Intestinal flora marker model for predicting the efficacy of drug immunotherapy and its construction method

Abstract

The present invention provides a combination of intestinal flora markers for predicting the efficacy of immunotherapy with immune checkpoint inhibitors anti-PD-1 and / or PD-L1 antibodies, wherein the intestinal flora markers are selected from 16 bacterial targets. The present invention also provides a kit for predicting the efficacy of immunotherapy with immune checkpoint inhibitors anti-PD-1 and / or PD-L1 antibody drugs, the kit containing reagents for detecting 16 intestinal flora markers, and calculating the immunotherapy benefit score Response score (RS) based on the model microbial abundance, and using the Youden index to calculate the cutoff value of the efficacy benefit score (RS=0.445071). If the patient's treatment benefit score is greater than 0.445071, it indicates that the patient belongs to a group with a high immunotherapy benefit rate, otherwise it belongs to a group with a low immunotherapy benefit rate. The present invention constructs a microbial prediction model related to immunotherapy response, and the benefit score calculated according to the model can predict the efficacy of immunotherapy in patients receiving immunotherapy.

Description

Technical Field

[0001] The present invention belongs to the field of bioengineering and relates to an intestinal flora marker model, specifically an intestinal flora marker model for predicting the efficacy of immunotherapy with immune checkpoint inhibitors PD-1 and / or PD-L1 drugs and a method for constructing the same. Background Art

[0002] With the rapid development of tumor immunology, the application of tumor immunotherapy, especially immune checkpoint inhibitors (ICIs), has reshaped the new paradigm of cancer treatment, providing new treatment options and disease relief for many patients with refractory advanced tumors. Traditional ICIs, including anti-cytotoxic T lymphocyte-associated antigen 4 (CTLA-4) and checkpoint inhibitors targeting programmed death receptor 1 (PD-1) / programmed death receptor-ligand 1 (PD-L1), have been approved for the treatment of various cancers, such as melanoma and non-small cell lung cancer.

[0003] However, studies have found that the inhibition rate of ICIs against solid tumors is only 10% to 40%, indicating that a large proportion of patients do not benefit from immunotherapy. Therefore, finding predictive biomarkers that determine the clinical efficacy of ICIs and designing reasonable combination treatment plans of ICIs with other therapeutic drugs are the keys to enhancing the overall efficacy of ICIs.

[0004] A growing body of research indicates that the gut microbiome plays a crucial role in tumor development and progression and can synergize with ICIs to enhance anti-tumor efficacy. Supplementation with specific probiotics can enhance the efficacy of immune checkpoint inhibitors in tumor-bearing mice. Fecal microbiota transplantation (FMT) has been shown to enhance the efficacy of immune checkpoint inhibitors in preclinical studies and some phase I clinical trials. Therefore, using the gut microbiome as a novel biomarker to assess patient benefit from ICIs holds great promise. Recently, several studies have proposed disease-specific prediction models and ICI efficacy based on the gut microbiome. A set of baseline gut microbiome biomarkers can predict patients who are likely to benefit most from ICI treatment. However, these retrospective data analyses have primarily used microbiome 16S rRNA sequencing, relatively conventional analytical methods, and a small number of subjects. The development of efficient and reproducible microbiome prediction models remains a work in progress. Moreover, due to the diversity of intestinal flora, there are large differences in the characteristic disease-related microbial communities screened out in different research cohorts. This is to some extent limited by factors such as the number of samples in different research cohorts, the source of patients, and the sequencing platform and sequencing strategy. Comprehensive analysis based on larger-scale and more accurate microbial data will have important clinical significance, so that clinicians can formulate treatment plans that maximize patient benefits based on the individual differences of cancer patients.

[0005] Fecal gut microbiome quantification technologies, including high-throughput sequencing (16S rRNA sequencing and shotgun metagenomic sequencing) and real-time quantitative PCR, can specifically measure the relative abundance of the gut microbiome in feces. With the widespread application of high-throughput technologies, their efficiency, wide reach, and comprehensive microbiome coverage allow for rapid analysis of gut microbiome abundance across multiple samples. 16S rRNA sequencing primarily relies on PCR-amplified bacterial 16S rRNA fragments, focusing on studying species composition, evolutionary relationships between species, and community diversity. Shotgun metagenomic sequencing, on the other hand, captures the complete gene sequence of a microbial community. By assembling these sequences, it provides structural and functional information about the genome, primarily focusing on the composition and function of microbial communities. Compared to 16S rRNA sequencing, metagenomic sequencing demonstrates greater accuracy in identifying microbial community structure and function at the species level. Understanding the composition of the gut microbiome using high-throughput sequencing is essential for scientific research and clinical application in linking the gut microbiome to disease. Integrating metagenomic sequencing data from multiple research cohorts and conducting specific microbial community screening and analysis can help develop more efficient microbial prediction models to identify populations that benefit from immune checkpoint inhibitors and assist clinicians in developing personalized treatment plans. Summary of the Invention

[0006] In response to the above-mentioned technical problems in the prior art, the present invention provides intestinal flora markers for predicting the efficacy of immunotherapy with immune checkpoint inhibitors anti-PD-1 and / or PD-L1 antibody drugs and a method for constructing the same. The intestinal flora markers for predicting the efficacy of immunotherapy with immune checkpoint inhibitors PD-1 and / or PD-L1 drugs and a method for constructing the same are intended to solve the technical problem that the prior art cannot timely predict the immunotherapy effect of tumors.

[0007] The present invention provides a combination of intestinal flora markers for predicting the efficacy of immunotherapy with immune checkpoint inhibitors PD-1 and / or PD-L1 antibody drugs, consisting of sixteen bacterial targets, which are:

[0008]

[0009]

[0010] The present invention also provides a kit for predicting the efficacy of immunotherapy with immune checkpoint inhibitor anti-PD-1 and / or PD-L1 antibody drugs, the kit containing reagents for detecting the following intestinal flora markers: Akkermansiamuciniphila, Bifidobacterium bifidum, [Eubacterium]hallii, Veillonella parvula, Lactobacillus mucosae, Hungatella hathewayi, Lactobacillus crispatus, Anaerobutyricum hallii, Eisenbergiella tayi, Roseburia intestinalis, Coprobacillus cateniformis, Streptococcus salivarius, Lachnospira eligens, Absiella dolichum, Faecalibacterium prausnitzii, and Coprococcus eutactus.

[0011] The present invention also provides a use of a combination of intestinal flora markers for predicting the efficacy of immunotherapy with immune checkpoint inhibitors anti-PD-1 and / or PD-L1 antibodies in the preparation of a kit for predicting the efficacy of immunotherapy with immune checkpoint inhibitors anti-PD-1 and / or PD-L1 antibodies.

[0012] The present invention also provides a method for constructing an intestinal flora marker model for predicting the efficacy of immunotherapy with immune checkpoint inhibitors PD-1 / PD-L1 drugs, comprising the following steps:

[0013] 1) Based on the abundance of fecal intestinal microorganisms at baseline before treatment, bacterial genera reported in the literature to be associated with the efficacy of immune checkpoint inhibitors were selected;

[0014] 2) Use machine learning algorithms to build and train models to obtain stable microbiome markers across different studies for predicting the efficacy of immune checkpoint inhibitors;

[0015] 3) The relative abundance characteristics of the fecal microbiome were obtained using 16S rRNA sequencing or metagenomic sequencing. Sixteen intestinal bacteria were selected as microorganisms to predict treatment efficacy; the intestinal flora markers were:

[0016] Akkermansia muciniphila, Bifidobacterium bifidum, [Eubacterium]hallii, Veillonella parvula, Lactobacillus mucosae, Hungatella hathewayi, Lactobacilluscrispatus, Anaerobutyricum hallii, Eisenbergiella tayi, Roseburia intestinalis, Coprobacillus cateniformis, Streptococcussalivarius, Lachnospira eligens, Absiella dolichum, Faecalibacteriumprausnitzii, Coprococcus eutactus;

[0017] 4) The abundance of sixteen intestinal bacteria in the microbiome that predicts treatment efficacy can predict the efficacy of immune checkpoint inhibitors;

[0018] 5) The model prediction score calculation formula requires the use of the Random Forest function in Python, the predict() function to predict the immunotherapy benefit score (Response score), and the Youden function to calculate the benefit score cutoff value RS = 0.445071. Patients with a score higher than this value have a high benefit rate for immune checkpoint inhibitors, and vice versa, have a low benefit rate for immunotherapy.

[0019] The present invention systematically reviewed the currently published studies (including clinical information such as the therapeutic efficacy of patients with anti-PD-1 / PD-L1 antibody immunotherapy and baseline fecal microbial metagenomic sequencing information, N=568), and used the software mOTUs3 to process all the raw metagenomic sequencing data and annotate the bacterial genome, and performed statistical analysis on the bacterial abundance information at the species level. Finally, bacterial genera related to the efficacy of immunotherapy were screened out in four cohorts (a total of 568 patients), and a microbial prediction model related to immunotherapy response was constructed. There may be commonalities between the characteristics of the intestinal flora and the efficacy of PD-1 / PD-L1 antibody treatment in different types of tumors. After data supplementation and reanalysis, the immune checkpoint inhibitor treatment efficacy prediction model constructed by the intestinal flora of the present invention has application prospects in various types of tumors, such as colorectal cancer / lung cancer / melanoma / gastric cancer, and may benefit patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1After iterative feature screening, the best microbiome was used for inter-study cross-validation and leave-one-out validation of the results.

[0021] Figure 2 Figure 3 is the ROC curve and maximum area under the curve (AUC) value of the 10-fold cross-validation of the microbiome features with the best AUC in the model after iterative feature screening.

[0022] Figure 3 . Use external data to validate the random forest model derived from the training set. DETAILED DESCRIPTION

[0023] The following is a further detailed description of an intestinal microbial model related to predicting the efficacy of immunotherapy and its construction method of the present invention, in conjunction with the accompanying drawings and specific examples, in order to more clearly understand its structural type and usage. However, this does not limit the scope of protection of the patent of the present invention.

[0024] Example 1 Experimental method

[0025] External validation using a newly recruited Pan-Cancer cohort

[0026] Patient recruitment and clinical sample collection

[0027] A total of 150 patients (pan-cancer types) who are interested in clinical immunotherapy (anti-PD-1 and / or PD-L1 antibody drugs) will be recruited. The details of patient recruitment are as follows:

[0028] [Inclusion Criteria] 1) Tumor patients clearly diagnosed by histopathology; 2) patients with clear and measurable lesions; 3) patients with clear outpatient and inpatient clinical data and regular admission to the hospital for efficacy evaluation (including imaging data such as magnetic resonance imaging, chest and abdominal enhanced CT, and bone scintigraphy); 4) patients with tumors receiving tumor immune checkpoint inhibitors such as anti-PD1 / PD-L1 inhibitors.

[0029] [Exclusion criteria] 1) Severe systemic infection, nasopharyngeal and oral inflammation (such as periodontitis, gingivitis, tonsillitis, etc.), respiratory tract infection, soft tissue or skin infection, abscess, endocarditis in the past 3 months; 2) Use of the following drugs in the past 3 months: antibiotics, nonsteroidal anti-inflammatory drugs, probiotics, hormones, immunosuppressants for more than 1 week; 3) Severe constipation, diarrhea, or significant changes in bowel habits in the past 3 months; 4) Personal history of tumors (oral cancer, pharyngeal cancer, esophageal cancer) Cancer, etc.), organ transplantation, or history of severe parasitic diseases, or other digestive system diseases (such as inflammatory bowel disease, cirrhosis, etc.); 5) History of trauma or surgery within 3 months; 6) Digestive bleeding, obstruction, perforation, etc. within the past 3 months; 7) Uncontrolled chronic metabolic, infectious, or endocrine diseases (such as hypertension, diabetes, hyperlipidemia, hyperuricemia, hyperpurineemia, thyroid dysfunction, etc.); 8) Vegetarian or significant changes in dining habits within 3 months.

[0030] Baseline sample data (including plasma and stool specimens) and corresponding clinical information (including gender, age, body fat content, underlying diseases, and histopathological characteristics) were collected from each patient before receiving immunotherapy. Stool was collected longitudinally at the end of each treatment cycle to establish an immunotherapy sample library. Tumor tissue specimens (including biopsy and surgical specimens) were collected from some patients at the time of initial diagnosis. The fecal specimen collection method was as follows: To prevent the collected feces from being contaminated by urine, the subjects were instructed to urinate cleanly before collecting feces. The fecal specimens weighed at least 5-10g and were stored in disposable sterile containers in a cool place (avoiding light and high temperatures) and immediately frozen in a -80°C refrigerator.

[0031] Results: A total of 150 patients who received anti-PD-1 / PD-L1 antibody immunotherapy were collected; metagenomic sequencing analysis was performed on the feces of patients who met the requirements. The details are as follows:

[0032] Metagenomic sequencing technology and data analysis

[0033] 1. Fecal DNA extraction:

[0034] Total microbial genomic DNA was extracted from 100–200 mg of fecal samples using the HiPure Stool DNA Mini Kit (China). DNA extracts were stored at −80°C.

[0035] 2. DNA fragmentation (ultrasound fragmentation of genomic DNA)

[0036] 1) Genomic DNA quantification; 1-500ng genomic DNA (50ul system)

[0037] 2) Ultrasonic shearing of genomic DNA: 50 μl of genomic DNA was ultrasonically sheared to 250-300 bp. The ultrasonic shearing instrument used 4°C water temperature, 30 s sonication with 30 s intervals. After 3 cycles, the sample was removed and vortexed to mix thoroughly. 50 μl of the sample was purified using 1XVAHTS DNA Clean Beads (washed twice with 70% ethanol), and 50 μl of the sample was eluted with 54 μl of water.

[0038] 3. DNA library construction

[0039] Step 1: End Preparation

[0040] This step blunts the ends of the input DNA, phosphorylates the 5' end, and adds a dA tail to the 3' end.

[0041] 1) Thaw the following reagents and mix thoroughly by inversion. Prepare the following reaction in a sterile PCR tube:

[0042]

[0043] 2) Use a pipette to gently pipette and mix (do not oscillate). Centrifuge briefly to collect the reaction mixture at the bottom of the tube.

[0044] 3) Place the PCR tube in a PCR instrument and perform the following reactions:

[0045] temperature time Heated cover 105℃ On 20℃ 30min 65℃ 30min 4℃ Hold

[0046] Step 2: Adapter Ligation

[0047] This step will connect the Adapter to the end of the End Preparation product.

[0048] 1) Dilute the adapter to the appropriate concentration according to Table 2 based on the amount of input DNA.

[0049] 2) Thaw the Rapid Ligation buffer, mix thoroughly by inversion, and place on ice until ready for use.

[0050] 3) Prepare the following reaction in the End Preparation PCR tube:

[0051] Components volume End Preparation product 60 μl Ligation Buffer* 30 μl PCR-grade water* 5μl DNA Ligase* 10 μl DNA Adapter X 5μl total 110 μl

[0052] 4) Use a pipette to gently pipette and mix (do not oscillate). Centrifuge briefly to collect the reaction mixture at the bottom of the tube.

[0053] 5) Place the PCR tube in a PCR instrument and perform the following reactions:

[0054] temperature time Heated cover 105℃ On 20℃ 15min 4℃ Hold

[0055] 6) Purify the reaction product using VAHTS DNA Clean Beads:

[0056] i. After the magnetic beads have equilibrated to room temperature, vortex and mix the VAHTS DNA Clean Beads.

[0057] ii. Pipette 60µl of VAHTS DNA Clean Beads into 110µl of Adapter Ligation product and mix thoroughly by vortexing or pipetting 10 times.

[0058] iii. Incubate at room temperature for 5 minutes.

[0059] iv. Briefly centrifuge the PCR tube and place it on a magnetic rack to separate the beads and liquid. After the solution has clarified (approximately 5 minutes), carefully remove the supernatant.

[0060] v. Keep the PCR tube in the magnetic rack and add 200 μl of freshly prepared 80% ethanol to rinse the magnetic beads. Incubate at room temperature for 30 seconds and carefully remove the supernatant.

[0061] vi. Repeat step v for a total of two rinses. Keep the PCR tube in the magnetic rack. Open the lid and air-dry the beads for 5 minutes until no ethanol remains. vii. Add 23 μl of water and elute 20 μl.

[0062] Step 3: PCR enrichment of the library

[0063] 1) Thaw PCR Primer Mix 3 and VAHTS HiFi Amplification Mix, mix thoroughly by inversion, and prepare the following reaction in a sterile PCR tube:

[0064]

[0065] 2) Use a pipette to gently pipette and mix (do not oscillate). Centrifuge briefly to collect the reaction mixture at the bottom of the tube.

[0066] 3) After the reaction, take 3 μl of the sample and run agarose gel electrophoresis to identify the bands.

[0067]

[0068]

[0069] 3) Purify the reaction product using VAHTS DNA Clean Beads:

[0070] i. After the magnetic beads have equilibrated to room temperature, vortex and mix the VAHTS DNA Clean Beads.

[0071] ii. Pipette 45 μl of VAHTS DNA Clean Beads into 50 μl of Library Amplification product and mix thoroughly by vortexing or pipetting gently 10 times.

[0072] iii. Incubate at room temperature for 5 minutes;

[0073] iv. Briefly centrifuge the PCR tube and place it on a magnetic rack to separate the beads and liquid. Once the solution has clarified (approximately 5 minutes), carefully remove the supernatant.

[0074] v. Keeping the PCR tube in the magnetic rack, add 200 μl of freshly prepared 80% ethanol to rinse the magnetic beads. Incubate at room temperature for 30 seconds and carefully remove the supernatant.

[0075] vi. Repeat step v for a total of two rinses;

[0076] vii. Keep the PCR tube in the magnetic rack at all times, open the lid and air-dry the magnetic beads for 5 minutes until no ethanol residue remains;

[0077] viii. Remove the PCR tube from the magnetic stand and wash with 33ul of water to elute 30ul.

[0078] ix. Quantify the library and store it in a -20°C freezer.

[0079] Quality Inspection

[0080] Qualified standards for use on the machine: nucleic acid mass ≥ 100ng, concentration greater than 2ng / μl, volume greater than 10μl. Agarose electrophoresis band size ≥ 500bp.

[0081] Illumina Nova6000 sequencing

[0082] Use Nova6000 (illumnia, USA) to run the system for 2 days and then convert the raw data into Fastq format.

[0083] Illumina Novo-seq 6000 sequencing and raw data processing and analysis

[0084] All metagenomic libraries were sequenced using the Illumina Novaseq 6000 platform by Shanghai Nier Biotechnology Co., Ltd. The average off-machine data volume for each sample was 11.9 GB. The on-machine sequencing results were processed as follows: Sequencing was then performed using the Illumina Novaseq 6000 platform using a 2x150 bp paired-end sequencing strategy. The raw data were filtered using Trimmomatic and aligned to the human genome (Homo sapiens genome assembly hg37) using Bowtie2. Host DNA sequences were removed to generate metagenomic DNA sequences for the gut microbiota. mOTUs2.0 was then used to estimate species composition and abundance at the species level (phylum, class, order, family, genus, and species) for each sample.

[0085] The DIAMOND algorithm was used to align the intestinal flora metagenomic DNA sequences with the UniRef90 database, and the HUMAnN2 hierarchical retrieval strategy was used to calculate the relative abundance of each sample in the KEGG Orthologous (Kyoto Encyclopedia of Genes and Genomes Orthologous) database.

[0086] Verify the results

[0087] Using a newly collected fecal metagenomic data cohort (N=150), the relative abundance of intestinal flora was calculated, and 16 characteristic microbial communities were screened for inclusion in the model. The model classifier derived from the training set was used to predict each patient's likelihood of response to immune checkpoint inhibitors, and the maximum area under the curve (AUC=0.85) of the random forest model classifier based on the training data set was calculated. This suggests that the model developed by the inventors for predicting the efficacy of immune checkpoint inhibitor treatment in patients receiving immune checkpoint inhibitors based on 16 intestinal microbial characteristics has certain feasibility and reproducibility.

[0088] Figure 1 Figure 3 is the ROC curve and maximum area under the curve (AUC) value of the 10-fold cross-validation of the microbiome features with the best AUC in the model after iterative feature screening.

[0089] Figure 2 After iterative feature screening, the best microbiome was used for inter-study cross-validation and leave-one-out validation of the results.

[0090] Figure 3 It is the result of validating the model prediction score derived from the training set using external data.

[0091] Example 2

[0092] The purpose of this invention is to discover and develop highly effective new biomarkers, providing a set of gut microbiome biomarker models and methods for their construction for predicting the efficacy of ICIs. These models can predict and stratify patients receiving immune checkpoint inhibitors (anti-PD1 / PD-L1 antibody drugs) based on baseline gut microbial diversity and identify high- and low-benefit groups. Furthermore, the authors propose potential functional mechanisms for analyzing gut microbiome function.

[0093] This study reviewed existing studies on programmed death receptor-1 / programmed death ligand-1 (PD1 / PD-L1) immune checkpoint inhibitors (N > 40). The included studies had to meet the following requirements: 1. Patients treated solely with anti-PD1 / PDL1 antibody drugs; 2. The studies could provide detailed clinical information (especially including the efficacy outcomes of patients receiving immunotherapy); and 3. The studies provided raw metagenomic sequencing data that could be downloaded and analyzed.

[0094] Treatment of confounding factors

[0095] Given that the population characteristics (including tumor type, gender ratio, age and BMI) of the research subjects included in different research institutes may vary, and the sequencing time, sequencing platform, sequencing depth, etc. of different research cohorts may also vary, this study treated many factors that may affect the research objectives as confounding factors, blocked the intervals of confounding factors with greater influence, and finally included the research cohorts with the largest number of studies (N>40) as the training set (N=568).

[0096] Screening of important microbial markers

[0097] Considering the significant confounding factor of "study," the inventors used the Unified Pipeline Meta-Analysis Method for Heterogeneity in Microbiome Studies (MMUPHin) to perform a batch correction of the microbial abundance matrix based on "study." They then used the two-sided blocked Wilcoxon rank sum test implemented in the "coin" package in R software (V.4.1.2) to calculate the significance of differential abundance for each microbial marker. Finally, 65 gut microbes associated with immunotherapy efficacy with a P value < 0.05 were included for subsequent feature screening and model construction.

[0098] Microbial feature screening and model building

[0099] The inventors integrated and analyzed the microbial characteristics and immunotherapy efficacy outcomes of patients collected in the training set. For the training set data (N=568), the inventors performed iterative feature screening on 61 microbial markers related to immunotherapy efficacy. The machine learning algorithm, the random forest algorithm, was used to construct a decision tree model. The number of root nodes (N=501) was used, and each child node occupied 10% of the previous node. The iterative feature extraction resulted in 16 optimal microbial features for model construction, such as Figure 1 As shown in Figure 3, the maximum area under the receiver operating curve (ROC) can reach 0.85.

[0100] To further verify that the 16 important microbial characteristics in the screened models can be stably applied in different studies, such as Figure 2 As shown, the inventors further evaluated the stability and reproducibility of the model across different study cohorts by conducting study-by-study cross-validation and leave-one-subject-out (LOSO) cross-validation across all studies in the training set. The results showed that the 16 important characteristic microbial groups screened by the inventors were able to be applied relatively stably across different study cohorts.

[0101] Microbial efficacy prediction score and its application

[0102] The inventors used the training set data, input the 16 selected model microbial feature abundance matrices, and used the predict() function in RandomFroest to predict the probability of patients benefiting from immune checkpoint inhibitor treatment. They obtained a continuous variable - the response score (RS). The Youden index was used to calculate the cutoff value (RS = 0.445071) of the response score in the training set data. This cutoff value (RS = 0.445071) was used as the threshold for subsequent external validation model evaluation.

[0103] Furthermore, the inventors used the newly collected metagenomic data of 150 patients to externally validate the established prediction model, and used the relative abundance of 16 microorganisms in the model to calculate the immune checkpoint inhibitor benefit score of newly admitted cases. Patients with RS scores above the cutoff value of 0.445071 were predicted to have a higher treatment benefit rate, and vice versa. Finally, the predicted values ​​were compared with the actual outcomes. The maximum area under the curve of the microbial model for predicting the efficacy of anti-PD1 / PD-L1 antibody immunotherapy in patients using RS scores can reach 0.84 ( Figure 3 ).

Claims

1. A combination of intestinal flora markers for predicting the efficacy of immunotherapy with immune checkpoint inhibitors anti-PD-1 and / or PD-L1 antibodies, characterized in that: The invention is composed of sixteen bacterial targets, which are: Akkermansiamuciniphila, Bifidobacterium bifidum, [Eubacterium]hallii, Veillonella parvula, Lactobacillus mucosae, Hungatella hathewayi, Lactobacillus crispatus, Anaerobutyricum hallii, Eisenbergiella tayi, Roseburia intestinalis, Coprobacillus cateniformis, Streptococcus salivarius, Lachnospira eligens, Absiella dolichum, Faecalibacterium prausnitzii, and Coprococcus eutactus.

2. A kit for predicting the efficacy of immune checkpoint inhibitor anti-PD-1 and / or PD-L1 antibody immunotherapy, characterized in that: The kit contains reagents for detecting the following intestinal flora markers: Akkermansiamuciniphila, Bifidobacterium bifidum, [Eubacterium]hallii, Veillonella parvula, Lactobacillus mucosae, Hungatella hathewayi, Lactobacillus crispatus, Anaerobutyricum hallii, Eisenbergiella tayi, Roseburia intestinalis, Coprobacillus cateniformis, Streptococcus salivarius, Lachnospira eligens, Absiella dolichum, Faecalibacterium prausnitzii, and Coprococcus eutactus.

3. Use of a combination of intestinal flora markers for predicting the efficacy of immunotherapy with immune checkpoint inhibitors anti-PD-1 and / or PD-L1 antibodies in the preparation of a kit for predicting the efficacy of immunotherapy with immune checkpoint inhibitors anti-PD-1 and / or PD-L1 antibodies.

4. A method for constructing an intestinal flora marker model for predicting the efficacy of immunotherapy with immune checkpoint inhibitors PD-1 / PD-L1, characterized in that: The steps include: 1) Based on the abundance of fecal intestinal microorganisms at baseline before treatment, bacterial genera reported in the literature to be associated with the efficacy of immune checkpoint inhibitors were selected; 2) Use machine learning algorithms to build and train models to obtain stable microbiome markers across different studies for predicting the efficacy of immune checkpoint inhibitors; 3) The relative abundance of the fecal microbiome was determined using 16S rRNA sequencing or metagenomic sequencing. Sixteen intestinal bacteria were selected as predictors of treatment efficacy. The intestinal microbiome markers were: Akkermansia muciniphila, Bifidobacterium bifidum, [Eubacterium] hallii, Veillonella parvula, Lactobacillus mucosae, Hungatella hathewayi, Lactobacillus crispatus, Anaerobutyricum hallii, Eisenbergiella tayi, Roseburia intestinalis, Coprobacillus cateniformis, Streptococcus salivarius, Lachnospira eligens, Absiella dolichum, Faecalibacterium prausnitzii, and Coprococcus eutactus. 4) The abundance of sixteen intestinal bacteria in the microbiome that predicts treatment efficacy can predict the efficacy of immune checkpoint inhibitors; 5) The model prediction score calculation formula requires the use of the Random Forest function in Python, the predict() function to predict the immunotherapy benefit score (Response score), and the Youden function to calculate the benefit score cutoff value RS = 0.445071. Patients with a score higher than this value have a high benefit rate for immune checkpoint inhibitors, and vice versa, have a low benefit rate for immunotherapy.

Images