Study of Gut Flora Changes in Patients with Colorectal Cancer by Microbiome and Machine Learning

Zeyue Yu^*

College of Life Sciences, Qingdao University, Qingdao, China, 266071

^*Corresponding author: yuzeyue190013421@126.com

Keywords: Colorectal Cancer, Gut Microbiome, Microbial Co-Occurrence Network, Metabolome, Random Forest.

Abstract: This study investigates the structural changes in the gut microbiome of colorectal cancer (CRC) patients and their association with disease progression by integrating microbiome and metabolomics data. The findings reveal that while there was no significant difference in alpha diversity (p=0.977) between CRC patients and healthy individuals, beta diversity showed distinct variations (P=0.005), indicating unique microbial community structures in CRC patients.Microbial co-occurrence network analysis demonstrated that the CRC group exhibited a more complex network structure, with greater network diameter (11 vs. 9), density, and average clustering coefficient, suggesting that tumor pressure alters microbial interaction patterns. The random forest model (AUC=0.61) identified 10 key bacterial genera as biomarkers, with Coprococcus ranking first—a butyrate-producing genus potentially associated with anti-cancer effects. Metabolomic analysis via OPLS-DA revealed significant differences in metabolites between the two groups (P=0.3852, R²=0.4178).The study highlights the ecological shift in CRC patients’ gut microbiota from a “keystone species-dominated” framework to a “collaborative structure.” Specific bacterial genera (e.g., Faecalibacterium, Streptococcus thermophilus) were found to influence CRC progression through immune modulation and metabolite secretion. These insights provide a theoretical foundation for microbiome-based CRC diagnostics and targeted therapies. Future research should focus on optimizing the random forest model and exploring the mechanistic roles of microbiome-metabolite interactions in CRC development.

1. Introduction

Findings indicate that only a relatively small proportion, specifically 10%–15%, of colorectal cancer cases have a hereditary basis. This rather striking fact serves to underscore, in a profound manner, the significant role that environmental factors play in the intricate processes of epigenetic and genetic regulation during the development of CRC [3]. In essence, it implies that a vast majority of CRC cases are influenced by external environmental elements rather than just genetic inheritance.

The gut microbiome is an incredibly highly complex microbial community that resides within the human gastrointestinal tract. It is like a bustling metropolis of microscopic life, consisting of thousands upon thousands of microbial species. These include bacteria, which come in a wide variety of shapes and functions, viruses that can have both beneficial and harmful effects, and minor eukaryotes that add another layer of complexity to this ecosystem ^[4]. When compared to any other body site, this microbiota stands out as it boasts the highest microbial density and the greatest species diversity. It’s almost as if the gut is a unique habitat that provides the perfect conditions for such a rich and diverse community of microorganisms to thrive ^[5]. Humans host a vast symbiotic microbiota, with the gut inhabitants being the dominant and most influential component. The number of these microorganisms is truly astonishing, exceeding 10 trillion. This number is an order of magnitude higher than the total number of human cells, highlighting the sheer scale of this microbial population within our bodies. Moreover, the gastrointestinal tract serves as an enormous repository of microbial genetic information. The gut microbiota encompasses over 1,500 species, which are distributed among more than 50 distinct phyla. Each of these phyla represents a different evolutionary branch, and the presence of such a wide range of phyla further emphasizes the complexity of the gut microbiome ^[6]. From a genomic perspective, the number of gut microbial genes is approximately 500 – fold that of human genes, indicating the vast genetic potential within this microbial community.

Alterations in this enteric ecosystem have been strongly linked to various human diseases, and CRC is no exception. CRC accounts for over 1.8 million new cases each year globally, making it a significant public health concern. It is also the second leading cause of cancer – related mortality, which means that a large number of people lose their lives to this disease every year ^[7]. For example, diet can have a profound impact on the development of CRC by influencing the gut flora and, in turn, the development of CRC. Studies have shown that the chronic consumption of high – fat, high – protein diets, which are often characteristic of Western – style diets, significantly increases the risk of CRC. Such diets stimulate bile secretion in the body. When bile is secreted in excessive amounts due to these types of diets, it promotes the overgrowth of specific anaerobic bacteria. These bacteria have the ability to convert bile acids into lithocholic acid and deoxycholic acid. As these secondary bile acids gradually accumulate over time, it leads to dysregulation of the gut metabolome. This dysregulation disrupts the normal metabolic processes in the gut, and ultimately, it contributes to the development of carcinogenesis ^[8].

A deeper understanding of the gut microbiome and metabolome in the context of colorectal cancer has paved the way for the development of microbiome – derived diagnostics and therapeutics. Characterizing the diversity and distribution of the human microbiota under different physiological conditions and disease states is a complex but crucial task. It involves using advanced techniques to identify and analyze the various microbial species present in the gut. Along with this, clarifying the associations between the composition of the microbial community and clinical manifestations is equally important. By doing so, we can gain insights into how changes in the gut microbiome are related to the symptoms and progression of CRC. This knowledge will undoubtedly advance the strategies for disease detection, diagnosis, and treatment, as it allows for more targeted and personalized approaches ^[9].

Numerous studies, including diverse animal models and extensive clinical investigations, have firmly established that the gut microbiome and metabolomics play crucial roles in the initiation, progression, and therapeutic response of CRC. These studies have used a variety of methods, such as genetic sequencing and biochemical analysis, to understand the complex interactions between the gut microbiota and the development of CRC.

However, previous research has predominantly focused on elucidating the pathogenic mechanisms of individual microbial species. For instance, some studies have been dedicated to proving the production of a toxin substance ^{by a particular microbial species [10]}. They use sophisticated laboratory techniques to detect and quantify these toxins and understand how they affect the cells in the gut. Other studies have been focused on revealing the impact of certain environmental factors, such as exposure to pollutants or specific dietary components, on the gut microbiome and the development of CRC ^[11]. There are also studies that focus on the association between the physiological state of a strain and the probability of developing CRC ^[12]. Despite all this research on individual aspects, the analysis of the macro perspective of microbial community structure has been neglected. In addition, it is common for current studies to be based on a one – sided approach. For example, histological studies often target only the metabolites of the gut flora ^[13][14]. This narrow focus leaves a significant gap in holistic perspectives that integrate multi – omics data to comprehensively characterize the relationship between the gut microbiota and CRC. If joint multi – omics studies were conducted from a holistic view of the microbiological structure of the gut flora, our knowledge of the associations between gut flora and CRC disease characteristics could be greatly improved.

In this study, we utilized public omics data sets (microbiome & metabolome) to investigate the association between the microbiome and metabolome. Our aim is to explore their potential role in the pathogenesis of colorectal cancer in a more in – depth and comprehensive manner. We also intend to postulate the causative mechanisms behind the identified associations, which could potentially lead to the development of new prevention and treatment strategies for CRC. 

Methods and Materials

2.1 Data collection

The OUT data and metabolomic data used in this study were derived from the research findings published by Sinha R et al. (2016) ^[15]. Fecal samples collected from CRC patients (CRC group, n = 50) and healthy volunteers (H group, n = 50) were subjected to microbiome (16S rRNA gene sequencing) and metabolome (gas chromatography-mass spectrometry, GC-MS) analyses. The datasets were analyzed individually and integrated for combined analysis using various bioinformatics approaches. For the microbiome data, an abundance screening was first conducted to eliminate outliers from the samples. As for the metabolomic data, standardization processing was performed using the MetaboAnalystR package (version 4.0) in R ^[16]. This step was aimed at minimizing biases that might arise from low classification units or insufficient sample sizes.

2.2 Statistical analysis

All statistical analyses were carried out by making use of a variety of R packages. These R packages are well – known in the field of statistical analysis for their robustness and wide – ranging functionality, which can handle complex data sets such as those obtained from microbial community studies.

The vegan package (version 2.7 – 1), a powerful and commonly used tool in ecological and microbiome research, was specifically employed to calculate the α – diversity indices of the microbial community. α – diversity indices are crucial metrics that provide insights into the richness and evenness of species within a particular sample. Specifically, the diversity() function from the vegan package was used to compute the Shannon index for each sample. The Shannon index is a widely recognized measure that takes into account both the number of different species (richness) and their relative abundances (evenness) in a community. It gives a more comprehensive understanding of the complexity of the microbial community in each sample.

Subsequently, an independent samples t – test (t.test()) was utilized to compare the differences in Shannon indices between the CRC group and the Healthy group. This statistical test is a fundamental method for determining whether there is a significant difference between the means of two independent groups. By comparing the Shannon indices of the CRC group and the Healthy group, we can gain valuable information about how the microbial diversity might be affected in the context of colorectal cancer (CRC).

In addition, the vegan package was used to calculate Bray – Curtis distances. The Bray – Curtis distance is a measure of dissimilarity between two samples, which can effectively quantify the differences in the composition of microbial communities between different samples. Based on these distances, Principal Coordinates Analysis ((PCoA) was performed. PCoA is a multivariate statistical technique that helps to visualize the relationships between samples in a lower – dimensional space. It allows us to observe patterns and groupings among the samples based on their microbial community compositions.

To assess the significance of the grouping effect in PCoA and other related analyses, the Adonis test function in the vegan package was applied to conduct statistical analysis on the intestinal microbiome data (bray, permutations = 9999). The Adonis test is a useful tool for testing whether there are significant differences in the composition of microbial communities between pre – defined groups. With 9999 permutations, this test can provide a reliable assessment of the statistical significance of the observed groupings.

For predictive modeling, the randomForest package (version 4.7 – 1.2) was used to predict the top 10 important genera. Random forest is an ensemble learning method that can handle high – dimensional data and is effective in identifying important variables. By using this package, we can predict which genera are most likely to have a significant impact on the outcome, such as the presence or absence of CRC.

All final figures were plotted using the ggplot2 package (version 3.5.1). The ggplot2 package is a popular and highly customizable data visualization tool in R. It allows for the creation of high – quality and aesthetically pleasing figures that can effectively present the results of our statistical analyses and predictive modeling.

3 Results

To evaluate the differences in bacterial diversity between the two groups, we estimate alpha diversity and beta diversity between CRC group and healthy group. Though there were little difference in the Shannon (p = 0.977) between the CRC and Healthy groups (Figure 1A). However Principal co-ordinates analysis (PCoA) plots showed a separation of the two groups (Figure 1B). The adonis results suggest that the diversity of gut microbiota could be strongly influenced by the Colorectal cancer (P = 0.005, F = 3.623, R² = 0.027).

Figure 1. Gut microbiome diversity and structure analysis. (A) α diversity differences between the Colorectal cancer (CRC) and Healthy groups, Shannon; NS, not significant. CRC, CRC patient group; Healthy, Healthy volunteer group. (B) Principal co-ordinates analysis (PCoA) of Gut Bacterial β-Diversity of CRC and Health group. CRC, CRC patient group; Healthy, Healthy volunteer group.

Figure 2. Impact of Colorectal cancer on gut microbial co-occurrence patterns. (A-B, (A) CRC group, (B) healthy group) Variations in gut microbial cooccurrence network in the and (C-H) Network topology of subnetworks inferred from trimmed microbiome abundance datasets of CRC patient group and Healthy volunteer group (contains edges, nodes, diameter, density, average clustering coefficient, average path length).

As can be seen from figs. 1 and 2, across all the samples, the microbial co-occurrence networks in the gut in the CRC group displayed greater network diameter, network density, and average clustering coefficient than those in the Healthy group. This indicates that tumor pressure complicates the connections among the gut microbiota.

Figure 3. Random forest based biomarker selection. (A) The performance of Random forest model on the external validation cohort. (B) ROC curve for classification of CRC patients and healthy volunteer. (C) Biomarker taxa are ranked in descending order of importance to the accuracy of the model.

By considering the variations in microbial species across CRC and Healthy group, the model demonstrates improved classification performance compared to prior metrics, with balanced sensitivity and specificity (Figure 3B). The ROC curve likely reflects enhanced discriminatory power, supported by the optimized biomarker panel ranked in Figure 3C. Notably, the reduction in FN (38%) and FP (31%) rates suggests better calibration between CRC and healthy cohorts. The biomarker taxa (Figure 3C) appear critical for model accuracy, though further refinement could enhance clinical applicability (Figure 3A). Additionally, our model achieved a performance of accuracy: 0.61 from CRC and healthy group (Figure.3B). Biomarker taxa are ranked in descending order of importance to the accuracy of the model (Figure.3C). Coprococcus is first and Oscillospira is the last. The relevant characters are shown in Table 1.

Table 1. Functions of representative bacterial genera

Figure 4. Metabolite differences between CRC and Health groups. (A) The orthogonal partial least square diseriminant analysis diagram (OPLS-DA) of CRC and Health groups, where the abscissa is the first principal component interpretation and the ordinate is the second principal component interpretation; (B) Volcanic map of differential metabolites in CRC and Health groups. The horizontal coordinate indicates the multiple change of metabolites in different groups, and the vertical coordinate indicates the statistical significance of the change difference in the number of metabolites.

The orthogonal partial least square diseriminant analysis diagram (OPLS-DA) shows that the two groups of samples were aggregated separately; the boundaries between CRC group and healthy group were clear and the separation was more pronounced (Figure 4A). (P = 0.3852, F = 92.5913, R² = 0.4178) Based on the variable importance (VIP) values obtained by orthogonal partial least squares analysis (OPLS-DA), combined with the results of t-test and folds of difference (FC), the differential metabolites were screened according to the criteria of VIP value > 1, P value 2 or < 0.5(Figure 4B).

4 Discussion

Our differential analysis of microbiota alpha and beta diversity revealed no significant differences in alpha diversity but significant variations in beta diversity, leading us to speculate that colorectal cancer patients harbor distinct microbial communities compared to healthy individuals.

Through microbial co-occurrence network analysis comparing gut microbial ecological networks between colorectal cancer (CRC) patients and healthy individuals, we revealed that CRC significantly alters network architecture, with the CRC network exhibiting larger-scale topology (180 nodes and 404 edges) versus the healthy group (172 nodes, 392 edges), indicating increased microbial taxa richness and enhanced interconnectivity complexity. The higher clustering coefficient in CRC further suggests it drives structural differentiation of microbial communities, increasing ecological heterogeneity and shifting from a “dominance-driven framework” (controlled by keystone species) toward a “collaborative structure”. Concurrently, CRC demonstrated greater network diameter (11 vs. 9) and longer average path length (4.107 vs. 3.246), reflecting reduced network compactness and compromised efficiency in metabolic exchange or signaling transduction. Therefore, future investigations should incorporate topology metrics (e.g., average degree, network density) to elucidate how CRC reshapes microbial interaction patterns and amplifies interspecies network complexity.

Operational Taxonomic Unit (OTU) abundance data serves as the most commonly used input feature for analyzing associations between microbial communities and host traits. Machine learning approaches based on random forests leverage OTU data to elucidate links between microbial taxa and host or environmental attributes. The ROC curve demonstrates that the random forest model in this study achieved an AUC value of 0.61, indicating that the constructed model holds significance for differentiating colorectal cancer (CRC) patients from healthy individuals. When applied to novel samples, our model can effectively differentiate between these groups and predict important bacteria genus.

The identification of cancer biomarkers is important in cancer treatment, with identifying key microbial strains and critical compounds holding promise for innovative approaches to CRC pathological assessment and therapeutic strategies ^[35]. Employing machine learning, we screened ten key microbial strains (e.g., Coprococcus) within the gut microbiota of CRC patients. These distinctive biomarkers offer potential optimization pathways for existing CRC treatment regimens. It can be seen from table 1, screened microbial species show certain characteristic changes during the process of CRC.The reduction of abundance of Prevotella and Ruminococcus could be find in the CRC group ^[36]. Furthermore, there are significantly different level of P. gingivalis between CRC patients and healthy people which were found in faeces of compared to controls ^[37]. These features can be used to distinguish the CRC group from the Healthy group. At the same time, the metabolic activities of these microorganisms are closely related to CRC. Butyrogenic microbes, such as Faecalibacterium and Coprococcus play a vital role in various gut-associated metabolisms which can maintain the normal physiological environment of the gut and play an anti-cancer role. So Butyrogenic microbes have the potential to be novel microbial therapeutics ^[38]. Streptococcus thermophilus could secreting β-Galactosidase inhibited cell proliferation and promoted apoptosis of cultured CRC cells to inhibits colorectal tumor ^[39]. Gut microbes may influence CRC progression through immune modulation. For example, Prevotella spp. and Bacteroides spp. are, respectively, positively and negatively correlated with IL-9 which is associated with inhibit tumorigenesis ^[40]. In addition, the gut flora may affect the outcome of CRC. Bacteroides fragilis promotes chemoresistance in CRC, which is a main cause of CRC treatment failure. We can undo this effect by using B. fragilis-targeting phage VA7 to selectively suppress. This suggests that precise gut microbiota manipulation can be seen as a potential method for the clinical management of CRC ^[41].

5 Conclusion

In the context of contemporary scientific research, our study delves deeply into the complex interplay within the microbial ecosystem that has been significantly perturbed by colorectal cancer (CRC). By harnessing integrated gut metagenomic and metabolomic datasets, we employ advanced bioinformatics approaches. These approaches are not just simple analytical tools but rather sophisticated methods that allow us to meticulously explore the intricate networks existing within this perturbed microbial ecosystem.

This work is of great significance as it provides a solid theoretical foundation for the development of CRC interventions. Specifically, by precisely targeting the modulation of the gut microbiota, we aim to find effective ways to prevent and treat CRC. This is not a straightforward process but involves in – depth understanding of the various components and interactions within the gut microbiota.

This study further reveals the critical and far – reaching role of alterations in gut microbial and metabolomic profiles in the pathogenesis of colorectal cancer. It is not just about identifying these alterations but also understanding how they interact with each other and contribute to the development and progression of the disease. Moreover, it underscores the interaction between the microbiome and metabolome as a highly valuable research approach. This interaction is like a hidden code that, once deciphered, can unlock new insights into CRC.

Precision modulation of microbial communities, which is informed by biomarker – driven diagnostics, offers a truly promising strategy for CRC interception and personalized therapy. Biomarker – driven diagnostics can accurately identify the specific characteristics of an individual’s CRC, and based on this, we can precisely adjust the microbial communities in the gut. This is a targeted and personalized approach that has the potential to revolutionize CRC treatment.

Future research directions should focus on more in-depth and comprehensive investigations. Regarding the Random Forest model, not only should further optimization be pursued, but its application should also be explored across various aspects of CRC research. The model should be tested on multiple datasets and its performance compared with other models to identify optimal improvement strategies.

Reference

[1] Siegel, R. L., Wagle, N. S., Cercek, A., Smith, R. A. & Jemal, A. Colorectal cancer statistics [J]. CA Cancer J Clin, 2023; 73(3): 233-254.

[2] Riedesser, J. E., Ebert, M. P. & Betge, J. Precision medicine for metastatic colorectal cancer in clinical practice [J]. Ther Adv Med Oncol, 2022, 14, 17588359211072703.

[3] J, L., X, M., D, C., S, S. & Ra, D. Genetic and biological hallmarks of colorectal cancer [J]. Genes & development, 2021; 35(11-12): 787-820.

[4] Kuziel, G. A. & Rakoff-Nahoum, S. The gut microbiome [J]. Curr Biol, 2022; 32(6): R257-R264.

[5] Methé, B. A. et al. A framework for human microbiome research [J]. Nature, 2012; 486(7402): 215-221.

[6] Jaswal, K., Todd, O. A. & Behnsen, J. Neglected gut microbiome: interactions of the non-bacterial gut microbiota with enteric pathogens [J]. Gut Microbes, 2023; 15(1): 2226916.

[7] Saeed, M. et al. Microbe-based therapies for colorectal cancer: Advantages and limitations [J]. Semin Cancer Biol, 2022; 86(Pt 3): 652-665.

[8] Zhou, E. & Rifkin, S. Colorectal Cancer and Diet: Risk Versus Prevention, Is Diet an Intervention?    [J] Gastroenterol Clin North Am, 2021; 50(1): 101-111.

[9] Kim, J. & Lee, H. K. Potential Role of the Gut Microbiome In Colorectal Cancer Progression [J]. Front Immunol, 2021, 12, 807648.

[10] Dubinsky V, Dotan I, Gophna U. Carriage of Colibactin-producing Bacteria and Colorectal Cancer Risk [J]. Trends Microbiol. 2020; 28(11): 874-876

[11] Greathouse KL, White JR, Padgett RN, et al. Gut microbiome meta-analysis reveals dysbiosis is independent of body mass index in predicting risk of obesity-associated CRC [J]. BMJ Open Gastroenterol. 2019; 6(1): e000247. 

[12] Song, Mingyang et al. “Influence of the Gut Microbiome, Diet, and Environment on Risk of Colorectal Cancer.” [J] Gastroenterology vol. 158, 2 (2020): 322-340.

[13] Chen F, Dai X, Zhou CC, et al. Integrated analysis of the faecal metagenome and serum metabolome reveals the role of gut microbiome-associated metabolites in the detection of colorectal cancer and adenoma [J]. Gut. 2022; 71(7): 1315-1325

[14] Bai X, Wei H, Liu W, et al. Cigarette smoke promotes colorectal cancer through modulation of gut microbiota and related metabolites [J]. Gut. 2022; 71(12): 2439-2450. 

[15] Sinha, R. et al. Fecal Microbiota, Fecal Metabolome, and Colorectal Cancer Interrelations [J]. PLoS One, 2016 Mar 25; 11(3): e0152126.

[16] R Core Team (2024). _R: A Language and Environment for Statistical Computing_. R Foundation for Statistical Computing, Vienna, Austria. <https://www.R-project.org/>.

[17] Yang, J. et al. Oscillospira – a candidate for the next-generation probiotics [J]. Gut Microbes, 2021 Jan-Dec; 13(1): 1987783.

[18] Chen, Y.-R. et al. High Oscillospira abundance indicates constipation and low BMI in the Guangdong Gut Microbiome Project [J]. Sci Rep 2020 Jun 9; 10(1): 9364.

[19] Mysak, J. et al. Porphyromonas gingivalis: major periodontopathic pathogen overview [J]. J Immunol Res, 2014, 2014, 476068.

[20] Liébana-García, R. et al. Gut commensal Phascolarctobacterium faecium retunes innate immunity to mitigate obesity and metabolic disease in mice [J]. Nat Microbiol, 2025 Jun; 10(6): 1310-1322.

[21] L, A. et al. Stratification of human gut microbiomes by succinotype is associated with inflammatory bowel disease status [J]. Microbiome 2024 Sep 30; 12(1): 186.

[22] Zafar, H. & Saier, M. H. Gut Bacteroides species in health and disease [J]. Gut Microbes, 2021 Jan-Dec; 13(1): 1-20.

[23] Wang, C. et al. Roles of intestinal bacteroides in human health and diseases [J]. Crit Rev Food Sci Nutr, 2021; 61(21): 3518-3536.

[24] Crost, E. H., Coletto, E., Bell, A. & Juge, N. Ruminococcus gnavus: friend or foe for human health [J]. FEMS Microbiol Rev, 2023 Mar 10; 47(2): fuad014.

[25] Zhao, S. et al. Microbial perspective of multidisciplinary collaborative weight management approach: Ruminococcus gnavus may serve as a key target for weight loss [J]. Gut Microbes, 2025 Dec; 17(1): 2442038.

[26] Deng, Q., Wong, H. M. & Peng, S. Effect of adiposity status on salivary Streptococcus mutans and Streptococcus sobrinus [J]. Oral Dis, 2024.

[27] Tett, A., Pasolli, E., Masetti, G., Ercolini, D. & Segata, N. Prevotella diversity, niches and interactions with the human host [J]. Nat Rev Microbiol, 2021; 19(9): 585-599.

[28] Gálvez, E. J. C. et al. Distinct Polysaccharide Utilization Determines Interspecies Competition between Intestinal Prevotella spp [J]. Cell Host Microbe, 2020; 28(6): 838-852.e6.

[29] Yang, C. et al. Characteristics of gut microbiota in patients with metabolic associated fatty liver disease [J]. Sci Rep, 2023; 13(1): 9988.

[30] Mi, H. T. N., Kim, H., Lee, J. S., Eser, B. E. & Han, J. Flavonoids Biotransformation by Human Gut Bacterium Dorea sp. MRG-IFC3 Cell-Free Extract [J]. J Microbiol Biotechnol, 2024; 34(6): 1270-1275.

[31] Martín, R. et al. Faecalibacterium: a bacterial genus with promising human health applications [J]. FEMS Microbiol Rev, 2023; 47(4): fuad039.

[32] Effendi, R. M. R. A. et al. Akkermansia muciniphila and Faecalibacterium prausnitzii in Immune-Related Diseases [J]. Microorganisms, 2022; 10(12): 2382.

[33] Notting, F., Pirovano, W., Sybesma, W. & Kort, R. The butyrate-producing and spore-forming bacterial genus Coprococcus as a potential biomarker for neurological disorders [J]. Gut Microbiome (Camb), 2023, 4, e16.

[34] R, Y. et al. Coprococcus eutactus, a Potent Probiotic, Alleviates Colitis via Acetate-Mediated IgA Response and Microbiota Restoration [J]. Journal of agricultural and food chemistry, 2023.

[35] Salva de Torres, C. et al. Current and Emerging Treatment Paradigms in Colorectal Cancer: Integrating Hallmarks of Cancer [J]. Int J Mol Sci, 2024 Jun 26; 25(13): 6967.

[36] Zhang, H. et al. Disease-associated gut microbiome and critical metabolomic alterations in patients with colorectal cancer [J]. Cancer Med, 2023 Jul; 12(14): 15720-15735.

[37] Kerdreux, M. et al. Porphyromonas gingivalis in Colorectal Cancer and its Association to Patient Prognosis [J]. J Cancer, 2023, 21; 14(9): 1479-1485.

[38] Singh, V. et al. Butyrate producers, ‘The Sentinel of Gut’: Their intestinal significance with and beyond butyrate, and prospective use as microbial therapeutics [J]. Front Microbiol, 2023, 13, 1103836.

[39] Li, Q. et al. Streptococcus thermophilus Inhibits Colorectal Tumorigenesis Through Secreting β-Galactosidase [J]. Gastroenterology, 2021; 160(4): 1179-1193.e14.

[40] Niccolai, E. et al. Significant and Conflicting Correlation of IL-9 With Prevotella and Bacteroides in Human Colorectal Cancer [J]. Front Immunol, 2020, 11, 573158.

[41] Ding, X. et al. Bacteroides fragilis promotes chemoresistance in colorectal cancer, and its elimination by phage VA7 restores chemosensitivity [J]. Cell Host Microbe,2025, 33(6): 941-956.e10.