Computer apparatus for predicting the efficacy of immunotherapy for HIV infection

Abstract

The application discloses a computer device for predicting the curative effect of HIV infection immune adjuvant therapy. It belongs to the field of computer information. The technical problem solved by the application is how to predict the curative effect of HIV infection immune adjuvant therapy. The application provides a computer program stored in a memory in a data processing device: S1, receiving sample data, the characteristic values of any one or several combinations of 7 metabolites and 3 bacteria in an ex vivo fecal sample of a subject; S2, inputting the sample data into an immune regulation effect prediction model of HIV immune non-responsive patients treated with shujibing; S3, outputting the probability value of the immune regulation effect prediction of the subject treated with shujibing by the immune regulation effect prediction model of HIV immune non-responsive patients treated with shujibing; and deriving the immune regulation effect of the subject treated with shujibing based on the probability value. The application can accurately and efficiently predict the curative effect of HIV infection immune adjuvant therapy.

Description

Technical Field

[0001] This invention belongs to the field of computer information and relates to a computer device for predicting the efficacy of adjuvant immunotherapy for HIV infection. Background Technology

[0002] By the end of 2023, 95% of HIV-infected individuals in my country had received antiretroviral therapy. While antiretroviral therapy effectively suppresses viral replication, 10-40% of infected individuals still fail to regain their immune status, often accompanied by chronic inflammation and abnormal immune activation, leading to an increased risk of non-AIDS-related complications and death. However, the mechanisms underlying this immune non-response are not fully understood, and there are currently no definitively proven effective treatments. Studies have found that some immune-adjuvant drugs can promote CD4+ in this population. + T cell recovery and reduced T cell activation are observed, but individual heterogeneity exists. In tumor immunotherapy, differences in gut microbiota can influence the efficacy of immune checkpoint inhibitors to some extent. However, in HIV adjuvant immunotherapy, there are no studies on the relationship between gut microbiota and the efficacy of adjuvant immunotherapy drugs, nor are there gut microbiota models that can effectively predict the effects of HIV adjuvant immunotherapy. Summary of the Invention

[0003] The technical problem solved by this invention is to prepare a computer device or model for predicting the efficacy of adjuvant immunotherapy for HIV infection.

[0004] A first aspect of the present invention provides a data processing apparatus, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to perform the following steps:

[0005] S1. Receiving data: Receiving sample data, wherein the sample data is the feature value of any one or any combination of 10 factors in the subject's ex vivo fecal sample;

[0006] The 10 factors include 7 metabolites and 3 bacteria;

[0007] The three bacteria are Phenylobacterium, Latilactobacillus, and Pseudactinotale;

[0008] The seven metabolites are 7-Oxocholesterol, Pipertipine, Hydrolarazine pyruvatehydrazone, Nevirapine, Arbutin, Vanilloylglycine, and cis-5,8,11,14,17-Eicosapentaenoic acid.

[0009] The characteristic value of each bacterium is its abundance.

[0010] The characteristic value of each metabolite is its metabolite content;

[0011] S2. Input data: Input the sample data into the immunomodulatory effect prediction model for silymarin treatment of HIV immune non-responsive patients;

[0012] The immunomodulatory effect prediction model for silybin treatment in HIV-unresponsive patients is constructed according to the following steps: using the feature values ​​of the 10 factors or any combination thereof in the ex vivo fecal samples of subjects who are known to have responded to silybin treatment after antiviral therapy and subjects who are known to have not responded to silybin treatment after antiviral therapy, as training samples, to train the immunomodulatory effect prediction model for silybin treatment in HIV-unresponsive patients.

[0013] The immunomodulatory effect is defined as either a response to immunomodulation or a lack of response to immunomodulation.

[0014] S3. Output Results: The immunomodulatory effect prediction model for silymarin treatment in HIV immune non-responsive patients is used to output the probability value of the subject predicting the immunomodulatory effect of silymarin treatment; and the immunomodulatory effect of silymarin treatment in the subject is then calculated based on the probability value.

[0015] The subjects were HIV-immune non-response patients after antiviral treatment.

[0016] In the above text, the abundance refers to ASV abundance.

[0017] In the above text, the content of metabolites refers to the relative content of metabolites, which can be specifically represented by the signal intensity (the intensity of the peaks in the mass spectrum) of each metabolite.

[0018] In the above text, both the known HIV immune non-responsive patients after antiviral therapy and the subjects who responded to silymarin treatment for immunomodulation, and the known HIV immune non-responsive patients after antiviral therapy who did not respond to silymarin treatment for immunomodulation, were patients with abnormal liver function.

[0019] In the above text, the data processing device may be a computer device.

[0020] In a second aspect, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the steps in the data processing apparatus described in the first aspect.

[0021] Thirdly, the present invention provides a computer-readable storage medium storing a computer program that causes a computer to perform the steps in the data processing apparatus described in the first aspect.

[0022] Fourthly, the present invention provides an apparatus for predicting the immunomodulatory effects of silymarin adjuvant therapy in HIV immune-unresponsive patients, the apparatus comprising:

[0023] S1, Data receiving module: used to receive sample data, wherein the sample data is the characteristic values ​​of 10 factors or any one or any combination thereof in the ex vivo fecal sample of the subject;

[0024] The 10 factors include 7 metabolites and 3 bacteria;

[0025] The three bacteria are Phenylobacterium, Latilactobacillus, and Pseudactinotale;

[0026] The seven metabolites are 7-Oxocholesterol, Pipertipine, Hydrolarazine pyruvatehydrazone, Nevirapine, Arbutin, Vanilloylglycine, and cis-5,8,11,14,17-Eicosapentaenoic acid.

[0027] The characteristic value of each bacterium is its abundance.

[0028] The characteristic value of each metabolite is its metabolite content;

[0029] S2, Data Input Module: Used to input the sample data into the immunomodulatory effect prediction model for silymarin treatment of HIV immune non-responders;

[0030] The immunomodulatory effect prediction model for silybin treatment in HIV-unresponsive patients is constructed according to the following steps: using the feature values ​​of the 10 factors or any combination thereof in the ex vivo fecal samples of subjects who are known to have responded to silybin treatment after antiviral therapy and subjects who are known to have not responded to silybin treatment after antiviral therapy, as training samples, to train the immunomodulatory effect prediction model for silybin treatment in HIV-unresponsive patients.

[0031] The immunomodulatory effect is defined as either a response to immunomodulation or a lack of response to immunomodulation.

[0032] S3. Result Output Module: Used to output the probability value of the subject predicting the immunomodulatory effect of silymarin treatment in HIV immune non-responsive patients through the immunomodulatory effect prediction model; and then calculate the immunomodulatory effect of silymarin treatment in the subject based on the probability value.

[0033] The subjects were HIV-immune non-response patients after antiviral treatment.

[0034] Fifthly, the present invention provides a method for predicting or assisting in predicting the immunomodulatory effect of silymarin treatment in subjects, the method comprising the following steps:

[0035] S1. Obtain the feature values ​​of any one or any combination of 10 factors in the isolated fecal sample of the subject;

[0036] The 10 factors include 7 metabolites and 3 bacteria;

[0037] The three bacteria are Phenylobacterium, Latilactobacillus, and Pseudactinotale;

[0038] The seven metabolites are 7-Oxocholesterol, Pipertipine, Hydrolarazine pyruvatehydrazone, Nevirapine, Arbutin, Vanilloylglycine, and cis-5,8,11,14,17-Eicosapentaenoic acid.

[0039] The characteristic value of each bacterium is its abundance.

[0040] The characteristic value of each metabolite is its metabolite content;

[0041] S2. Input the sample data into the immunomodulatory effect prediction model for silymarin treatment of HIV immune non-responders;

[0042] The immunomodulatory effect prediction model for silybin treatment in HIV-unresponsive patients is constructed according to the following steps: using the feature values ​​of the 10 factors or any combination thereof in the ex vivo fecal samples of subjects who are known to have responded to silybin treatment after antiviral therapy and subjects who are known to have not responded to silybin treatment after antiviral therapy, as training samples, to train the immunomodulatory effect prediction model for silybin treatment in HIV-unresponsive patients.

[0043] The immunomodulatory effect is defined as either a response to immunomodulation or a lack of response to immunomodulation.

[0044] S3. The immunomodulatory effect prediction model for silymarin treatment in HIV immune non-responsive patients is used to output the probability value of the subject predicting the immunomodulatory effect of silymarin treatment; and the immunomodulatory effect of silymarin treatment in the subject is then calculated based on the probability value.

[0045] The subjects were HIV-immune non-response patients after antiviral treatment.

[0046] In a sixth aspect, the present invention provides a method for constructing a predictive model of the immunomodulatory effect of silymarin in treating HIV-unresponsive patients, comprising the following steps: using the feature values ​​of the 10 factors or any combination thereof in ex vivo fecal samples from subjects who are known to have responded to immunomodulation with silymarin treatment after antiviral therapy and subjects who are known to have not responded to immunomodulation with silymarin treatment after antiviral therapy as training samples to train the predictive model of the immunomodulatory effect of silymarin in treating HIV-unresponsive patients;

[0047] The immunomodulatory effect is defined as either a positive or negative immune response.

[0048] The methods described above may not include the step of obtaining biological samples from animals. All methods may not target living human or animal bodies, but only data. All methods may be information processing methods in which all steps are performed by a data processing device such as a computer.

[0049] In the above text, any one or any combination of the 10 factors is as follows:

[0050] The 10 factors mentioned in A1);

[0051] The seven metabolites mentioned in A2);

[0052] A3)7-Oxocholesterol, Hydralazine pyruvate hydrazone, Nevirapine, Arbutin and cis-5,8,11,14,17-Eicosapentaenoic acid;

[0053] The three bacteria mentioned in A4);

[0054] A5)Phenylobacterium and 7-Oxocholesterol;

[0055] A6) Phenylobacterium, 7-Oxocholesterol and Hydralazine pyruvate hydrazone;

[0056] A7) Phenylobacterium, 7-Oxocholesterol, Hydralazine pyruvate hydrazone, Nevirapine, Arbutin and cis-5,8,11,14,17-Eicosapentaenoic acid.

[0057] In a seventh aspect, the present invention provides the application of silymarin in the preparation of products for restoring the immunomodulatory response in HIV-immune non-responders;

[0058] Alternatively, this invention provides silymarin for preparing CD4 in HIV-recovering or enhancing individuals with unresponsive HIV immune systems. + Application in T-cell count products.

[0059] Given the lack of effective treatment options for HIV-infected patients with unresponsive immune systems, the individual heterogeneity of efficacy of various immunomodulatory therapies, and the absence of methods for early prediction of the efficacy of adjuvant immunotherapy, this invention demonstrates a plant extract with immunomodulatory effects—silymarin—which promotes CD4 activation. + While T-cell recovery plays a role, given the continued heterogeneity in efficacy, establishing a classification model based on gut microbiota to predict the efficacy of adjuvant drugs can help doctors implement individualized precision treatment. For patients who are not sensitive to drugs, further intervention through gut microbiota can improve efficacy or enable timely replacement of other drugs, restore patients' immune function, and reduce the risk of non-AIDS chronic complications.

[0060] This invention provides a method for constructing a predictive model for the efficacy of adjuvant immunotherapy for HIV infection. By combining pre-treatment gut bacteria and related metabolites, it predicts the regulatory effect of adjuvant immunotherapy on HIV immune non-response, thereby guiding precision treatment, reducing excessive immune activation and chronic inflammation in HIV-infected individuals, lowering the risk of non-AIDS chronic complications and death, and helping doctors to carry out individualized precision interventions to effectively correct immune disorders and reduce related complications in patients. Attached Figure Description

[0061] Figure 1 Flowchart for a model to predict the efficacy of adjuvant immunotherapy for HIV infection.

[0062] Figure 2 Changes in T cell subsets, including CD4, in HIV-infected immune non-responders after silymarin treatment. + T cells and CD8 +T cell count, CD4 + T cells and CD8 + The ratio of T cells. In the figure, SIL represents silybin, 6m pre SIL represents 6 months before silybin treatment, SIL MO represents the baseline of silybin treatment, and SIL M3 represents 3 months of silybin treatment.

[0063] Figure 3 This figure shows the changes in T-cell immunophenotypes in HIV-infected immune non-responders after silybin treatment, including Treg cells, Th17 cells, and cell activation, senescence, and differentiation. In the figure, "pre" represents the baseline of silybin treatment, and "post" represents 3 months of silybin treatment.

[0064] Figure 4 This figure shows the difference in peripheral blood T cell subsets between subjects who responded to and did not respond to silybin treatment. In the figure, R represents subjects who responded to silybin treatment, NR represents subjects who did not respond to silybin treatment, pre represents the baseline of silybin treatment, and post represents 3 months of silybin treatment.

[0065] Figure 5 The baseline fecal bacterial spore composition for silymarin treatment in both the immunomodulatory responsive and non-responsive groups was determined.

[0066] Figure 6 To identify differentially expressed bacterial genera in the feces of subjects who responded to silybin treatment and those who did not, the LefSe differential analysis method was used to screen for differentially expressed bacterial genera based on an LDA score >2.

[0067] Figure 7 Differential metabolites in feces at baseline were identified between silybin-treated immunomodulatory response and non-responder groups. Univariate analysis was used to assess fold-change and t-test statistically, combined with multivariate statistical analysis (PLS-DA) to obtain the VIP (Variable Important for the Projection) value. Differential metabolites (relative amounts of each metabolite obtained from non-targeted metabolites) were screened based on criteria of P < 0.05, fold-change > 2, and VIP > 1.

[0068] Figure 8To effectively distinguish the gut microbiota characteristics of subjects responding to and not responding to silymarin treatment for immunomodulation, a random forest algorithm was used to screen for differentially expressed bacteria and metabolites between the two groups. The top 10 factors with the highest classification power were selected. Mean Decrease Accuracy, representing the degree of decrease in prediction accuracy, was used to assess the importance of the features.

[0069] Figure 9 Receiver operating characteristic (ROC) curves for single gut microbiota factors were plotted. ROC curves were generated for each of the 10 factors, and the area under the ROC curve (AUC) was calculated to determine the factor with the best classification effect.

[0070] Figure 10 Receiver operating characteristic (ROC) curves were plotted for different combinations of gut microbiota factors. ROC curves were generated using different factor combinations, and the area under the ROC curve (AUC) was calculated to determine the factor combination with the best classification effect. In the figure, A represents the Phenylobacterium + 7-Oxocholesterol combination; B represents...

[0071] Phenylobacterium+7-Oxocholesterol+Hydralazine pyruvate hydrazone combination; C represents Phenylobacterium+Latilactobacillus+Pseudactinotalea combination; D represents 7-Oxocholesterol+Pipertipine+Hydralazine pyruvate hydrazone+Nevirapine+Arbutin+Vanilloylglycine+cis-5,8,11,14,17-Eicosapentaenoic acid combination; E represents 7-Oxocholesterol+Hydralazine pyruvatehydrazone+Nevirapine+Arbutin+cis-5,8,11,14,17-Eicosapentaenoic acid combination; F stands for Phenylobacterium+7-Oxocholesterol+Hydralazine pyruvatehydrazone+Nevirapine+Arbutin+cis-5,8,11,14,17-Eicosapentaenoic acid combination. Detailed Implementation

[0072] The present invention will now be described in further detail with reference to specific embodiments. The given embodiments are merely illustrative of the invention and not intended to limit its scope. The embodiments provided below can serve as a guide for further improvements by those skilled in the art and do not constitute a limitation on the invention in any way.

[0073] Unless otherwise specified, the experimental methods used in the following examples are conventional methods, performed according to the techniques or conditions described in the literature in this field or according to the product instructions. Unless otherwise specified, the materials and reagents used in the following examples are commercially available.

[0074] Example 1: A computer device for predicting the immunomodulatory effects of silymarin in treating HIV-immune unresponsive patients.

[0075] I. Patient inclusion criteria and CD4 + T cell changes classification

[0076] 1. Patient enrollment, follow-up, and sample collection

[0077] HIV-infected individuals who have not responded to antiretroviral therapy were screened from the HIV treatment clinic of a tertiary hospital according to the following inclusion criteria: ① Confirmed HIV infection, ② Age over 18 years, Chinese nationality, ③ Received standard antiretroviral therapy for more than 2 years, plasma viral load less than 20 copies / mL, CD4+ + T cell count less than 500 cells / µL; ④ fatty liver or abnormal liver function; ⑥ voluntarily sign informed consent form and guarantee to accept follow-up; Exclusion criteria: ① intellectual disability or inability to communicate effectively with doctors; ② non-Chinese nationality; ③ currently participating in other interventional clinical trials; ④ having taken antibiotics in the two weeks prior to enrollment.

[0078] Patients meeting the above criteria were given oral silymarin for 3 consecutive months (70 mg, three times daily; silymarin capsules, Tianjin Tasly Shengte Pharmaceutical Co., Ltd., National Drug Approval Number H20040299). Demographic information and clinical test results, including T-cell subsets, were collected from 53 patients within one year prior to silymarin treatment. Peripheral blood samples (20 ml) and baseline stool samples (4 g) were collected from patients at baseline and at the 3-month follow-up point.

[0079] The clinical indications for silymarin are abnormal liver function in acute and chronic hepatitis and fatty liver. Only patients with abnormal liver function can be used as target cases in this model. Although HIV-infected individuals or those undergoing long-term antiviral therapy may develop abnormal liver function, this is related to individual differences, viral load, immune status, and co-infections, and not all HIV-infected individuals will develop abnormal liver function. Studies have reported that the incidence of elevated transaminases in HIV-infected individuals is 32%, and the proportion of abnormal liver function decreases and tends to stabilize with the extension of treatment time. Currently, the incidence of HIV immune non-response is 10-40%, so multiple factors such as abnormal liver function, HIV immune non-response, and adherence must be considered. This model included 53 HIV immune non-response patients who were able to undergo standard silymarin treatment and completed a 3-month follow-up.

[0080] 2. Peripheral blood T cell detection and classification of immune response effects

[0081] EDTA-anticoagulated whole blood was prepared from 20 ml of peripheral blood collected from the 53 patients in section 1 above.

[0082] Add 10 μL of CD4 / CD8 / CD3 dye (340499, BD) and 50 μL of EDTA-anticoagulated whole blood to an absolute counting tube (340334, BD), mix well, and incubate at room temperature in the dark for 15 minutes. Add 450 μL of erythrocyte lysis buffer (349202, BD), mix well, and incubate in the dark for 15 minutes. Use flow cytometry to detect the absolute number of peripheral blood T cell subsets.

[0083] Peripheral blood mononuclear cells (PBMCs) were isolated using Ficoll density gradient centrifugation. The following dyes were used: BD PerCPCD3 (552851), FITC CD4 (566911), PE CD38 (555460), PE-Cy7 CD8 (557750), APC CD25 (555434), PE CD127 (557938), PE CD194 (551120), APC CD196 (560619), BV421 CD183 (562558), PE-Cy7 CD45RA (560675), Pacific Blue CD4 (558116), APC CD45RO (559865), FITC CD57 (555619), PE CD28 (555729), APC-Cy7 CD8 (557760), APC HLA-DR (5599868), and Fixable. Viability Stain 700 (564997) antibody detection of T cell activation, senescence, differentiation and other phenotypes.

[0084] The results are as follows Figures 2-4As shown, a comparison of 53 HIV-infected individuals who received 3 months of standard silymarin treatment revealed that, overall, CD4 levels decreased after treatment. + The number of T cells increased significantly, while the level of T cell activation decreased significantly.

[0085] Further based on CD4 + Regarding T cell changes, the 53 infected individuals could be clearly divided into two groups: an immunomodulatory responsive group (SIL-R, n=27) and an immunomodulatory non-responsive group (SIL-NR, n=26), suggesting that silymarin can improve incomplete immune reconstitution or modulate the immune system (specifically by increasing CD4 counts). + (T cell count), but there is heterogeneity in treatment efficacy among individuals.

[0086] According to CD4 + Changes in T cells can be categorized into two types:

[0087] The silymarin-treated immunomodulatory responsive group (SIL-R) was defined as the group with significantly increased CD4 counts after 3 months of silymarin treatment compared to before treatment. + The number of T cells increased by 50 or 20% compared to before treatment;

[0088] The silymarin-treated immunomodulatory non-responsive group (SIL-NR) was defined as the group with significantly lower CD4 counts after 3 months of silymarin treatment compared to before treatment. + The number of T cells did not increase or increased by less than 50 cells compared with that before treatment, or the increase was less than 20%.

[0089] II. Determination of factors predicting the immunomodulatory effects of silymarin treatment in HIV-immune unresponsive patients

[0090] 1. Analysis of fecal bacterial diversity

[0091] DNA was extracted from feces collected from 53 patients in section 1 above using the CTAB method. Primers 341F (5'-CCTACGGGNGGCWGCAG-3') and 805R (5'-GACTACHVGGGTATCTAATCC-3') were used to amplify the V3-V4 regions of 16S rDNA. Band size was detected by 2% agarose gel electrophoresis, and PCR products were quantified using a UV spectrophotometer. PCR products were purified using AMPure XT beads (Beckman) and quantified using Qubit 2.0. Sequencing adapters were added to the ends of the PCR products using the TruSeq™ DNA Sample Prep Kit (Illunina). Library fragment size was quality controlled using an Agilent 2100, and the library was quantified using quantitative real-time PCR. Amplicon libraries were mixed in equimolar amounts and paired-end sequencing was performed using a Novaseq sequencer with a PE250 sequencer. First, the raw data is split based on the barcode information, and connectors and barcode sequences are removed. The paired reads are then concatenated using overlap regions to obtain contigs. Quality control, removal of sequences shorter than 100bp, and chimera filtering are performed according to Q20 to obtain high-quality "CleanData". "DADA2" is then used for length filtering and noise reduction to obtain the ASV abundance table and feature sequences.

[0092] Bacterial classification and annotation were performed using the SILVA (Release 138) and NT-16S databases based on ASV characteristic sequence files. The abundance of each species in each sample was statistically analyzed using the ASV abundance table, with a confidence level greater than 0.7. Differential bacterial analysis was performed using the Mann-Whitney U test, linear discriminant analysis, and Lefse analysis.

[0093] 16S rDNA results as follows Figure 5 and Figure 6As shown, the gut bacterial composition differs between the immune response group (referred to as the silymarin-treated immunomodulatory responsive group, SIL-R, n=27) and the immune non-response group (referred to as the silymarin-treated immunomodulatory unresponsive group, SIL-NR, n=26). Specifically, at the genus level, the two groups show differences in the composition of Prevotella, Escherichia-Shigella, Megamonas, Dialisterella, Ruminococcus, and Streptococcus. The dominant genera were *Streptococcus*, *Bifidobacterium*, *Bifidobacterium*, *Bifidobacterium*, and *Fusobacterium*. However, compared to the SIL-NR group, the SIL-R group had higher abundances of *Streptococcus*, *Bifidobacterium*, *Allprevotella*, and *Rombotusia*, while having lower abundances of *Akkermansia*. Combined statistical and LefSe differential analysis revealed statistical differences in 13 bacteria at the genus level, with 8 enriched in the SIL-R group and 5 enriched in the SIL-NR group.

[0094] 2. Fecal non-targeted metabolomics analysis

[0095] Methanol extraction of metabolites from ex vivo fecal samples: Fecal samples collected from 49 patients (with sufficient samples) in step 1 above were thawed on ice. 100 mg of sample was transferred to a 2 ml EP tube, and 1 ml of pre-cooled 50% methanol aqueous solution was added. The tube was vortexed for 1 minute, allowed to stand at room temperature for 10 minutes, and then incubated at -20°C overnight. After protein precipitation, the extract was centrifuged at 4000g for 20 minutes. The supernatant was transferred to a new 96-well plate as the test sample. Simultaneously, 10 μL of each test sample was mixed to prepare a quality control (QC) sample. All test samples were stored frozen at -80°C before loading.

[0096] Reversed-phase separation: The supernatant (sample) obtained above was subjected to reversed-phase separation using an UltiMate 3000 UPLC system (ThermoFisher Scientific) with an ACQUITY UPLC T3 column (100mm*2.1mm, 1.8μm, Waters). The column temperature was maintained at 40℃. 5mM ammonium acetate and 5mM acetic acid were added, along with mobile phase B (acetonitrile). The flow rate was 0.3ml / min. The gradient elution settings were: 0-0.8 min, 2% mobile phase B; 0.8-2.8 min, 2%-70% mobile phase B; 2.8-5.6 min, 70-90% mobile phase B; 5.6-6.4 min, 90-100% mobile phase B; 6.4-8 min, 100% mobile phase B; 8.0-8.1 min, 100-2% mobile phase B; 8.1-10 min, 2% mobile phase B.

[0097] Mass spectrometry analysis was performed using Q-Exative (Thermo Fisher Scientific) in both positive and negative ion modes: ion precursor spectra were collected at a resolution of 70,000 m / s (70–1050 m / s), with an AGC of 3 × 10⁻⁶ m / s. 6 The maximum injection time was set to 100ms; the top 3 data acquisition settings were configured in DDA mode; fragment spectra were acquired at 17500 resolution, and AGC reached 1*10. 5 The maximum injection time was set to 80 ms. MSConver software was used to convert the raw LC-MS data into mzXML format. R software was used to run the XCMS toolkit for peak extraction and retention time correction, the CAMERA toolkit for additive ion analysis, and the metaX toolkit for ion identification based on retention time and m / z data. Metabolites were matched and annotated using the online KEGG (Kyoto Encyclopedia of Genes and Genomes), HMDB (Human Metabolome Database), and in-house databases. Univariate analysis was used for fold-change and t-test statistical tests, combined with multivariate statistical analysis of the VIP (Variable Important for the Projection) value obtained from PLS-DA. Differential metabolites were screened based on P < 0.05, fold change > 2, and VIP > 1, and the relative abundance (peak intensity in the mass spectrum) of each metabolite was obtained.

[0098] The results are as follows Figure 7As shown, compared with the SIL-NR group, the SIL-R group had 105 significantly upregulated and 78 significantly downregulated metabolites, of which 19 were annotable (10 upregulated and 9 downregulated). The 10 upregulated metabolites were mainly classified as organosulfur compounds, benzene compounds, organooxygen compounds, organoheterocyclic compounds and lipid molecules in the HMDB secondary classification, while the 9 downregulated metabolites were mainly classified as organonitrogen compounds in the secondary classification.

[0099] 3. Factor determination for the classification model

[0100] The random forest algorithm was used to screen the top 10 factors with the highest predictive power from the differentially expressed bacteria obtained in step 1 and the differentially expressed metabolites obtained in step 2. Figure 8 The study includes 7 metabolites and 3 bacteria: Phenylobacterium (denoted as g_Phenylobacterium in the figure), Latilactobacillus (denoted as g_Latilactobacillus in the figure), and Pseudactinotalea (denoted as g_Pseudactinotalea in the figure), and the metabolites 7-Oxocholesterol, Pipertipine, Hydrolarazine pyruvate hydrazone, Nevirapine, Arbutin, Vanilloylglycine, and cis-5,8,11,14,17-Eicosapentaenoic acid.

[0101] Table 1 lists the structural formulas, CAS numbers, and commercial sources of the seven metabolites.

[0102]

[0103]

[0104]

[0105] III. Establishment of a model for predicting the immunomodulatory effect of silymarin treatment on HIV immune-unresponsive patients undergoing antiviral therapy. Figure 1 Flowchart for a model to predict the efficacy of adjuvant immunotherapy for HIV infection.

[0106] 1. Detection factor eigenvalues

[0107] The characteristic values ​​of 10 factors from 3 categories in the above two categories were detected in the ex vivo feces of 53 patients in the above category 1: characteristic values ​​of 3 bacteria (from 53 patients) and characteristic values ​​of 7 metabolites (from 49 patients);

[0108] The characteristic value of the above three bacteria is the abundance of the three bacteria (ASV abundance);

[0109] The method for detecting ASV abundance in bacteria is as follows:

[0110] All DNA was extracted from ex vivo fecal samples and used as a template. The V3-V4 region of the bacterial 16S rDNA was amplified using primers 341F (5'-CCTACGGGNGGCWGCAG-3') and 805R (5'-GACTACHVGGGTATCTAATCC-3'). After sequencing, length filtering and noise reduction were performed, and the abundance of ASV was calculated.

[0111] The characteristic value of the seven metabolites is the content of the seven metabolites;

[0112] The detection methods for the above metabolite content are as follows:

[0113] Metabolites were extracted from isolated fecal samples using methanol according to method 2.1. A Thermo Fisher Scientific UltiMate 3000 UPLC system was used for reversed-phase separation with an ACQUITY UPLC T3 column (100 mm * 2.1 mm, 1.8 μm, Waters). The column temperature was maintained at 40 °C. 5 mM ammonium acetate and 5 mM acetic acid were added, along with mobile phase B (acetonitrile). The flow rate was 0.3 mL / min. The gradient elution settings were: 0–0.8 min, 2% mobile phase B; 0.8–2.8 min, 2%–70% mobile phase B; 2.8–5.6 min, 70–90% mobile phase B; 5.6–6.4 min, 90–100% mobile phase B; 6.4–8 min, 100% mobile phase B; 8.0–8.1 min, 100–2% mobile phase B; 8.1–10 min, 2% mobile phase B. A Q-Exative (Thermo Fisher Scientific) mass spectrometer was used in both positive and negative ion modes (collecting ion precursor spectra (70-1050 m / z) at a resolution of 70,000, with an AGC of 3*10^6 m / z). 6 The maximum injection time was set to 100ms; the top 3 data acquisition settings were configured in DDA mode; fragment spectra were acquired at 17500 resolution, and AGC reached 1*10. 5 The maximum injection time was set to 80ms. The contents of 7 metabolites were detected and calculated (the relative contents of each metabolite are expressed by the intensity of the peaks in the mass spectrum).

[0114] 2. A model for predicting the modulatory effect of silymarin on the immune response in HIV-immune unresponsive patients treated with antiviral therapy was developed based on a combination of features.

[0115] 1) Constructing a prediction model using the eigenvalues ​​of 10 factors

[0116] The eigenvalues ​​of the 10 factors for each patient mentioned above and the known immune response efficacy of silymarin treatment (CD4) for each sample were combined. + The immune response group (SIL-R) and the immune non-response group (SIL-NR) differentiated by T cell changes were used as input data for modeling. The random forest (RF) computation method was used to train the model and output the predicted probability value of the immune regulation effect of silymarin treatment.

[0117] The characteristic values ​​of 10 factors for 53 patients are shown in Table 2:

[0118] Table 2 shows the eigenvalues ​​of 10 factors for 53 patients and the predicted probability values ​​of the prediction model constructed from these eigenvalues.

[0119]

[0120]

[0121]

[0122]

[0123] In the table above, factors 1-10 represent Phenylobacterium, Latilactobacillus, Pseudactinotalea, metabolite 7-Oxocholesterol, Pipertipine, Hydrolarazine pyruvatehydrazone, Nevirapine, Arbutin, Vanilloylglycine, and cis-5,8,11,14,17-Eicosapentaenoic acid, respectively.

[0124] In the table above, the values ​​of factors 1-3 correspond to the ASV abundances of Phenylobacterium, Latilactobacillus, and Pseudactinotalea, respectively, while the values ​​of factors 4-10 correspond to the relative contents (specifically, the peak intensities in the mass spectra) of 7-Oxocholesterol, Pipertipine, Hydrolarazine pyruvate hydrazone, Nevirapine, Arbutin, Vanilloylglycine, and cis-5,8,11,14,17-Eicosapentaenoic acid.

[0125] Random Forest Model: This study uses the random forest algorithm to train the input data and outputs the predicted probability value of the immunomodulatory effect of silymarin treatment. The optimized parameters of the model were determined through grid search. The key parameters finally selected include: the number of decision trees (nestimators) is 500, the maximum tree depth (max_depth) is 5, the minimum number of samples per split node (min_samples_split) is 2, the minimum number of samples per leaf node (min_samples_leaf) is 1, and the splitting criterion is calculated based on the Gini coefficient (criterion = gini). The model was trained using ten-fold cross-validation with ten repetitions to ensure the robustness and generalization ability of the model.

[0126] Based on the predicted probability value of the immunomodulatory effect of silymarin treatment for each sample, receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) value was calculated (AUC = 1, Figure 10 The best classification effect is achieved when using all (or all).

[0127] The threshold for this corresponding prediction model is 0.500;

[0128] If the predicted probability value of a subject is greater than the set threshold, then the subject is considered to be or a candidate for silymarin treatment with an immune-modulated response (SIL-R').

[0129] If the predicted probability value of a subject is less than or equal to a set threshold, then the subject is or is a candidate for silymarin treatment-immunomodulatory non-responsive (SIL-NR').

[0130] 2) Constructing prediction models using eigenvalues ​​of individual metabolite factors

[0131] Unlike 1) above, the only difference is that the input data consists of the characteristic values ​​of individual metabolite factors from the 53 patients in 1) above, and the known immunomodulatory effects of silymarin treatment (CD4+) for each sample. + Using the T cell changes as input data, predictive models for different metabolite factors were constructed. (The immune regulation responsive group SIL-R and the immune regulation non-responsive group SIL-NR were distinguished.)

[0132] Based on the predictive models of different metabolite factors, the predicted probability values ​​of silymarin treatment immune response for each sample were used to plot receiver operating characteristic (ROC) curves and calculate the area under the curve (AUC) value. Figure 9 ),

[0133] The results are as follows:

[0134] The prediction model constructed with 7-Oxocholesterol has an AUC (area under the curve) value of 0.78.

[0135] The prediction model built by Pipertipine has an AUC (area under the curve) value of 0.75.

[0136] The prediction model constructed with hydralazine pyruvate hydrazone has an AUC (area under the curve) value of 0.65.

[0137] The prediction model built by Nevirapine has an AUC (area under the curve) value of 0.62.

[0138] The prediction model constructed by Arbutin has an AUC (area under the curve) value of 0.64.

[0139] The prediction model constructed by Vanilloylglycine has an AUC (area under the curve) value of 0.70.

[0140] The prediction model constructed using cis-5,8,11,14,17-Eicosapentaenoic acid had an AUC (area under the curve) value of 0.65.

[0141] 3) Constructing a predictive model using eigenvalues ​​of metabolite factor combinations.

[0142] Unlike 1) above, the only difference is that the input data is a combination of the characteristic values ​​of metabolite factors from the 53 patients in 1) above, and the known immunomodulatory effects of silymarin treatment (CD4) of each sample. + Using the T cell changes as input data, a predictive model for different combinations of metabolite factors was constructed. (The model distinguishes between the immune-responsive group SIL-R and the immune-unresponsive group SIL-NR.)

[0143] Different combinations of metabolite factors include combinations of all 7 metabolites and combinations of 7-Oxocholesterol, Hydrolazinepyruvate hydrazone, Nevirapine, Arbutin, and cis-5,8,11,14,17-Eicosapentaenoic acid.

[0144] Based on the predicted probability values ​​of silymarin treatment immune response for each sample in the prediction model with different combinations of metabolite factors, receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) value was calculated.

[0145] The results are as follows:

[0146] (1) Predictive models constructed from all 7 metabolite combinations ( Figure 10 The AUC (area under the curve) value is 0.96.

[0147] The predicted probability values ​​of the immune regulation effect given by the corresponding prediction model are shown in Table 3.

[0148] Table 3 shows the predicted probability values ​​of the prediction model constructed from the combination of 7 metabolite characteristics of 53 patients.

[0149]

[0150] The threshold for this corresponding prediction model is 0.367;

[0151] If the predicted probability value of a subject is greater than the set threshold, then the subject is considered to be or a candidate for silymarin treatment with an immune-modulated response (SIL-R').

[0152] If the predicted probability value of a subject is less than or equal to a set threshold, then the subject is or is a candidate for silymarin treatment-immunomodulatory non-responsive (SIL-NR').

[0153] (2) A predictive model constructed using a combination of 7-Oxocholesterol, Hydrolarazine pyruvate hydrazone, Nevirapine, Arbutin, and cis-5,8,11,14,17-Eicosapentaenoic acid. Figure 10 The area under the curve (AUC) value is 0.93.

[0154] The predicted probability values ​​of the immune regulation effect given by the corresponding prediction model are shown in Table 4.

[0155] Table 4 shows the predicted probability values ​​of the prediction model constructed based on the combination of five metabolite characteristics in 53 patients.

[0156]

[0157] The threshold for this corresponding prediction model is 0.513;

[0158] If the predicted probability value of a subject is greater than the set threshold, then the subject is considered to be or a candidate for silymarin treatment with an immune-modulated response (SIL-R').

[0159] If the predicted probability value of a subject is less than or equal to a set threshold, then the subject is or is a candidate for silymarin treatment-immunomodulatory non-responsive (SIL-NR').

[0160] 4) Constructing a prediction model using the feature values ​​of a single bacterium.

[0161] Unlike 1) above, the only difference is that the input data consists of the characteristic values ​​of individual bacteria from the 53 patients in 1 above and the known immunomodulatory effects of silymarin treatment (CD4+) of each sample. +Using the T cell changes (differentiating between the immune-responsive group SIL-R and the immune-unresponsive group SIL-NR) as input data for modeling, a predictive model for a single bacterium was constructed.

[0162] Based on the predicted probability value of silymarin treatment immune response for each sample in the prediction model for a single bacterium, receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) value was calculated.

[0163] The results are as follows Figure 9 :

[0164] The prediction model constructed using Phenylobacterium (denoted as g_Phenylobacterium in the figure) has an AUC (area under the curve) value of 0.65.

[0165] The prediction model constructed using Latilactobacillus (denoted as g_Latilactobacillus in the figure) has an AUC (area under the curve) value of 0.58.

[0166] The prediction model constructed by Pseudactinotalea (denoted as g_Pseudactinotalea in the figure) has an AUC (area under the curve) value of 0.65.

[0167] 5) Constructing a prediction model using the eigenvalues ​​of bacterial combinations.

[0168] Unlike 1) above, the only difference is that the input data consists of the characteristic values ​​of three bacteria from 53 patients in 1 above, and the known immunomodulatory effects of silymarin treatment (CD4+) of each sample. + Using the T cell changes (differentiating between the immune-responsive group SIL-R and the immune-unresponsive group SIL-NR) as input data for modeling, predictive models for the three bacteria were constructed.

[0169] Based on the predicted probability values ​​of silymarin treatment immune response for each sample in the prediction model for the three bacteria, receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) value was calculated.

[0170] The results are as follows: The prediction model constructed from the three bacteria had an AUC (area under the curve) value of 0.82. Figure 10 C);

[0171] The predicted probability values ​​of the immune regulation effect given by the corresponding prediction model are shown in Table 5.

[0172] Table 5 shows the predicted probability values ​​of the prediction model constructed from the three bacterial feature combinations in 53 patients.

[0173]

[0174] The threshold for this prediction model is 0.485;

[0175] If the predicted probability value of a subject is greater than the set threshold, then the subject is considered to be or a candidate for silymarin treatment with an immune-modulated response (SIL-R').

[0176] If the predicted probability value of a subject is less than or equal to a set threshold, then the subject is or is a candidate for silymarin treatment-immunomodulatory non-responsive (SIL-NR').

[0177] 6) Constructing a predictive model using eigenvalues ​​of metabolites and bacterial combinations.

[0178] Unlike 1) above, the only difference is that the input data respectively includes the characteristic values ​​of 3 bacteria and 7 metabolites from 53 patients in 1 above, and the known immunomodulatory effects of silymarin treatment (CD4) of each sample. + Using the T cell changes (differentiating between the immunomodulatory responsive group SIL-R and the immunomodulatory non-responsive group SIL-NR) as input data for modeling, a predictive model of bacterial and metabolite combinations was constructed.

[0179] Based on the predicted probability values ​​of silymarin treatment immune response for each sample in the predictive model of bacteria and metabolites, receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) value was calculated.

[0180] The results are as follows:

[0181] (1) The prediction model constructed by combining Phenylobacterium and 7-Oxocholesterol had an AUC (area under the curve) value of 0.87. Figure 10 A);

[0182] The predicted probability values ​​of the immune regulation effect given by the corresponding prediction model are shown in Table 6.

[0183] Table 6 shows the predicted probability values ​​of the prediction model constructed from the combination of one bacterial and one metabolite feature in 53 patients.

[0184]

[0185] The threshold for this corresponding prediction model is 0.504;

[0186] If the predicted probability value of a subject is greater than the set threshold, then the subject is considered to be or a candidate for silymarin treatment with an immune-modulated response (SIL-R').

[0187] If the predicted probability value of a subject is less than or equal to a set threshold, then the subject is or is a candidate for silymarin treatment-immunomodulatory non-responsive (SIL-NR').

[0188] (2) The prediction model constructed using a combination of Phenylobacterium, 7-Oxocholesterol, and Hydrolarazine pyruvate hydrazone had an AUC (area under the curve) value of 0.92. Figure 10 B);

[0189] The predicted probability values ​​of the immune regulation effect given by the corresponding prediction model are shown in Table 7.

[0190] Table 7 shows the predicted probability values ​​of the prediction model constructed from the combination of one bacterium and two metabolite features in 53 patients.

[0191]

[0192]

[0193] The threshold for this corresponding prediction model is 0.270;

[0194] If the predicted probability value of a subject is greater than the set threshold, then the subject is considered to be or a candidate for silymarin treatment with an immune-modulated response (SIL-R').

[0195] If the predicted probability value of a subject is less than or equal to a set threshold, then the subject is or is a candidate for silymarin treatment-immunomodulatory non-responsive (SIL-NR').

[0196] (3) The prediction model constructed using a combination of Phenylobacterium, 7-Oxocholesterol, Hydrolarazine pyruvate hydrazone, Nevirapine, Arbutin, and cis-5,8,11,14,17-Eicosapentaenoic acid had an AUC (area under the curve) value of 0.98. Figure 10 (F);

[0197] The predicted probability values ​​of the immune regulation effect given by the corresponding prediction model are shown in Table 8.

[0198] Table 8 shows the predicted probability values ​​of the prediction model constructed from the combination of one bacterium and five metabolite features in 53 patients.

[0199]

[0200] The threshold for this corresponding prediction model is 0.505;

[0201] If the predicted probability value of a subject is greater than the set threshold, then the subject is considered to be or a candidate for silymarin treatment with an immune-modulated response (SIL-R').

[0202] If the predicted probability value of a subject is less than or equal to a set threshold, then the subject is or is a candidate for silymarin treatment-immunomodulatory non-responsive (SIL-NR').

[0203] Therefore, a model with an AUC value greater than 0.8 was selected as the model for predicting the modulating effect of silymarin treatment on the immune response in HIV-immune unresponsive patients undergoing antiviral therapy. Specifically:

[0204] A predictive model was constructed using the eigenvalues ​​of 10 factors; the sensitivity and specificity were both 100%.

[0205] The prediction model was constructed using all seven metabolite combinations; its sensitivity was 84% ​​and its specificity was 96%.

[0206] A predictive model constructed using the combination of 7-Oxocholesterol, Hydrolazine pyruvate hydrazone, Nevirapine, Arbutin, and cis-5,8,11,14,17-Eicosapentaenoic acid was used; the sensitivity was 88% and the specificity was 83%.

[0207] A predictive model constructed using a combination of three bacteria showed a sensitivity of 96% and a specificity of 67%.

[0208] The prediction model constructed using the combination of Phenylobacterium and 7-Oxocholesterol had a sensitivity of 92% and a specificity of 75%.

[0209] The prediction model constructed using the combination of Phenylobacterium, 7-Oxocholesterol, and Hydrolarazine pyruvate hydrazone had a sensitivity of 76% and a specificity of 92%.

[0210] A predictive model was constructed using a combination of Phenylobacterium, 7-Oxocholesterol, Hydrolarazine pyruvate hydrazone, Nevirapine, Arbutin, and cis-5,8,11,14,17-Eicosapentaenoic acid. The sensitivity was 92%, and the specificity was 96%.

[0211] The present invention has been described in detail above. For those skilled in the art, the invention can be practiced in a wide range of ways with equivalent parameters, concentrations, and conditions without departing from its spirit and scope, and without requiring unnecessary experiments. Although specific embodiments have been given, it should be understood that further modifications can be made to the invention. In summary, according to the principles of the invention, this application is intended to include any changes, uses, or improvements to the invention, including changes made using conventional techniques known in the art that depart from the scope disclosed herein. Some of the essential features can be applied within the scope of the following appended claims.

Claims

1. A data processing apparatus, comprising a memory, a processor, and a computer program stored in the memory, characterized in that: The processor executes the computer program to perform the following steps: S1. Receiving data: Receiving sample data, wherein the sample data is the feature value of any one or any combination of 10 factors in the subject's ex vivo fecal sample; The 10 factors include 7 metabolites and 3 bacteria; The three bacteria are Phenylobacterium, Latilactobacillus, and Pseudactinotale; The seven metabolites are 7-Oxocholesterol, Pipertipine, Hydrolarazine pyruvatehydrazone, Nevirapine, Arbutin, Vanilloylglycine, and cis-5,8,11,14,17-Eicosapentaenoic acid. The characteristic value of each bacterium is its abundance. The characteristic value of each metabolite is its metabolite content; S2. Input data: Input the sample data into the immunomodulatory effect prediction model for silymarin treatment of HIV immune non-responsive patients; The immunomodulatory effect prediction model for silybin treatment in HIV-unresponsive patients is constructed according to the following steps: using the feature values ​​of the 10 factors or any combination thereof in the ex vivo fecal samples of subjects who are known to have responded to silybin treatment after antiviral therapy and subjects who are known to have not responded to silybin treatment after antiviral therapy, as training samples, to train the immunomodulatory effect prediction model for silybin treatment in HIV-unresponsive patients. The immunomodulatory effect is defined as either a response to immunomodulation or a lack of response to immunomodulation. S3. Output Results: The immunomodulatory effect prediction model for silymarin treatment in HIV immune non-responsive patients is used to output the probability value of the subject predicting the immunomodulatory effect of silymarin treatment; and the immunomodulatory effect of silymarin treatment in the subject is then calculated based on the probability value. The subjects were HIV-immune non-response patients after antiviral treatment.

2. A computer program product, including a computer program, characterized in that: When the computer program is executed by the processor, it performs the steps in the data processing apparatus of claim 1.

3. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program that causes a computer to perform the steps in the data processing apparatus of claim 1.

4. A device for predicting the immunomodulatory effect of silymarin adjuvant therapy in HIV immune-unresponsive patients, the device comprising: S1, Data receiving module: used to receive sample data, wherein the sample data is the characteristic values ​​of 10 factors or any one or any combination thereof in the ex vivo fecal sample of the subject; The 10 factors include 7 metabolites and 3 bacteria; The three bacteria are Phenylobacterium, Latilactobacillus, and Pseudactinotale; The seven metabolites are 7-Oxocholesterol, Pipertipine, Hydrolarazine pyruvatehydrazone, Nevirapine, Arbutin, Vanilloylglycine, and cis-5,8,11,14,17-Eicosapentaenoic acid. The characteristic value of each bacterium is its abundance. The characteristic value of each metabolite is its metabolite content; S2, Data Input Module: Used to input the sample data into the immunomodulatory effect prediction model for silymarin treatment of HIV immune non-responders; The immunomodulatory effect prediction model for silybin treatment in HIV-unresponsive patients is constructed according to the following steps: using the feature values ​​of the 10 factors or any combination thereof in the ex vivo fecal samples of subjects who are known to have responded to silybin treatment after antiviral therapy and subjects who are known to have not responded to silybin treatment after antiviral therapy, as training samples, to train the immunomodulatory effect prediction model for silybin treatment in HIV-unresponsive patients. The immunomodulatory effect is defined as either a response to immunomodulation or a lack of response to immunomodulation. S3. Result Output Module: Used to output the probability value of the subject predicting the immunomodulatory effect of silymarin treatment in HIV immune non-responders through the immunomodulatory effect prediction model; and then calculate the immunomodulatory effect of silymarin treatment in the subject based on the probability value. The subjects were HIV-immune non-response patients after antiviral treatment.

5. A method for predicting or assisting in predicting the immunomodulatory effects of silymarin treatment on subjects, characterized in that: The method includes the following steps: S1. Obtain the feature values ​​of any one or any combination of 10 factors in the isolated fecal sample of the subject; The 10 factors include 7 metabolites and 3 bacteria; The three bacteria are Phenylobacterium, Latilactobacillus, and Pseudactinotale; The seven metabolites are 7-Oxocholesterol, Pipertipine, Hydrolarazine pyruvatehydrazone, Nevirapine, Arbutin, Vanilloylglycine, and cis-5,8,11,14,17-Eicosapentaenoic acid. The characteristic value of each bacterium is its abundance. The characteristic value of each metabolite is its metabolite content; S2. Input the sample data into the immunomodulatory effect prediction model for silymarin treatment of HIV immune non-responders; The immunomodulatory effect prediction model for silybin treatment in HIV-unresponsive patients is constructed according to the following steps: using the feature values ​​of the 10 factors or any combination thereof in the ex vivo fecal samples of subjects who are known to have responded to silybin treatment after antiviral therapy and subjects who are known to have not responded to silybin treatment after antiviral therapy, as training samples, to train the immunomodulatory effect prediction model for silybin treatment in HIV-unresponsive patients. The immunomodulatory effect is defined as either a response to immunomodulation or a lack of response to immunomodulation. S3. The immunomodulatory effect prediction model for silymarin treatment in HIV immune non-responsive patients is used to output the probability value of the subject predicting the immunomodulatory effect of silymarin treatment; and the immunomodulatory effect of silymarin treatment in the subject is then calculated based on the probability value. The subjects were HIV-immune non-response patients after antiviral treatment.

6. A method for constructing a predictive model of the immunomodulatory effect of silymarin in treating HIV immune-unresponsive patients, characterized in that: The method includes the following steps: using the feature values ​​of the 10 factors or any combination thereof in the ex vivo fecal samples of subjects who are known to be HIV immune non-responsive after antiviral treatment and who are known to be HIV immune non-responsive after antiviral treatment and who are not responsive to silybin treatment for immunomodulation as training samples, to train an immunomodulatory effect prediction model for silybin treatment in HIV immune non-responsive patients. The immunomodulatory effect is defined as either a positive or negative immune response.

7. The data processing apparatus according to claim 1, the computer program product according to claim 2, the computer-readable storage medium according to claim 3, the apparatus according to claim 4, or the method according to claim 5 or 6, characterized in that: Any one or any combination of the 10 factors is as follows: The 10 factors mentioned in A1); The seven metabolites mentioned in A2); A3)7-Oxocholesterol, Hydralazine pyruvate hydrazone, Nevirapine, Arbutin and cis-5,8,11,14,17-Eicosapentaenoic acid; The three bacteria mentioned in A4); A5)Phenylobacterium and 7-Oxocholesterol; A6) Phenylobacterium, 7-Oxocholesterol and Hydralazine pyruvate hydrazone; A7) Phenylobacterium, 7-Oxocholesterol, Hydralazine pyruvate hydrazone, Nevirapine, Arbutin and cis-5,8,11,14,17-Eicosapentaenoic acid.

8. Application of silymarin in the preparation of products for restoring immunomodulatory responses in HIV-insensitive individuals; Alternatively, silymarin may be used in the preparation of CD4 cells in HIV-inhibiting or immune-enhancing individuals. + Application in T-cell count products.

Images