population to obtain the weighted difference. Select genes with positive weighted differences as potential markers, and screen out genes with weighted differences of 0 or negative to perform a second round of gene filtering.
[0045] This step assigns weights based on the composition ratio of CXCL13+CD8T cells in the microenvironment and the positive expression rate of the gene in the subpopulation. The weighted expression score is the product of the subpopulation proportion, expression level, and positive cell rate within the subpopulation. For the first weighted expression score, the subpopulation proportion can be obtained using techniques such as flow cytometry or single-cell sequencing to determine the percentage of CXCL13+CD8T cells in the target cell population. For example, if CXCL13+CD8T cells account for X% of all cells in the target cell population, then the subpopulation proportion is X%. The expression level can be measured using microarrays or sequencing technology to determine the expression level of the target gene in the cells. RNA expression level can be measured using qPCR or RNA-seq, and the expression levels of different samples are normalized to eliminate sample differences. The positive cell rate within the subpopulation can be determined using methods such as immunohistochemical staining or flow cytometry to determine the proportion of cells within the CXCL13+CD8T cell subpopulation that are positive for the target gene out of the total number of cells in that subpopulation. Multiplying these three parameters yields the first weighted expression score. The calculation process for the second weighted expression score can refer to the above process, and will not be repeated here in this embodiment.
[0046] This step ensures that the selected genes have both target cell population-specific pro-immune effects and are not affected by expression in other cell types. This means that when the weighted genes show an increase in RNA-seq levels, it is due to the upregulation of their expression in the corresponding cell type, rather than the upregulation of their expression in other cell types.
[0047] S4, based on multiple machine learning algorithms, performs a third round of filtering on the gene features that have passed the second round of filtering to construct a dataset.
[0048] This embodiment employs four machine learning algorithms to further filter these features: without machine learning filtering, a first feature genome (containing 57 feature genes) is obtained; a second feature genome (containing 41 feature genes) is obtained based on random forest filtering; a third feature genome (containing 9 feature genes) is obtained based on support vector machine filtering; a fourth feature genome (containing 22 feature genes) is obtained based on elastic network filtering; and a fifth feature genome (containing 6 feature genes) is obtained based on logistic regression filtering.
[0049] like Figure 3 As shown, (a) represents the process of random forest algorithm to select genes with importance score > 0; (b) represents the set of genes with minimum penalty value when elastic network is used to select genes; (c) represents the number of genes when support vector machine has the highest selection efficiency; and (d) represents the optimization process of logistic regression to select genes.
[0050] S5. Using the constructed dataset, the anti-PD-1 prognostic prediction models built based on different machine learning models are trained and tested to obtain the optimal prognostic prediction model. The anti-PD-1 prognostic prediction model takes the screened gene features as input and the anti-PD-1 treatment results as output.
[0051] These five gene sets were used as features, and the biomarker was filtered and optimized using public data. Seven machine learning algorithms were used to construct anti-PD-1 prognostic prediction models, resulting in 35 models predicting drug response to PD-1 monoclonal antibody treatment for malignant melanoma. These models were then tested and validated on a validation set consisting of 49 transcriptome datasets, and the model with the optimal area under the curve (AUC) was selected. Figure 4 The average AUC of 35 models on the validation and test sets is shown. Figure 5 The stability of AUC, i.e., 1 - range rate, is demonstrated. Model selection is based on this result. In one embodiment, Figure 6 and Figure 7 The AUC of seven models trained using random forest selection is shown on the validation and test sets. Then, the models were tested again on a test set containing 11 patients (an RNA sequencing cohort obtained by pseudoRNA calculation based on single-cell sequencing data from these 11 patients), and the model with the highest stability and accuracy was selected as the optimal prognostic prediction model.
[0052] This embodiment uses a weighted gene score based on single-cell data to filter characteristic gene modeling, which is accurate and stable. It has good accuracy in predicting the treatment of malignant melanoma with PD-1 monoclonal antibody. The optimal prognostic prediction model has an ROC of 0.857 on the validation set, an ROC of 0.896 on the test set, and a 1-range of 95.586.
[0053] S6. The optimal prognostic prediction model is used to predict the treatment effect.
[0054] The optimal prognostic prediction model is based on the R language and stored in an .Rdata file. It can predict whether patients with malignant melanoma will experience disease progression after receiving PD-1 monoclonal antibodies based on their transcriptome data, thereby helping them decide whether to choose this therapy.
[0055] The following is a code example of using a .Rdata file to predict treatment effectiveness:
[0056] ## Loading R packages
[0057] library(caret)
[0058] library(class)
[0059] library(pROC)
[0060] library(adabag)
[0061] library(ROCR)
[0062] library(survival)
[0063] library(survminer)
[0064] ## Read the test set and the model
[0065] load("set.Rdata")
[0066] load("comprehensive_AdaBoost.Rdata")
[0067] load("survival information.Rdata")
[0068] ## Predict and evaluate
[0069] y_pred<-predict.boosting(model,newdata=test_meta[,c("response",geneset)],type='prob')
[0070] predictFull_rf<-prediction(y_pred$prob[,1],test_meta$responseNUM)
[0071] perfFull_rf<-performance(predictFull_rf,measure="tpr",x.measure="fpr")
[0072] AUC=performance(predictFull_rf,"auc")@y.values[[1]]
[0073] ## Predict survival
[0074] y_pred<-predict.boosting(model,newdata=meta2[,c("response",combined_genes)],type='prob')
[0075] meta2$predict=y_pred$class
[0076] meta2$predict==meta2$response
[0077] meta2$OS=as.numeric(meta2$OS)
[0078] meta2$OS = meta2$OS / 30
[0079] meta2$dead=as.numeric(meta2$dead)
[0080] sfit<-survfit(Surv(OS,dead)~predict,data=meta2)
[0081] ggsurvplot(sfit, pval = TRUE)
[0082] To use this method, import the RNA-seq and clinical information of the patients to be predicted into R language and organize it into test_meta (row names are patient names, column names are the gene names required for prediction, including: ABCC1, ABCD2, ACSL5, ASB2, CARD11, CATSERB, CCDC141, CDYL2, CLNK, CMIP, DOCK2, DOCK8, EOMES, EVL, FCRL3, GAB3, GBP5, IKZF3, ITGAD, ITGAL, JAKMIP1, MYO1F, NFATC2, OSBPL3, PARP15, PARVG, PCED1B, PDCD1, PLCG2, PLEK, PRKCB, PRKCH, PTK2B, RPTOR, SAMD3, SH2D3C, SIRPG, SKAP1, SLA2, SLAMF7). Running the above code will output the predicted risk of progression after anti-PD-1 treatment in melanoma patients.
[0083] The results obtained by running the code are as follows: View the y_pred object in R language. The results are shown in Table 1 below, which is the predicted disease progression score for patients. For each row (each patient), the first column is the patient's disease progression risk score. The higher the score, the greater the patient's risk of disease progression.
[0084] Table 1
[0085] Sample number PD NPD Pt2_Pre_AD101150-6 0.418 0.582 Pt59_Pre_AD823915-5 0.196 0.804 Pt82_Pre_AD823914-8 0.290 0.710 Pt66_Pre_AD667850-6 0.446 0.554 Pt67_Pre_AD506074-6 0.440 0.560 Pt94_Pre_AD732850-6 0.552 0.448 Pt62_Pre_AD608303-5 0.724 0.276 Pt90_Pre_AD467873-6 0.450 0.550 Pt9_Pre_E9021024-6 0.404 0.596 Pt18_Pre_E9024732-6 0.120 0.880 Pt24_Pre_AD436687-5 0.638 0.362 Pt30_Pre_AD497503-5 0.364 0.636
[0086] Figure 8 This is a result diagram of survival analysis using model prediction results in one embodiment. As can be seen, the method proposed in this invention can perform prognostic prediction very well, providing effective and accurate analytical basis for survival analysis.
[0087] The above is an introduction to the method embodiments. The following system embodiments will further illustrate the solution of the present invention.
[0088] An anti-PD-1 prognostic prediction system for malignant melanoma, comprising:
[0089] Data acquisition module: Acquires pre-treatment sequencing data and corresponding anti-PD-1 treatment results of patients with malignant melanoma, including disease progression and no disease progression;
[0090] The triple-filtering module performs the following steps: First, it calculates the upregulated genes in non-progressing patients compared to those in progressive patients in the public ICI cohort, performing a first round of filtering on subpopulation marker genes. For CXCL13+CD8T cells that have passed the first round of filtering, it calculates their first weighted expression score in the target cell population and the difference between this first weighted expression score and the second weighted expression scores of other cell types in the microenvironment within the target cell population, obtaining a weighted difference. Based on this weighted difference, it performs a second round of filtering on genes. Finally, based on multiple machine learning algorithms, it performs a third round of filtering on the gene features that have passed the second round of filtering, constructing the dataset.
[0091] Model training module: The constructed dataset is used to train and test the anti-PD-1 prognostic prediction models built based on different machine learning models to obtain the optimal prognostic prediction model. The anti-PD-1 prognostic prediction model takes the screened gene features as input and the anti-PD-1 treatment results as output.
[0092] Prediction module: Utilizes the optimal prognosis prediction model to predict treatment outcomes.
[0093] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0094] The electronic device of this invention includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in read-only memory (ROM) or loaded from a storage unit into random access memory (RAM). The RAM may also store various programs and data required for device operation. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.
[0095] Multiple components in the device are connected to the I / O interface, including: input units such as keyboards and mice; output units such as various types of displays and speakers; storage units such as disks and optical discs; and communication units such as network interface cards (NICs), modems, and wireless transceivers. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0096] The processing unit executes the various methods and processes described above, such as methods S1 to S6. For example, in some embodiments, methods S1 to S6 may be implemented as computer software programs tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via ROM and / or a communication unit. When the computer program is loaded into RAM and executed by the CPU, one or more steps of methods S1 to S6 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S6 by any other suitable means (e.g., by means of firmware).
[0097] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.
[0098] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.
[0099] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more 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 of the foregoing.
[0100] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.
[0101] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A method for predicting the prognosis of malignant melanoma using anti-PD-1 inhibitors, characterized in that, The method includes the following steps: Obtain pre-treatment sequencing data and corresponding anti-PD-1 treatment results of patients with malignant melanoma, including disease progression and no disease progression. Calculate the genes upregulated in patients with no disease progression in the public ICI cohort compared to patients with disease progression, and perform the first round of filtering for subgroup marker genes; For CXCL13+CD8T cells that have undergone the first round of filtering, their first weighted expression score in the target cell population is calculated, and the difference between the first weighted expression score and the second weighted expression score of other cell types in the microenvironment in the target cell population is calculated to obtain the weighted difference. The genes are then filtered in the second round based on the weighted difference. Based on multiple machine learning algorithms, a third round of filtering is performed on the gene features that have undergone the second round of filtering to construct a dataset; The constructed dataset was used to train and test anti-PD-1 prognostic prediction models based on different machine learning models to obtain the optimal prognostic prediction model. The anti-PD-1 prognostic prediction model takes the screened gene features as input and the anti-PD-1 treatment result as output. The optimal prognostic prediction model is used to predict treatment outcomes.
2. The method for predicting the prognosis of malignant melanoma using anti-PD-1 therapy according to claim 1, characterized in that, The weighted expression score is the product of the subpopulation proportion, expression level, and positive cell rate within the subpopulation.
3. The method for predicting the prognosis of malignant melanoma using anti-PD-1 therapy according to claim 1, characterized in that, The second round of gene filtering based on weighted difference specifically involves selecting genes with positive weighted difference values as potential biomarkers and filtering out genes with weighted difference values of 0 or negative.
4. The method for predicting the prognosis of malignant melanoma using anti-PD-1 therapy according to claim 1, characterized in that, The third round of filtering of gene features after the second round of filtering, based on multiple machine learning algorithms, specifically involves: obtaining the first feature genome without filtering, obtaining the second feature genome based on random forest filtering, obtaining the third feature genome based on support vector machine filtering, obtaining the fourth feature genome based on elastic network filtering, and obtaining the fifth feature genome based on logistic regression filtering.
5. An anti-PD-1 prognostic prediction system for malignant melanoma, characterized in that, include: Data acquisition module: Acquires pre-treatment sequencing data and corresponding anti-PD-1 treatment results of patients with malignant melanoma, including disease progression and no disease progression; The triple-filtering module performs the following steps: First, it calculates the upregulated genes in non-progressing patients compared to those in progressive patients in the public ICI cohort, performing a first round of filtering on subpopulation marker genes. For CXCL13+CD8T cells that have passed the first round of filtering, it calculates their first weighted expression score in the target cell population and the difference between this first weighted expression score and the second weighted expression scores of other cell types in the microenvironment within the target cell population, obtaining a weighted difference. Based on this weighted difference, it performs a second round of filtering on genes. Finally, based on multiple machine learning algorithms, it performs a third round of filtering on the gene features that have passed the second round of filtering, constructing the dataset. Model training module: The constructed dataset is used to train and test the anti-PD-1 prognostic prediction models built based on different machine learning models to obtain the optimal prognostic prediction model. The anti-PD-1 prognostic prediction model takes the screened gene features as input and the anti-PD-1 treatment results as output. Prediction module: Utilizes the optimal prognosis prediction model to predict treatment outcomes.
6. The anti-PD-1 prognostic prediction system for malignant melanoma according to claim 5, characterized in that, The weighted expression score is the product of the subpopulation proportion, expression level, and positive cell rate within the subpopulation.
7. The anti-PD-1 prognostic prediction system for malignant melanoma according to claim 5, characterized in that, The second round of gene filtering based on weighted difference specifically involves selecting genes with positive weighted difference values as potential biomarkers and filtering out genes with weighted difference values of 0 or negative.
8. The anti-PD-1 prognostic prediction system for malignant melanoma according to claim 5, characterized in that, The third round of filtering of gene features after the second round of filtering, based on multiple machine learning algorithms, specifically involves: obtaining the first feature genome without filtering, obtaining the second feature genome based on random forest filtering, obtaining the third feature genome based on support vector machine filtering, obtaining the fourth feature genome based on elastic network filtering, and obtaining the fifth feature genome based on logistic regression filtering.
9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 4.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 4.