Application of substance for detecting microorganisms in preparation of cervical cancer and cervical precancerous lesion products

Abstract

The invention discloses an application of a substance for detecting microorganisms in preparation of cervical cancer and cervical cancer precancerous lesion products, and the microorganisms are selected from at least one of micromonas KA00067, anaerococcus PH9, dilisteria microaerogenes, protobacterium pemorini and lactobacillus inervis. The abundance of the five microorganisms is obviously different between a CIN2-group sample and a CIN3 + group sample; then, a random forest algorithm is adopted to construct a cervical cancer and precancerous lesion detection model, meanwhile, the generalization ability and anti-overfitting performance of the cervical cancer and precancerous lesion detection model are improved, it is determined that the five microorganisms can serve as markers, the conditions of the cervical cancer and precancerous lesion are predicted and evaluated very accurately, and the detection accuracy of the cervical cancer and precancerous lesion detection model is improved. Therefore, doctors can formulate more effective treatment schemes, and the survival rate and life quality of patients are improved.

Description

Technical Field

[0001] This invention belongs to the field of biotechnology, and specifically relates to the application of substances for detecting microorganisms in the preparation of products for cervical cancer and precancerous cervical lesions. Background Technology

[0002] Cervical intraepithelial neoplasia (CIN) is a precancerous lesion. The classification and management of CIN is highly dependent on histopathological diagnosis, but the invasiveness of the procedure and inter-observer variability remain challenges that urgently need to be addressed in clinical practice.

[0003] Recent studies have revealed that dysbiosis in the cervical epithelial / vaginal surface microenvironment may directly affect CIN progression through mechanisms such as regulating local immune responses and metabolic reprogramming. Cervical microbes directly colonize the neoplastic area, and changes in their composition better reflect biological processes more closely related to disease progression. The vaginal flora, as the first line of defense for cervical health, maintains local microecological balance and effectively prevents ascending pathogen invasion. Simultaneously, disturbances in the vaginal microbiota can cause dysbiosis of the uterine microecology. This may encourage harmful microorganisms to ascend into the uterus and other reproductive organs, triggering ascending infections and consequently affecting the health of the female reproductive system, such as cervical intraepithelial neoplasia.

[0004] Early detection and treatment can significantly improve patients' survival rates and quality of life. Studies have shown that vaginal microbiome dysbiosis is closely related to the occurrence and development of cervical cancer. Therefore, there is an urgent need to find a microbial biomarker as a basis for early diagnostic assessment of cervical cancer. By analyzing the composition and functional changes of microbial biomarkers in patients, cancer signs can be detected earlier, thereby assisting current cervical cancer screening methods to reduce unnecessary overtreatment. Summary of the Invention

[0005] The purpose of this invention is to provide the application of substances for detecting microorganisms in the preparation of products for cervical cancer and precancerous cervical lesions, and to provide new microbial markers for the early screening or diagnosis of cervical cancer and precancerous cervical lesions.

[0006] To achieve the above-mentioned objectives, the technical solution of the present invention is as follows:

[0007] The application of a substance for detecting microorganisms in the preparation of products for cervical cancer and precancerous cervical lesions, wherein the microorganisms are selected from at least one of the following: *Parvimonas* sp. KA00067, *Anaerococcus* sp. PH9, *Dialister microaerophilus*, *Prevotella timonensis*, and *Lactobacillus iner*. The substance for detecting microorganisms is selected from at least one of the following reagents, devices, and equipment used for detecting the abundance of the aforementioned microorganisms.

[0008] This invention first extracts biomarkers related to the progression of cervical cancer and precancerous lesions from metagenomic data using differential analysis. It found significant differences in the abundance of five microorganisms—Parvimonas_sp._KA00067, Anaerococcus_sp._PH9, Dialister_micraerophilus, Prevotella_timonensis, and Lactobacillus iner—between CIN2- group samples (including healthy, CIN1, and CIN2 samples) and CIN3+ group samples (including CIN3 and cervical cancer samples). Then, a random forest algorithm is used to construct a cervical cancer and precancerous lesion detection model, which enhances the model's generalization ability and resistance to overfitting. The invention confirms that these five microorganisms can serve as markers to accurately predict and assess the condition of cervical cancer and precancerous lesions, helping doctors develop more effective treatment plans and improving patient survival rates and quality of life.

[0009] In this invention, the microorganisms are taken from at least one of the cervical tissue and the vaginal surface.

[0010] In this invention, the cervical precancerous lesion refers to cervical intraepithelial neoplasia.

[0011] In this invention, the cervical cancer and precancerous lesion product package is a cervical cancer and precancerous lesion screening product, a cervical cancer and precancerous lesion diagnostic product, or a cervical cancer and precancerous lesion staging product; wherein, staging can distinguish between CIN2- stage and CIN3+ stage.

[0012] In this invention, the abundance of four microorganisms—Parvimonas sp. KA00067, Anaerococcus sp. PH9, Dialister micraerophilus, and Prevotella timonensis—was significantly increased in the CIN3+ group samples, while the abundance of Lactobacillus iner was significantly decreased in the CIN3+ group samples. According to the detection model analysis, the above five microbial markers all showed good single-bacterial diagnostic efficacy for cervical cancer and precancerous cervical lesions. Among them, Lactobacillus iner, obtained from cervical tissue, had the highest ROC curve area (AUC) of 0.952 (mean AUC of 0.813), demonstrating excellent performance.

[0013] When the above five microorganisms are randomly combined as markers, their diagnostic efficacy is better than that of a single bacterium. Specifically, the preferred microorganisms consist of *Parvimonas* sp. KA00067, *Anaerococcus* sp. PH9, *Dialister micraerophilus*, *Prevotella timonensis*, and *Lactobacillus iner*. ROC curve analysis showed an area under the curve (AUC) of 0.923 for the training set and 0.894 for the validation set, indicating that this combination of five microorganisms performs well in distinguishing between positive and negative samples (i.e., CIN2- and CIN3+ samples), with a sensitivity of 0.928 and a specificity of 0.889.

[0014] Compared with the prior art, the technical effects of the present invention are reflected in:

[0015] This invention first extracts biomarkers related to the progression of cervical cancer and precancerous lesions from metagenomic data using differential analysis. It found significant differences in the abundance of five microorganisms—Parvimonas_sp._KA00067, Anaerococcus_sp._PH9, Dialister_micraerophilus, Prevotella_timonensis, and Lactobacillus iner—between CIN2- group samples (including healthy, CIN1, and CIN2 samples) and CIN3+ group samples (including CIN3 and cervical cancer samples). Then, a random forest algorithm is used to construct a cervical cancer and precancerous lesion detection model, which enhances the model's generalization ability and resistance to overfitting. The invention confirms that these five microorganisms can serve as markers to accurately predict and assess the condition of cervical cancer and precancerous lesions, helping doctors develop more effective treatment plans and improving patient survival rates and quality of life. Among them, Lactobacillus iner from cervical tissue showed the highest single-bacterial diagnostic efficacy, with the highest area under the ROC curve reaching 0.952 (mean AUC reaching 0.813); while the combination of the above five microorganisms from cervical tissue and vaginal surface showed the highest combined diagnostic efficacy, with an area under the curve AUC of 0.923 for the training set, an AUC of 0.894 for the validation set, a sensitivity of 0.928, and a specificity of 0.889. Attached Figure Description

[0016] Figure 1 This is an analysis of the importance of variables for each differentially expressed bacterium in the random forest model constructed in Embodiment 2 of the present invention;

[0017] Here, Variable Importance represents the importance of the variable, Mean Decrease Gini represents the influence of each variable on the heterogeneity of the observations at each node in the classification tree, and the larger the value, the greater the importance of the variable. Parvimonas_sp._KA00067 represents Micromonas KA00067, Anaerococcus_sp._PH9 represents Anaerococcus PH9, Dialister_micraerophilus represents Microaerophilus, Prevotella_timonensis represents Prevotella timonensis, and Lactobacillus iner represents Lactobacillus inertis.

[0018] Figure 2 ROC curves for the training set of a random forest model based on five differentially expressed microbial biomarkers;

[0019] Where ROC Curve represents the ROC curve, AUC represents the area under the curve, Specificity represents specificity, and Sensitivity represents sensitivity; the same applies below.

[0020] Figure 3 ROC curves for the validation set of a random forest model based on five differentially expressed microbial biomarkers. Detailed Implementation

[0021] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] Example 1: Screening candidate markers

[0023] 1. Data Collection and Processing

[0024] Data on the human reproductive tract microbiome related to cervical intraepithelial neoplasia were retrieved from PubMed's SRA (Sequence Read Archive). 16S rRNA data from 3 cervical tissue samples and 19 vaginal surface samples were collected. The SRA Toolkit software was installed on a Linux operating system. The required sequencing data were downloaded in batches from the SRA database using the "prefetch" command provided by the SRA Toolkit. The BioProject number, sampling site (cervical tissue or vaginal surface), and disease status (Normal / CIN1 / CIN2 / CIN3 / Cancer) were analyzed, processed, and recorded using the corresponding metadata provided by SRA and relevant published literature.

[0025] Among them, Normal means normal; CIN1 is Cervical Intraepithelial Neoplasia 1, which indicates mild dysplasia of cervical squamous epithelium; CIN2 is Cervical Intraepithelial Neoplasia 2, which indicates moderate dysplasia of cervical squamous epithelium; CIN3 is Cervical Intraepithelial Neoplasia 3, which indicates severe dysplasia of cervical squamous epithelium and carcinoma in situ; and Ca is Cervical cancer.

[0026] 2. Sequence file processing

[0027] All raw sequencing data were converted from SRA files to FastQ format using the "Fastqdump" parameter in SRAToolkit. Paired-end sequences were merged, while single-end sequences were left unprocessed. Redundancy was removed from sequences in each accession number file, and a minimum size of 10 was added to remove low-abundance sequences. Primer excision and quality control were performed on the FastQ sequence files using the dada2 function in QIIME 2 software. A full-length classifier was constructed from the Silva database (SILVA 138.1), and sequences were annotated. Phyloseq objects were created using the phyloseq package in R, importing the OTU table, sample data, and classification information. After merging the various phyloseq objects, samples with more than 1000 sequences were selected. Using the accession number as a reference, conditional quantile regression (ConQuR) was used to standardize the multi-cohort data to remove batch effects for subsequent analyses of alpha diversity, beta diversity, and species composition. The samples ultimately included in the downstream analysis were: 163 cervical conization tissue samples from the CIN2- group (i.e., including CIN2, CIN1, and Normal groups); 80 cervical conization tissue samples from the CIN3+ group (i.e., including CIN3 and Cancer groups); 715 vaginal surface samples from the CIN2- group; and 194 vaginal surface samples from the CIN3+ group.

[0028] 3. Screening of candidate biomarkers

[0029] Classical differential analysis methods (ALDEx2, Cornob, LefSe, ANCOM-II, ANCOM-BC, metagenomeSeq, MaAsLin2, and LinDA) were used to compare the differential bacterial populations at the species level in cervical tissue and vaginal surface between CIN2- and CIN3+ and to calculate the single-bacterial discrimination power (AUC value).

[0030] In the evaluation of single-bacterial classification performance, this embodiment employs a single-feature-based random forest model and a repeated validation strategy to calculate the AUC value of each bacterium. Specifically, a stratified sampling strategy is implemented using the `createDataPartition` function of the `caret` package, dividing the sample data into training and validation sets in a 7:3 ratio, ensuring that the sample proportions for each category are consistent with the original dataset. To evaluate the stability of single-bacterial classification performance, 100 independent random seeds are set. A single-feature random forest model is independently constructed for each random partition. The feature selection criteria require that feature values ​​in the training set have variability (group label > 2); features that do not meet the criteria are marked as NA and excluded from analysis. The maximum and mean AUC values ​​of single bacteria are calculated from multiple repeated experiments.

[0031] Ultimately, it was found that 71 bacteria in the cervical tissue microbiome had a mean AUC greater than 0.50 and a maximum AUC greater than 0.60; 23 bacteria in the vaginal microbiome had a mean AUC greater than 0.50 and a maximum AUC greater than 0.60, demonstrating significant CIN3+ predictive ability.

[0032] Further analysis of the differentially expressed bacterial species obtained from screening cervical tissue and vagina yielded five species with consistent differences as candidate biomarkers (see Table 1). Among them, four species were significantly enriched in the cervical tissue / vaginal tissue of CIN3+ patients: Parvimonas sp. KA00067, Anaerococcus sp. PH9, Dialister micraerophilus, and Prevotella timonensis. One species was significantly reduced in the cervical tissue / vaginal tissue of CIN3+ patients: Lactobacillusiners.

[0033] Table 1. Ability of differentially expressed bacteria to distinguish CIN2- and CIN3+

[0034]

[0035] Example 2: Clinical validation of the diagnostic efficacy of microbial markers

[0036] 1. Sample collection

[0037] A retrospective analysis was conducted on vaginal secretion samples collected by the Institute of Clinical Pharmacology, Xiangya Hospital, Central South University, between 2021 and 2022. The study included a control group of 10 patients with negative HPV and TCT tests, and 66 patients with cervical intraepithelial neoplasia (CIN1 group: 11 cases, CIN2 group: 10 cases, CIN3 group: 29 cases, and Ca group: 16 cases). These 66 patients with cervical intraepithelial neoplasia were divided into a training set and a validation set in a 7:3 ratio.

[0038] Inclusion criteria: women aged 25-55 years; voluntarily participating in the research project and signing the informed consent form; having a history of sexual activity; having not had sexual activity within 48 hours; not having undergone vaginal irrigation or medication within 3 days prior to the examination; not having undergone vaginal examination and not during menstruation; and having clear colposcopy / biopsy results.

[0039] Exclusion criteria: pregnant women; those with a history of cervical surgery (such as hysterectomy); those with a history of treatment for cervical and vaginal lesions; patients with other malignant tumors; patients with hematologic or digestive system diseases; those with no history of radiotherapy or chemotherapy; those who withdrew midway; and those with incomplete clinical information and related examination information.

[0040] 2. 16S third-generation sequencing analysis

[0041] Total microbial DNA from vaginal swab samples was amplified and a DNA library was constructed. Single-molecule real-time (SMRT) sequencing was performed on the PacBio Sequel platform. The `dada2` package in R software (4.2.2) was used for quality control, noise reduction, and chimera removal. First, primers were removed directionally using the `remove Primers` function. Then, the `filterAnd Trim` function was used for sequence quality control, selecting and retaining high-quality sequences between 1000-1600 bp in length. After noise reduction using the `dada` function, chimera removal was performed using the `remove Bimera Denovo` function. Finally, all valid sequences were aligned using the SILVA 138 database and the BLAST algorithm on the QIIME 2 platform for species annotation and subsequent basic bioinformatics analysis. The bacterial read count values ​​of the samples were standardized (using the DESeq2 standardization method: the normalization coefficient was calculated, and the sequence number divided by the normalization coefficient gave the relative abundance of the bacterial species after standardization) to obtain the relative abundance.

[0042] 3. Construct a random forest model

[0043] The abundance data of the five differentially expressed bacterial communities selected in Example 1 were standardized and used as feature variables for a random forest model. Machine learning modeling was performed using the random Forest package (version 4.7-1.1) in R. A classification model was constructed using the random Forest function with default hyperparameters to explore the relationship between the five differentially expressed bacterial communities and cervical cancer and precancerous lesions.

[0044] The random forest model is built based on the following default settings: number of decision trees generated (n) tree The number of features (m) is 500; the number of features randomly selected during each split is... try For classification problems, the parameter is set to the square root of the total number of predictors; the minimum number of samples for node splitting (nodesize) is 1; all other parameters remain at their default values. Ten-fold cross-validation is performed to confirm the robustness and reliability of the model. After determining the optimal parameters, a learning curve is plotted to measure the model's fit to the data, resulting in a cervical cancer and precancerous lesion detection model.

[0045] like Figure 1 As shown, "Mean Decrease Gini" is an important indicator in the cervical cancer and severe precancerous lesion detection model. It is used to calculate the impact of each variable on the heterogeneity of observations on the classification tree node and to compare the importance of variables. The larger the value, the more important the variable.

[0046] After analyzing the ability of each strain to distinguish between CIN2- and CIN3+ using ROC curves, the following results were found: Lactobacillus iners had a Max AUC value of 0.792 for distinguishing between CIN3+ clinical patients; Parvimonas sp. KA0006 had a Max AUC value of 0.583; Anaerococcus sp. PH9 had a Max AUC value of 0.667; Dialistermicraerophilus had a Max AUC value of 0.708; and Prevotella timonensis had a Max AUC value of 0.607. This indicates that among these five differentially expressed bacteria, Lactobacillus iners is the most important for the accurate diagnosis of cervical cancer and precancerous lesions.

[0047] Further ROC curves were plotted to evaluate the performance of the cervical cancer and precancerous lesion detection model, and the results are as follows: Figure 2 and Figure 3 As shown in the figure, the horizontal axis (X-axis) represents 1-Specificity, which is the false positive rate, ranging from 0 to 1, reflecting the proportion of cervical cancer detection models that incorrectly predict negative cases as positive cases; the vertical axis (Y-axis) represents Sensitivity, which is the true positive rate, measuring the ability of cervical cancer detection models to correctly identify positive cases.

[0048] from Figure 2 and Figure 3 As can be seen, the ROC curves are all close to the upper left corner. The area under the curve (AUC) of the training set is 0.923, and the AUC of the validation set is 0.894, indicating that the model can distinguish between positive and negative samples well, with a sensitivity of 0.928 and a specificity of 0.889.

Claims

1. The application of a substance for detecting microorganisms in the preparation of products for cervical cancer and precancerous cervical lesions, characterized in that, The microorganisms mentioned are selected from at least one of the following: Parvimonas sp. KA00067, Anaerococcus sp. PH9, Dialister micraerophilus, Prevotella timonensis, and Lactobacillus iner. The substance used to detect microorganisms is selected from at least one of the reagents, devices, and equipment used to detect the abundance of the aforementioned microorganisms.

2. The application as described in claim 1, characterized in that, The microorganism mentioned is *Lactobacillusiner*.

3. The application as described in claim 1, characterized in that, The microorganisms described consist of Parvimonas sp. KA00067, Anaerococcus sp. PH9, Dialister micraerophilus, Prevotella timonensis, and Lactobacillus inertia.

4. The application as described in claim 1, characterized in that, The microorganisms are taken from at least one site on the cervical tissue and the vaginal surface.

5. The application as described in claim 1, characterized in that, The aforementioned precancerous lesions of the cervix refer to cervical intraepithelial neoplasia.

6. The application as described in claim 1, characterized in that, The aforementioned cervical cancer and precancerous lesion product package includes cervical cancer and precancerous lesion screening products, cervical cancer and precancerous lesion diagnostic products, or cervical cancer and precancerous lesion staging products.

Images