Application of product for detecting abundance of intestinal microbiota
By detecting the abundance of intestinal microbiota and random forest models, combining intestinal metabolism and age-related microbiota, the problem of lack of guidance for personalized treatment in colorectal cancer diagnosis and prevention is solved, and the incidence is significantly reduced and survival is improved rate effect.
Patent Information
- Application Number
- CN202411163456.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-23
- Publication Date
- 2024-11-12
AI Technical Summary
The prior art is difficult to effectively utilize the abundance of intestinal microbiota for the diagnosis and prevention of colorectal cancer, and the lack of guidance on personalized treatment strategies, resulting in high morbidity and mortality.
Provide products and random forest models for detecting the abundance of the gut microbiota, combining intestinal metabolism and age-related microbiota for diagnosis or assisting in the diagnosis of colorectal cancer, guiding personalization by identifying the association of specific microbial community structures with cancer treat.
It significantly reduces the incidence of colorectal cancer, improves patient survival, provides feasible solutions for the prevention and treatment of CRC, and reveals the complex interaction between the intestinal microbiota and colorectal cancer.
Smart Images

Figure CN118932062A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of biomedicine technology, and specifically to the application of a product for detecting the abundance of intestinal microorganisms, and further to the application of a product for detecting the abundance of intestinal microorganisms in the preparation of a product for diagnosing or assisting in the diagnosis of colorectal cancer. Background Art
[0002] Colorectal cancer (CRC) ranks third in incidence worldwide and is the second leading cause of cancer-related death. Despite extensive research, the exact etiology and pathogenesis of colorectal cancer remain poorly understood. Advances in technologies such as high-throughput sequencing have enabled the discovery of changes in the type and abundance of the gut microbiome in patients with CRC. The gut microbiota consists of trillions of bacteria, viruses, fungi, and other microorganisms. Bacteroidetes, Firmicutes, Actinobacteria, Proteobacteria, and Coccus are the dominant bacterial phyla in the colon of healthy adults.
[0003] Given the unique involvement of the gut microbiome in the onset and progression of CRC, efforts are underway to develop microbiome-based biomarkers and therapeutic approaches targeting this “hidden organ.” Studies have shown that changes in the composition and function of the gut microbiota can significantly affect the occurrence, promotion, and progression of CRC.
[0004] Multiple lines of evidence from animal models and human studies have implicated the gut microbiota in the development of colorectal cancer. Specific pathogens, such as Fusobacteria and Bacteroides fragilis, as well as alterations in the overall community structure, have been associated with CRC. Multiple potential mechanisms may contribute to the carcinogenic effects of a dysregulated microbiome on cancer: microbial toxins directly damage DNA, produce metabolites that promote tumor growth, disrupt epithelial barrier function, induce chronic inflammation, and modulate antitumor immunity.
[0005] In contrast, a healthy gut microbiota provides anticancer effects by strengthening the mucosal layer, maintaining immune homeostasis, regulating proliferation and differentiation processes, detoxifying carcinogens, and activating tumor suppressor pathways.
[0006] In addition, the gut microbiome has a significant impact on the clinical response to immunotherapy, chemotherapy, and radiotherapy through multiple mechanisms, which provides a theoretical basis for the development of microbiome-based combination treatment strategies. For example, the genus Fusobacterium has been associated with the occurrence of CRC because it can enhance the adhesion and invasion of tumor cells, induce the production of inflammatory cytokines, and inhibit antitumor immune responses. This bacterium is commonly found in the oral cavity and gastrointestinal tract and has been shown to be more abundant in CRC tumor tissues than in normal adjacent tissues. In addition, certain intestinal bacteria can produce short-chain fatty acids (SCFAs), such as butyrate, through dietary fiber fermentation, which exhibit anticancer effects by inhibiting CRC cell growth, promoting cell apoptosis, and enhancing intestinal barrier integrity. Conversely, an imbalance in the gut microbiota may lead to a decrease in SCFAs, thereby promoting the development of CRC. Bifidobacteria and Lactobacilli have been shown to enhance immune surveillance and have anti-inflammatory effects, which may reduce the risk of CRC. These examples highlight the complex mechanisms of the gut microbiome on the development and progression of CRC.
[0007] Efforts to modulate the microbiome to prevent or treat CRC have made significant progress. Probiotics, prebiotics, symbiotics, and postbiotics have been used to promote the growth of beneficial bacteria such as bifidobacteria and butyrate-producing bacteria. Antibiotics and bacteriocins may help eliminate cancer-causing pathogens. One study showed that antibiotics have the potential to enhance clinical responses to immunotherapy. Succinic acid, derived from Fusobacterium nucleatum, which is highly prevalent in the intestines of patients with metastatic colorectal cancer, inhibits the cGAS-IFNβ pathway and reduces the levels of chemokines CCL5 and CXCL10, thereby limiting the migration of CD8+ T cells into the tumor microenvironment and hindering antitumor responses. After treatment with the antibiotic metronidazole, mice showed resensitization to immunotherapy due to reduced serum succinic acid levels. Fecal microbial transplantation to restore healthy community structure has shown promise in early trials.
[0008] The microbiome plays an important role in nutrient and drug metabolism, synthesis of amino acids and vitamins, development of the immune system, metabolic regulation, and protection of the intestinal epithelium. The composition of the gut microbiota is influenced by early life exposures and changes with age, diet, medication use, and other host and environmental factors.
[0009] Colorectal cancer is a common digestive tract malignancy, and its incidence and mortality increase significantly with age. Further exploration will help reveal the complex interactions between the intestinal microbiota and the development of colorectal cancer, guide the development of personalized treatment strategies across different age groups, significantly reduce the incidence and improve patient survival, and provide valuable insights into CRC prevention and treatment. Summary of the invention
[0010] In view of this, the present invention provides the use of a product for detecting the abundance of intestinal microbiota in the preparation of a product for diagnosing or assisting in the diagnosis of colorectal cancer, as well as a random forest model and system for diagnosing or assisting in the diagnosis of colorectal cancer.
[0011] In a first aspect, the present invention provides the use of a product for detecting the abundance of intestinal microbiota in the preparation of a product for diagnosing or assisting in the diagnosis of colorectal cancer, wherein the intestinal microbiota is an intestinal microbiota associated with intestinal metabolism and age.
[0012] In an optional embodiment, the patient to be diagnosed is less than 50 years old, and the intestinal microbiota comprises:
[0013] Rombus, Faecalibacterium, Phocaeicola, Lachnospiraceae, Klebsiella, GGB9747, Andeckia, GGB9573, GGB3429, Corynebacterium, Alternaria, GGB3363, Odorobacter, GGB9691, Zurichobacter, GGB51441, GGB35068, Odorobacter, Enterobacter, Enterocloster, Lansfieldella, Eubacteriales_Family_Xlll_Incertae_Ssdis_unclassified, Candidatus_Nanosynsacchari, Fusobacterium, Anaerobic Bacillus, and Mediterraneibacter.
[0014] In an optional embodiment, the age of the patient to be diagnosed is greater than or equal to 50 years old and less than 65 years old, and the intestinal microbiota includes:
[0015] Asha, Fusobacterium, Roseburia, Odorobacter, Gemini, Peptostreptococcus, Candidatus_Nanoper, Flavonoids, Porphyromonas, GGB2998, Clostridium, Enterocloster, Eubacteriales_Family_Xlll_Incertae_Ssdis_unclassified, GGB55957, GGB9561, Dysosmobacter, Streptospira, GGB9559, unclassified Oscillatory Spiraceae, Streptospira, Oscillobacter, and Abacterium.
[0016] In an optional embodiment, the age of the patient to be diagnosed is greater than or equal to 65 years old, and the intestinal microbiota includes:
[0017] Blautia, Peptostreptococcus, Enterocloster, Odorbacterium, Dorsalella, Micromonas, Roseburia, Anaerobic Bacillus, Candidatus_Nanoperiomorbus, GGB33928, Fusobacterium, Phocaeicola, Anaerobic Bacteria, Oscillating Bacterium, Flavonoids, Phocea, Another Cladosporium, Candidatus_Cibion, Vescimonas, GGB3612, Jimmyella, GGB34228, Unclassified Lachnospiraceae, Butyrivibrio, Eagle's Head, Massiliimalia.
[0018] In an optional embodiment, the product for detecting the abundance of intestinal microbiota includes at least one of a specific recognition primer, a probe, an antisense oligonucleotide, an aptamer or an antibody for intestinal microorganisms.
[0019] In a second aspect, the present application provides a random forest model for diagnosing or assisting in the diagnosis of colorectal cancer, characterized in that the training process of the random forest model includes:
[0020] Obtain the target gut microbiota abundance and colorectal cancer diagnosis results of at least two patients;
[0021] The target intestinal microbiome abundance of each patient is used as input data, and the corresponding colorectal cancer diagnosis result is used as output data to train a random forest model;
[0022] Wherein, the target intestinal microbiota is an intestinal microbiota associated with intestinal metabolism and age, and the colorectal cancer diagnosis result includes having colorectal cancer and not having colorectal cancer.
[0023] In an optional embodiment, the patient is less than 50 years old, and the target intestinal microbiota comprises:
[0024] Rombus, Faecalibacterium, Phocaeicola, Lachnospiraceae, Klebsiella, GGB9747, Andeckia, GGB9573, GGB3429, Corynebacterium, Alternaria, GGB3363, Odorobacter, GGB9691, Zurichobacter, GGB51441, GGB35068, Odorobacter, Enterobacter, Enterocloster, Lansfieldella, Eubacteriales_Family_Xlll_Incertae_Ssdis_unclassified, Candidatus_Nanosynsacchari, Fusobacterium, Anaerobic Bacillus, and Mediterraneibacter.
[0025] In an optional embodiment, the patient is greater than or equal to 50 years old and less than 65 years old, and the target intestinal microbiota includes:
[0026] Asha, Fusobacterium, Roseburia, Odorobacter, Gemini, Peptostreptococcus, Candidatus_Nanoper, Flavonoids, Porphyromonas, GGB2998, Clostridium, Enterocloster, Eubacteriales_Family_Xlll_Incertae_Ssdis_unclassified, GGB55957, GGB9561, Dysosmobacter, Streptospira, GGB9559, unclassified Oscillatory Spiraceae, Streptospira, Oscillobacter, and Abacterium.
[0027] In an optional embodiment, the patient is greater than or equal to 65 years old, and the intestinal microbiota comprises:
[0028] Blautia, Peptostreptococcus, Enterocloster, Odorbacterium, Dorsalella, Micromonas, Roseburia, Anaerobic Bacillus, Candidatus_Nanoperiomorbus, GGB33928, Fusobacterium, Phocaeicola, Anaerobic Bacteria, Oscillating Bacterium, Flavonoids, Phocea, Another Cladosporium, Candidatus_Cibion, Vescimonas, GGB3612, Jimmyella, GGB34228, Unclassified Lachnospiraceae, Butyrivibrio, Eagle's Head, Massiliimalia.
[0029] In a third aspect, the present application provides a system for diagnosing or assisting in the diagnosis of colorectal cancer, the system comprising an input device, a computing device, and an output device; wherein:
[0030] The input device is used to input the target intestinal microbial abundance of the patient to be diagnosed;
[0031] The output device is used to output the colorectal cancer diagnosis result of the patient to be diagnosed;
[0032] The computing device includes a memory and a processor; a computer program is stored in the memory, and when the computer program is run on an electronic device, the electronic device executes the random forest model as described in any one of claims 6 to 9.
[0033] The technical solution of this application has at least the following beneficial effects:
[0034] The present application provides the use of a product for detecting the abundance of intestinal microbiota associated with intestinal metabolism and age in the preparation of a product for diagnosing or assisting in the diagnosis of colorectal cancer, which reveals the complex interaction between intestinal microbiota associated with intestinal metabolism and age and the occurrence and development of colorectal cancer, helps guide the formulation of personalized treatment strategies across different age groups, can significantly reduce the incidence of colorectal cancer and improve patient survival, and provides a feasible solution for the prevention and treatment of CRC. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0036] Figure 1 , (A) α-diversity estimated by Pieiou even, Shannon, Simpson even index, p-value calculated by two-sided unpaired t-test; (B) Principal coordinates analysis (PCoA) of diseased and healthy groups based on Bray-Curtis distance showed that microbial composition was significant between groups (p=0.001) and between cohorts (p=0.000); p-value of β-diversity based on Bray-Curtis distance was calculated by two-sided test using PERMANOVA; (C) Principal coordinates analysis (PCoA) based on Bray-Curtis distance of age-specific CRC groups; (D) Principal coordinates analysis (PCoA) of healthy groups of different ages based on Bray-Curtis distance; (E) Venn diagram showing overlap between groups; (F) Inverted diagram showing the number of different bacterial species in each group, shared by the combination of data sets; the numbers above each column indicate the size of the differential species; the set size on the right indicates the number of differential species in each cohort;
[0037] Figure 2 , (A) phylum-level species composition of the first 15 species in each group; (B) class-level species composition of the first 15 Bacteroidetes in each group; (C) order-level species composition of the first 15 bacterial orders in each group; (D) family-level species composition of the first 15 bacterial families in each group; (E) genus-level species composition of the first 15 species in each group; (F) species composition of the first 15 bacterial species in each group;
[0038] Figure 3, (A) Heat map of 6 different methods for analyzing the differences at the genus level between the young CRC group and the young healthy group; (B) Heat map of 6 different ANOVA methods at the genus level between the middle-aged and elderly CRC group and the middle-aged healthy group; (C) Heat map of 6 different ANOVA methods at the genus level between the elderly CRC group and the elderly healthy group; (D) Heat map of 6 different ANOVA methods at the genus level between the young CRC group and the young healthy group; (E) Heat map of 6 different methods for analyzing the differences at the species level between the middle-aged CRC group and the middle-aged healthy group; (F) Heat map of 6 different ANOVA methods at the species level between the elderly CRC group and the elderly healthy group;
[0039] Figure 4 , (A) LDA histogram of the juvenile CRC group (C_Y) and the juvenile healthy group (N_Y) at the species level (LDA score>3.0, p<0.05); (B) Imprint map of the young CRC group and the young healthy group at the genus level (two-sided unpaired t-test, p<0.05); (C) LDA histogram of the middle-aged CRC group (C_M) and the middle-aged healthy group (N_M) at the genus level (LDA score>3.0, p<0.05); (D) Imprint map of the middle-aged CRC group and the middle-aged healthy group at the genus level (two-sided unpaired t-test, p<0.05); (E) LDA histogram of the genus level between the old CRC group (C_O) and the old healthy group (N_O) (LDA score>3.0, p<0.05); (F) Imprint map of the old CRC group and the old healthy group at the genus level (two-sided unpaired t-test, p<0.05);
[0040] Figure 5 , (A) CRC group with significant changes in age groups in PRJNA731589; (B) Line graph of CRC group with significant changes in age-specific microbiota in validation set PRJNA763023; (C) CRC group with significant changes in age groups in PRJNA731589 box plot, showing significant changes;
[0041] Figure 6 , (A) Line graph of the healthy group with significant changes in age groups in PRJNA731589; (B) Line graph of the healthy group with significant changes in age-specific microbiota in the validation set PRJNA763023; (C) The microbiota of different age groups in the healthy group changed significantly, and the box plot of PRJNA731589 showed obvious changes in the microbiota;
[0042] Figure 7, (A) Co abundance network of young disease group, absolute correlation higher than 0.8, significant threshold FDR<0.05; (B) For middle-aged CRC group, co abundance network, absolute correlation higher than 0.8, FDR<0.05 significant threshold; (C) For older CRC group, co abundance network, absolute correlation higher than 0.8, FDR<0.05 significant threshold; (D) Co abundance network of young healthy group, absolute correlation higher than 0.8, FDR<0.05 significant threshold; (E) Co abundance network for middle-aged healthy group, absolute correlation higher than 0.8, FDR<0.05 significant threshold; (F) Co abundance network in old healthy group, absolute correlation higher than 0.8, FDR<0.05 significant threshold;
[0043] Figure 8 , (A) Heatmap of different molars in healthy group versus CRC group; (B) KEGG age-differentiation pathways enriched in CRC group; (C) KEGG age-differentiation pathways enriched in healthy group; (D) Heatmap of species-level hub colonization with different molar associations; (E) Heatmap of genus-level hub colonization with different molar associations, The provided bar graphs illustrate the number of genes associated with each pathway, the length of the bar graphs represents the gene count, and the color gradient represents the significance of the enrichment, from less significant (blue) to more significant (red);
[0044] Fig. 9 , hierarchical histogram of (AQ)-level difference model;
[0045] Fig.10 , (AQ) Hierarchical histogram of genus-level difference moudel;
[0046] Fig.11, (A) 10-fold cross validation of the random forest model between the young CRC and young healthy groups during the training phase; (B) Circular plot of the top 30 differentially abundant markers selected as the optimal marker set for the young CRC and young healthy groups based on random forests; (C) AUC values of the random forest model for the young CRC and young healthy groups, and the error bars represent the 95% confidence intervals of the AUC values; (D) 10-fold cross validation of the random forest model between the middle-aged CRC and middle-aged healthy groups during the training phase; (E) Circular plot of the top 30 differentially abundant markers selected as the optimal marker set for the middle-aged CRC and middle-aged healthy groups based on random forests; (F) AUC values of the random forest model for the middle-aged CRC and middle-aged healthy groups, and the error bars represent the 95% confidence intervals of the AUC values ; (G) Random forest between the elderly CRC group and the elderly healthy group during the training phase, the model was cross-validated 10 times; (H) Based on random forest, the top 30 differentially abundant markers were selected as the optimal marker set for the elderly CRC group and the elderly healthy group; (I) AUC values of the random forest model for the elderly CRC group and the elderly healthy group, the error bars indicate the 95% confidence interval of the AUC value; (J) AUC values based on the random forest model for young samples (<50 years old) in the young CRC group and the young healthy group in the external independent validation set PRJNA763023; (K) AUC values based on the random forest model, applicable to middle-aged and elderly samples (>50 years old) in the external independent validation set PRJNA763023, based on the middle-aged and elderly CRC group and the middle-aged and elderly healthy group. DETAILED DESCRIPTION
[0047] The following examples are provided for a better understanding of the present invention, but are not intended to limit the best mode of implementation, nor to limit the content and protection scope of the present invention. Any product identical or similar to the present invention obtained by anyone under the inspiration of the present invention or by combining the features of the present invention with other prior arts shall fall within the protection scope of the present invention.
[0048] If no specific experimental steps or conditions are specified in the examples, the conventional experimental steps or conditions described in the literature in the field can be used. If no manufacturer is specified for the reagents or instruments used, they are all conventional reagent products that can be obtained commercially.
[0049] The present invention is further described in detail below in conjunction with specific embodiments. These embodiments should not be construed as limiting the scope of protection claimed by the present invention.
[0050] 1. Methods
[0051] 1. Data Overview
[0052] The data were classified into two large groups: normal group (N) and colorectal cancer group (C). According to age, the normal group (N) and colorectal cancer group (C) were further divided into three subgroups: young group (Y) under 50 years old, middle-aged group (M) aged 50 to 65 years old, and old group (O) aged 65 years old or more. In other words, the data were actually classified into six subgroups, namely, normal young group (N_Y), colorectal cancer young group (C_Y), normal middle-aged group (N_M), colorectal cancer middle-aged group (C_M), normal old group (N_O), and colorectal cancer old group (C_O).
[0053] The training dataset PRJNA731589 includes 12 healthy individuals in the normal young group (N_Y), 16 patients with colorectal cancer in the young group (C_Y), 52 healthy individuals in the normal middle-aged group (N_M), 36 patients with colorectal cancer in the middle-aged group (C_M), 20 healthy individuals in the normal elderly group (N_O) and 27 patients with colorectal cancer in the elderly group (C_O). For validation purposes, a separate dataset PRJNA763023 was included, which includes 100 young group individuals (50 healthy individuals and 50 colorectal cancer individuals) and 200 middle-aged and elderly group individuals (100 healthy individuals and 100 colorectal cancer individuals), as shown in Table 1.
[0054] Table 1 Overview of training set PRJNA731589 and validation set PRJNA763023 data
[0055]
[0056] Cohort: Young (Y), <50 years old; Middle-aged (M), ≥50<65 years old; Old (O), ≤65 years old.
[0057] 2. Metagenomic sequencing data analysis
[0058] The main execution software used in this example includes fastqc (v0.12), trimmomatic (v0.39) and bowtie2 (v2.3.5.1). These tools are used for quality control, filtering and comparison of data sets, respectively.
[0059] Trimmomatic was used to remove low-quality reads (SLIDINGWINDOW: 4:20MINLEN: 50); Bowtie2v2.3.5 was used to eliminate matching read lengths that might originate from host-related or laboratory-related sequences using the human reference database; MetaPhlAn v4.0.6 was used to calculate relative abundance based on the mpa_vOct22_CHOCOPhlAnSGB_202212 database; functional composition analysis was performed using the human3v3.8 and full_chocophlan.v296_201901b databases; to minimize potential false positives, taxonomic data with relative abundance values < 0.1% were filtered out from samples; for downstream analysis, the species abundance data of each dataset were integrated with the associated metadata using the Phyloseq package in R; the alpha and beta diversity indices for gut microbiota analysis were estimated based on species relative abundance profiles using the Vegan package in R and visualized using ggplot2.
[0060] 3. Identification of age-specific microbiota in CRC patients or healthy subjects
[0061] In this embodiment, statistics, machine learning and network-based methods are combined to identify potential biomarkers for colorectal cancer. Specifically, the multivariate correlation linear modeling algorithm MaAsLin2 is used to control confounding factors such as gender and body mass index (BMI), and TMM analysis is used for data standardization. Six analysis methods including t-test, Wilcoxon test, LEfSe (linear discriminant effect size analysis), Deseq2, Edger and MaAsLin2 were selected for variance analysis. According to the FDR-corrected P≤0.05 standard, microbiota with differential abundance between CRC patients and healthy individuals in different age groups were identified. In addition, LEfSe regards microbiota with LDA scores>4 and FDR-corrected P≤0.05 as significantly different. Further analysis focused on the family level and genus level.
[0062] This embodiment applies three machine learning algorithms, namely random forest RF, support vector machine SVM and generalized linear model GLM, and random forest shows excellent classification performance. In order to determine the microbial marker with the most information, five ten-fold cross-validation tests were performed using random forest, and then the critical point was selected based on the average minimum cross-validation error. The most discriminatory biomarkers were determined by averaging the average reduced accuracy of the random forest model feature importance score associated with the phenotype.
[0063] Microbial correlations were further explored within the framework of co-abundance networks using the ggCLusterNet package. Microbial correlation matrices were estimated using SparCC to construct modular microbial co-abundance networks. Correlations with FDR-corrected P values < 0.05 and magnitudes > 0.6 were selected for subsequent visualization and network analysis. The network layout was visualized by the igraph package.
[0064] 4. Species-Function Correlation
[0065] Spearman correlation coefficients were calculated for the CRC / healthy groups, and the FDR multiple testing was adjusted using the BH (Benjamini & Hochberg) method. Only correlation coefficients with an absolute value greater than 0.5 and a p value less than 0.05 were retained. Correlation heat maps were generated using the corrplot software package.
[0066] 2. Results
[0067] 1. There are significant differences in the distribution of colorectal cancer and normal groups in each age group
[0068] The bacterial richness of different cohorts was evaluated by comparative analysis of fecal microbial alpha diversity indicators. The analysis showed that the richness index of the N_M (normal middle-aged) group and the C_M (colorectal cancer middle-aged) group was significantly lower than that of the young group N_Y (normal young) group, as shown by a significant decrease in the number of observed operational taxonomic units (OTUs). On the other hand, the fecal microbial Simpson's uniformity index showed that the uniformity of the C_Y (colorectal cancer young) and C_O (colorectal cancer old) groups was significantly increased compared with the N_Y group, indicating that the overall uniformity of intestinal microorganisms in the colorectal cancer cohort was improved (results as shown in Figure 1 (as shown in A and B).
[0069] It is noteworthy that the Pielou uniformity index and the Shannon diversity index did not show significant changes between the groups. Using principal coordinates analysis (PCoA) based on Bray-Curtis dissimilarity and validated by permutational multivariate analysis of variance (PERMANOVA), it was found that there were significant differences in the distribution of microbiota structure between CRC and healthy cohorts (results as shown in Figure 1 B).
[0070] In addition, there were significant differences in the distribution between the CRC group and the normal group in each age group (results such as Figure 1 C and D). These differences suggest that each group has different gut microbial diversity and structure. This is further supported by the bacterial Venn diagram and UpSet diagram, which show that there are more unique microbial species in the healthy group compared to the CRC group, with the N_M group having the least species (results shown in Figure 1 E and F).
[0071] Figure 1 (A) Alpha diversity estimated by Pieiou odd, Shannon, Simpson odd index, p-value calculated by two-sided unpaired t-test. (B) Principal coordinate analysis (PCoA) of diseased and healthy groups based on Bray-Curtis distance showed that microbial composition was significant between groups (p = 0.001) and between cohorts (p = 0.000). The p-value of beta diversity based on Bray-Curtis distance was calculated by two-sided test using PERMANOVA. (C) Principal coordinate analysis (PCoA) based on Bray-Curtis distance of age-specific CRC groups. (D) Principal coordinate analysis (PCoA) of healthy groups of different ages based on Bray-Curtis distance. (E) Venn diagram showing overlap between groups. (F) The inverted diagram shows the number of different bacterial species in each group, shared by the combination of data sets. The numbers above each column indicate the size of the differential species. The set size on the right indicates the number of differential species in each cohort.
[0072] A powerful machine learning framework is used to evaluate the predictive power of classification models at various classification levels. The results are shown in Figure 2 The analysis showed that there was a surge in taxonomic groups in the CRC group compared with the healthy controls of the age groups, such as Bacteroidia, Bacteroidetes, Bacteroidales, Bacteroidaceae, Bacteroides, Phocaeicola, and Phocaeicola vulgatus, which increased in the CRC group relative to the healthy controls in all age groups. In contrast, the relative abundance of Gammaproteobacteria, Firmicutes, Proteobacteria, Enterobacterales, Enterobacteriaceae, Bacteroides, Prevotella copri clade A, and Klebsiella pneumoniae decreased in the CRC group compared with the healthy peers.
[0073] Figure 2(A) Phylum-level species composition of the top 15 species in each group. (B) Class-level species composition of the top 15 Bacteroidetes in each group. (C) Order-level species composition of the top 15 bacterial orders in each group. (D) Family-level species composition of the top 15 bacterial families in each group. (E) Genus-level species composition of the top 15 species in each group. (F) Species composition of the top 15 bacterial species in each group.
[0074] 2. Systematic classification characteristics of fecal microbiota in patients with colorectal cancer in different age groups
[0075] A systematic taxonomic characterization of fecal microbiota associated with colorectal cancer was performed, and only microbiota similar to more than 10% of the communities were selected. A multivariate correlation linear modeling algorithm (MaAsLin2) was used to control for gender and BMI factors, and then the trimmed mean (TMM) method of M value was used for data standardization. Six types of tests were performed: t-test, Wilcoxon test, effect size linear discriminant analysis (LEfSe), Deseq2, Edger, and MaAsLin2. These methods were used for analysis of variance. Based on the groups that frequently appeared in the test results, microbiota that were different between healthy individuals and CRC patients in different age groups were selected (results as shown in Figure 3 shown).
[0076] Figure 3 (A) Heatmap of 6 different methods analyzing the differences at the genus level between the young CRC group and the young healthy group. (B) Heatmap of 6 different ANOVA methods at the genus level between the middle-aged and elderly CRC group and the middle-aged healthy group. (C) Heatmap of 6 different ANOVA methods at the genus level between the elderly CRC group and the elderly healthy group. (D) Heatmap of 6 different ANOVA methods at the genus level between the young CRC group and the young healthy group. (E) Heatmap of 6 different methods analyzing the differences at the species level between the middle-aged CRC group and the middle-aged healthy group. (F) Heatmap of 6 different ANOVA methods at the species level between the elderly CRC group and the elderly healthy group.
[0077] Successive differential analysis revealed that 31 genera and 45 species were different in the CRC samples of the three age groups. The young age group showed biomarkers such as Parasutterella excrementihominis, Anaerotruncus colihominis, and Bacteroides cellulosilyticus; the middle-aged group showed Coprococcus comes, Fusicatenibactersaccharivorans, and Citrobacter freundii; while the elderly group suggested Blautia obeum, Roseburia intestinalis, and Bacteroides eggerthii (results shown in Figure 2). Figure 4 AF).
[0078] Figure 4 (A) LDA histogram at the species level between the juvenile CRC group (C_Y) and the juvenile healthy group (N_Y). (LDA score>3.0, p<0.05). (B) Imprint map at the genus level between the young CRC group and the young healthy group (two-sided unpaired t-test, p<0.05). (C) LDA histogram at the genus level between the middle-aged CRC group (C_M) and the middle-aged healthy group (N_M). (LDA score>3.0, p<0.05). (D) Imprint map at the genus level between the middle-aged CRC group and the middle-aged healthy group (two-sided unpaired t-test, p<0.05). (E) LDA histogram at the genus level between the old CRC group (C_O) and the old healthy group (N_O). (LDA score>3.0, p<0.05). (F) Imprint map at the genus level between the old CRC group and the old healthy group (two-sided unpaired t-test, p<0.05).
[0079] The results of the examination of the genus-level and species-level intestinal flora with the same trend and differentiation level in the PRJNA763023 dataset are shown in Tables 2 and 3, where Table 2 shows the results of the genus-level examination and Table 3 shows the results of the species-level examination. It can be seen from Tables 2 and 3 that the number of Bacteroides, Bilophila, and Fusobacterium increased in the CRC group of all age groups, and the number of Bacteroides thetaiotaomicron, Bilophila wadsworthia, and Clostridium symbiosum increased at the genus and species levels.
[0080] Table 2 Genus-level intestinal microbiota with the same trends and difference levels in the PRJNA731589 dataset and the PRJNA763023 dataset
[0081]
[0082]
[0083] Table 3 Species-level gut microbiota with the same trends and different levels in the PRJNA731589 dataset and the PRJNA763023 dataset
[0084]
[0085]
[0086] 3. Systematic classification characteristics of fecal microbiota in patients with colorectal cancer in different age groups
[0087] In this example, the fecal microbiota of colorectal cancer patients in three age groups was systematically characterized, focusing on 31 genera and 45 species. The analysis showed that six age-related bacteria in the CRC group had significant discriminatory potential (results as shown in Figure 5 A and C). The abundance of Coprococcus and Roseburia decreased with age, whereas the abundance of Fusicatenibacter, Parabacteroides, and Coprococcus cames, Fusicatenibacter saccharivorans, and Parasutterella excrementihominis increased with age. In contrast, seven bacterial indicators showed significant age-related changes in the healthy control group ( Figure 6A and C), the abundance of Coprococcus decreased with age, while the abundance of Erysipelatoclostridium, Fusicatenibacter, Lachnoclostridium, Bacteroidesnordii, Blautia obeum, and Roseburia intestinalis increased with age. Similar results were observed in the PRJNA763023 dataset (see Figure 2A and C). Figure 5 B and Figure 6 B).
[0088] Figure 5 (A) CRC groups with significant changes in age groups in PRJNA731589. (B) Line plot of CRC groups with significant changes in age-specific microbiota in validation set PRJNA763023. (C) CRC groups with significant changes in age groups in PRJNA731589 box plot showing significant changes.
[0089] Figure 6 (A) Line graph of the healthy group with significant changes in age groups in PRJNA731589. (B) Line graph of the healthy group with significant changes in age-specific microbiota in the validation set PRJNA763023. (C) The microbiota of different age groups in the healthy group changed significantly, and the box plot of PRJNA731589 showed obvious changes in the microbiota.
[0090] 4. Construction of symbiotic network
[0091] In order to understand the potential interactions between the intestinal microbial communities of different age groups and their possible role in the pathogenesis of CRC, this example performed a co-abundance association analysis. Among the groups studied, the young CRC group showed 14,048 associations in 695 species (e.g. Figure 7 A), the middle-aged CRC group had 6633 associations in 714 species (as shown in Figure 7 B), the elderly CRC group reported 10,791 associations between 752 species (as shown in Figure 7 In addition, the young healthy group showed 20,715 associations for 872 species (see Figure 2). Figure 7 D), there were 5059 associations in 622 species in the middle-aged healthy group (as shown in Figure 7 E), while there were 10,091 associations for 678 species in the elderly healthy group (as shown in Figure 7F). Overall, the networks of the younger age group appeared more complex and interconnected, while the networks of the middle-aged group were less connected. Compared with the healthy group, the correlations in the young CRC network decreased, but the correlations in the CRC networks of the middle-aged and older groups increased.
[0092] Figure 7 In (A) the Co abundance network of the young disease group, the absolute correlation is higher than 0.8, and the significant threshold FDR<0.05. (B) For the middle-aged CRC group, the Co abundance network, the absolute correlation is above 0.8, and the significant threshold FDR<0.05. (C) For the older CRC group, the Co abundance network, the absolute correlation is higher than 0.8, and the significant threshold FDR<0.05. (D) The Co abundance network of the young healthy group has an absolute correlation higher than 0.8, and the significant threshold FDR<0.05. (E) The Co abundance network for the middle-aged healthy group has an absolute correlation higher than 0.8, and the significant threshold FDR<0.05. (F) The Co abundance network in the old healthy group has an absolute correlation higher than 0.8, and the significant threshold FDR<0.05.
[0093] 5. Analysis of fecal microbial function
[0094] In the subsequent functional analysis of the fecal microbiota, the present example evaluated the changes in the Kyoto Gene and Genome Database (KEGG) gene and pathway levels and identified seventeen different modules (e.g. Figure 8 Our metagenomic differential analysis of the KEGG database elucidated significant enrichment of metabolic pathways within the CRC group. Notably, pathways including glyoxylate and dicarboxylic acid metabolism, cofactor biosynthesis, and lipopolysaccharide biosynthesis were highly enriched, as indicated by the expanded bars and red hues in the visualization, implying their importance in energy generation, molecular building block synthesis, and cellular structure establishment (e.g. Figure 8 In addition, the “Carbon metabolism” and “Glycine, serine, and threonine metabolism” pathways showed significant gene enrichment, indicating high carbon utilization and amino acid biosynthesis activities in the analyzed microbial communities.
[0095] The observed upregulation of the “Fructose and mannose metabolism” and “Amino acid biosynthesis” pathways indicates vigorous activities in carbohydrate metabolism and protein production. Moreover, the enrichment of genes in the “Tricarboxylic acid cycle (TCA cycle)” emphasizes the key role of obtaining energy through aerobic respiration. Genes related to nucleotide sugar metabolism, especially the “Amino sugar and nucleotide sugar metabolism” and “Biosynthesis of nucleotide sugars” pathways, were also significantly upregulated, emphasizing their importance in the synthesis of structural polysaccharides and glycoconjugates. In contrast, the healthy group showed fewer differentially enriched metabolic pathways, focusing on amino acid biosynthesis, carbon metabolism, phenylalanine, tyrosine and tryptophan biosynthesis, and nitrogen metabolism (such as Figure 8 C).
[0096] We identified 17 distinct modules, specifically for phosphatidylcholine (PC) biosynthesis, which were mainly attributed to different bacterial genera associated with age, such as Parasturtella, Agathobaculum, and Dorea. The genera group including Parabacteroides, Bacteroides, and Bilophila had significant biosynthetic contributions to threonine and ectoine. Candidatus Nanosynsachari played a dominant role in several biosynthetic pathways, including cobalamin biosynthesis and dermatan sulfate degradation, while Blautia was significantly involved in the pentose phosphate pathway and methionine biosynthesis. In addition, bacteria such as Parabacteroides distasonis and Lachnospiraceae contributed to respiratory complex III, among which Dysomobacter welbionis and Citrobacter freundii showed significant enrichment in dermatan sulfate and chondroitin sulfate degradation (e.g. Figure 8 D and E), which reflects the different functions and importance of different microbial communities. The differences at the genus and species levels and the corresponding hierarchical bar graphs are shown in Fig. 9 and 10 shown.
[0097] Figure 8 (A) Heatmap of different molars in healthy group versus CRC group. (B) KEGG age-differentiation pathways enriched in CRC group. (C) KEGG age-differentiation pathways enriched in healthy group. (D) Heatmap of species-level hub colonization with different molar associations. (E) Heatmap of genus-level hub colonization with different molar associations. The bars provided illustrate the number of genes associated with each pathway, the length of the bar represents the gene count, and the color gradient represents the significance of the enrichment, from less significant (blue) to more significant (red).
[0098] Fig. 9 Hierarchical histogram of the (AQ) level difference model.
[0099] Fig.10 In the figure, (AQ) hierarchical histogram of genus-level difference moudel.
[0100] In conclusion, through metagenomic differential analysis, we focused on central metabolic pathways that were abundantly expressed in CRC-associated microbial communities, elucidating their functional capabilities and possible bacterial adaptations in the intestinal ecosystem.
[0101] 6. Identification and validation of markers based on fecal microbial communities
[0102] Analyses were performed exclusively at the family and genus levels using three machine learning methods: random forest (RF), support vector machine (SVM), and generalized linear model (GLM). Random forest showed the strongest classification ability (see Fig.11 Five iterations of ten-fold cross-validation tests were performed using random forests to identify the best microbial biomarkers, and the key turning points were identified by averaging the minimum cross-validation error. The biomarkers that showed the most significant discrimination were selected using mean decreasing precision (feature importance score in random forests) and were considered to produce the least error.
[0103] Specifically, during the training of the random forest model, the target variable to be predicted (whether or not the patient has colorectal cancer) and the biomarker (intestinal flora abundance) as input are first defined. Then, a ten-fold cross-validation is set: the data set is randomly divided into 10 subsets, one subset is selected as the test set in each iteration, and the remaining nine are used as training sets. This process is repeated in five independent iterations, and 10 models are trained in each iteration and their validation errors are calculated. Next, the validation errors of each iteration are averaged to obtain the average cross-validation error of each iteration, and the results of the five iterations are averaged to obtain the average minimum cross-validation error of the entire process. In addition, the importance of each feature is evaluated, that is, its contribution to the model's predictive ability is assessed by looking at the performance of the feature in the model split node. The average decreasing accuracy of each feature in five iterations is calculated, and the best performing feature is selected as the key biomarker. Finally, the random forest model is rebuilt using these key biomarkers, and the performance of the new model is verified by further cross-validation to ensure the stability and accuracy of the selected model.
[0104] Key biomarkers corresponding to the young group include Fig.11As shown in B, including: Rombutzia, Faecalicatena, Phocaeicola, Lachnospira, Klebsiella, GGB9747, Andeckia, GGB9573, GGB3429, Corynebacterium, Another Branch, GGB3363, Odorobacter, GGB9691, Zurichobacter, GGB51441, GGB35068, Dysosmobacter, Enterobacter, Enterocloster, Lancefieldella, Eubacteriales_Family_Xlll_Incertae_Ssdis_unclassified, Candidatus_Nanosynsacchari, Fusobacterium, Anaerobic Bacillus and Mediterraneibacter.
[0105] The key biomarkers corresponding to the middle-aged group are as follows Fig.11 E, including: Lawsonibacter, Fusobacterium, Roseburia, Odorobacter, Gemini, Peptostreptococcus, Candidatus_Nanoper, Flavonoids, Porphyromonas, GGB2998, Clostridium, Enterocloster, Eubacteriales_Family_Xlll_Incertae_Ssdis_unclassified, GGB55957, GGB9561, Dysosmobacter, Streptospiraceae, GGB9559, unclassified Oscillatory Spiraceae, Streptobacter, Oscillatory Bacter and Agathobaculum.
[0106] Key biomarkers corresponding to the elderly group include Fig.11 H, including: Blautia, Peptostreptococcus, Enterocloster, Bacillus, Dorea, Micromonas, Roseburia, Anaerobic Bacillus, Candidatus_Nanoperiomorbus, GGB33928, Fusobacterium, Phocaeicola, Anaerotignum, Oscillating Bacterium, Phocea, Another Branch, Candidatus_Cibion, Vescimonas, GGB3612, Jimmyella, GGB34228, unclassified Lachnospiraceae, Butyrivibrio, Eagle's Bacteria, Massiliimalia.
[0107] The results are as follows Fig.11As shown, the results show that the AUC of the random forest model corresponding to the young group is 0.922, the AUC of the random forest model corresponding to the middle-aged group is 0.956, and the AUC of the random forest model corresponding to the elderly group is 0.949, which indicates that Fig.11 The random forest model presented had significant accuracy. In addition, the model divided the participants in the external validation set into two age groups based on age 50, with an AUC level of 0.849 for younger participants and 0.763 for middle-aged and older participants (see Fig.11 ).
[0108] exist Fig.11 Figure 2. (A) 10-fold cross validation of the random forest model between the young CRC and young healthy groups during the training phase. (B) Circular plot of the top 30 differentially abundant markers selected as the optimal marker set for the young CRC and young healthy groups based on random forest. (C) AUC values of the random forest model for the young CRC and young healthy groups, and the error bars represent the 95% confidence interval of the AUC values. (D) 10-fold cross validation of the random forest model between the middle-aged CRC and middle-aged healthy groups during the training phase. (E) Circular plot of the top 30 differentially abundant markers selected as the optimal marker set for the middle-aged CRC and middle-aged healthy groups based on random forest. (F) AUC values of the random forest model for the middle-aged CRC and middle-aged healthy groups, and the error bars represent the 95% confidence interval of the AUC values. (G) Random forest between the old CRC and old healthy groups during the training phase. The model was cross-validated 10 times. (H) The top 30 differentially abundant markers were selected as the optimal marker set for the old CRC and old healthy groups based on random forest. (I) AUC values of the random forest model for the elderly CRC group and the elderly healthy group; error bars indicate the 95% confidence interval of the AUC value. (J) AUC values of the young samples (<50 years old) of the young CRC group and the young healthy group in the external independent validation set PRJNA763023 based on the random forest model. (K) AUC values based on the random forest model, applicable to the middle-aged and elderly samples (>50 years old) in the external independent validation set PRJNA763023, based on the middle-aged and elderly CRC group and the middle-aged and elderly healthy group.
[0109] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.
Claims
1. The use of products for detecting the abundance of intestinal microbiota in the preparation of products for diagnosing or assisting in the diagnosis of colorectal cancer, It is characterized in that The intestinal microbiota is an intestinal microbiota associated with intestinal metabolism and age.
2. The use according to claim 1, It is characterized in that The patient to be diagnosed is less than 50 years old, and the intestinal microbiota includes: Rombus, Faecalibacterium, Phocaeicola, Lachnospiraceae, Klebsiella, GGB9747, Andeckia, GGB9573, GGB3429, Corynebacterium, Alternaria, GGB3363, Odorobacter, GGB9691, Zurichobacter, GGB51441, GGB35068, Odorobacter, Enterobacter, Enterocloster, Lansfieldella, Eubacteriales_Family_Xlll_Incertae_Ssdis_unclassified, Candidatus_Nanosynsacchari, Fusobacterium, Anaerobic Bacillus, and Mediterraneibacter.
3. The use according to claim 1, It is characterized in that The age of the patient to be diagnosed is greater than or equal to 50 years old and less than 65 years old, and the intestinal microbiota includes: Asha, Fusobacterium, Roseburia, Odorobacter, Gemini, Peptostreptococcus, Candidatus_Nanoper, Flavonoids, Porphyromonas, GGB2998, Clostridium, Enterocloster, Eubacteriales_Family_Xlll_Incertae_Ssdis_unclassified, GGB55957, GGB9561, Dysosmobacter, Streptospira, GGB9559, unclassified Oscillatory Spiraceae, Streptospira, Oscillobacter, and Abacterium.
4. The use according to claim 1, It is characterized in that The patient to be diagnosed is aged greater than or equal to 65 years old, and the intestinal microbiota comprises: Blautia, Peptostreptococcus, Enterocloster, Odorbacterium, Dorsalella, Micromonas, Roseburia, Anaerobic Bacillus, Candidatus_Nanoperiomorbus, GGB33928, Fusobacterium, Phocaeicola, Anaerobic Bacteria, Oscillating Bacterium, Flavonoids, Phocea, Another Cladosporium, Candidatus_Cibion, Vescimonas, GGB3612, Jimmyella, GGB34228, Unclassified Lachnospiraceae, Butyrivibrio, Eagle's Head, Massiliimalia.
5. The use according to any one of claims 1 to 4, It is characterized in that The product for detecting the abundance of intestinal microorganisms includes at least one of a specific recognition primer, a probe, an antisense oligonucleotide, an aptamer or an antibody for intestinal microorganisms.
6. A random forest model for diagnosing or assisting in the diagnosis of colorectal cancer. It is characterized in that The training process of the random forest model includes: Obtain the target gut microbiota abundance and colorectal cancer diagnosis results of at least two patients; The target intestinal microbiome abundance of each patient is used as input data, and the corresponding colorectal cancer diagnosis result is used as output data to train a random forest model; Wherein, the target intestinal microbiota is an intestinal microbiota associated with intestinal metabolism and age, and the colorectal cancer diagnosis result includes having colorectal cancer and not having colorectal cancer.
7. The random forest model according to claim 6, It is characterized in that The patient is less than 50 years old, and the target intestinal microbiota includes: Rombus, Faecalibacterium, Phocaeicola, Lachnospiraceae, Klebsiella, GGB9747, Andeckia, GGB9573, GGB3429, Corynebacterium, Alternaria, GGB3363, Odorobacter, GGB9691, Zurichobacter, GGB51441, GGB35068, Odorobacter, Enterobacter, Enterocloster, Lansfieldella, Eubacteriales_Family_Xlll_Incertae_Ssdis_unclassified, Candidatus_Nanosynsacchari, Fusobacterium, Anaerobic Bacillus, and Mediterraneibacter.
8. The random forest model according to claim 6, It is characterized in that The patient is aged greater than or equal to 50 years old and less than 65 years old, and the target intestinal microbiota includes: Asha, Fusobacterium, Roseburia, Odorobacter, Gemini, Peptostreptococcus, Candidatus_Nanoper, Flavonoids, Porphyromonas, GGB2998, Clostridium, Enterocloster, Eubacteriales_Family_Xlll_Incertae_Ssdis_unclassified, GGB55957, GGB9561, Dysosmobacter, Streptospira, GGB9559, unclassified Oscillatory Spiraceae, Streptospira, Oscillobacter, and Abacterium.
9. The random forest model according to claim 6, It is characterized in that The patient is older than or equal to 65 years old, and the intestinal microbiota includes: Blautia, Peptostreptococcus, Enterocloster, Odorbacterium, Dorsalella, Micromonas, Roseburia, Anaerobic Bacillus, Candidatus_Nanoperiomorbus, GGB33928, Fusobacterium, Phocaeicola, Anaerobic Bacteria, Oscillating Bacterium, Flavonoids, Phocea, Another Cladosporium, Candidatus_Cibion, Vescimonas, GGB3612, Jimmyella, GGB34228, Unclassified Lachnospiraceae, Butyrivibrio, Eagle's Head, Massiliimalia.
10. A system for diagnosing or assisting in the diagnosis of colorectal cancer, It is characterized in that The system includes an input device, a computing device and an output device; wherein, The input device is used to input the target intestinal microbial abundance of the patient to be diagnosed; The output device is used to output the colorectal cancer diagnosis result of the patient to be diagnosed; The computing device includes a memory and a processor; a computer program is stored in the memory, and when the computer program is run on an electronic device, the electronic device executes the random forest model as described in any one of claims 6 to 9.
