Title: Comprehensive Source Tracking of Human Microbiome Exchange Patterns Across Body Sites Using the FEAST Algorithm

Authors: 小雯, opencode (glm-5), zd200572

Abstract: The human microbiome plays a critical role in health and disease, with distinct microbial communities inhabiting various body sites. Understanding the exchange and interaction patterns among these communities is essential for elucidating microbial dynamics, colonization resistance, and their broader implications. This study employed the Fast Expectation-maximization microbial Source Tracking (FEAST) algorithm to quantitatively estimate the contribution of microbial sources from different body sites to target (sink) communities, utilizing 16S rRNA gene amplicon sequencing data from the Human Microbiome Project (HMP). Our analysis revealed intricate microbiome exchange patterns characterized by notable sex-specific differences. In male participants, a high bidirectional similarity was observed between the skin and nasal microbiomes (~58% reciprocal contribution), suggesting frequent microbial exchange driven by anatomical proximity and shared environmental exposures. The salivary microbiome also showed a substantial contribution from the nasal cavity (~35%). For female participants, a striking finding was the profound similarity between the vaginal and skin microbiomes (76.12% contribution from skin to vagina), indicating the skin as a primary source for vaginal colonization, potentially influenced by local anatomical contiguity. Consistent with male samples, the skin and nasal microbiomes in females also exhibited high bidirectional exchange. Furthermore, the skin emerged as a prominent multi-site source in females, contributing significantly to both vaginal and gut microbiomes. While core similarities—such as skin-nasal and saliva-nasal interactions—were conserved across sexes, distinct gender-specific ecological dynamics in overall source contributions were evident. These findings underscore the highly interconnected nature of the human microbiome, highlighting specific exchange routes and emphasizing the need to consider sex as a critical biological variable in microbiome research.

Keywords: Human Microbiome Project, FEAST, Microbiome Exchange, Source Tracking, 16S rRNA, Microbial Ecology, Sex Differences

1. Introduction The human body is colonized by trillions of microorganisms, collectively known as the human microbiome (Sender et al., 2016; Turnbaugh et al., 2007; The Human Microbiome Project Consortium, 2012). These microbial communities reside in diverse, specialized niches such as the gastrointestinal tract, skin, oral cavity, nasal passages, and urogenital tract, each characterized by distinct selective pressures (Cho & Blaser, 2012; Ley et al., 2008). These microbial communities are not isolated entities; they are dynamic ecosystems influenced by host genetics, environmental factors, lifestyle, and, crucially, continuous inter-site microbial dispersal (Costello et al., 2009; Lax et al., 2014). Understanding the sources and sinks of these communities—specifically, how microbes translocate, compete, and establish across different body sites—is fundamental to comprehending microbiome assembly, maintaining colonization resistance against pathogens (Belkaid & Hand, 2014), and elucidating the mechanisms of dysbiosis in disease states. Recent longitudinal studies from the Integrative Human Microbiome Project (iHMP) further highlight that microbiome dynamics correlate across body sites, suggesting systemic dynamics influenced by complex host-microbial-environment interactions (Zhou et al., 2024; Lloyd-Price et al., 2017; The Integrative HMP Research Network Consortium, 2019).

Traditional microbial ecology studies have predominantly focused on characterizing individual body site microbiomes in isolation. However, the inherent interconnectedness of these habitats necessitates advanced computational and analytical approaches to decipher their complex interaction networks. Microbial source tracking algorithms provide a powerful statistical framework to quantitatively estimate the proportional contributions of various potential source environments to a mixed, target microbial community (sink). While early methods like SourceTracker relied on Bayesian approaches (Knights et al., 2011), the Fast Expectation-maximization microbial Source Tracking (FEAST) algorithm, developed by Shenhav et al. (2019), represents a significant advancement. FEAST leverages an expectation-maximization (EM) framework, offering superior computational efficiency and the ability to accurately infer source contributions even when dealing with low-biomass sources or complex mixtures.

The Human Microbiome Project (HMP) has generated unprecedented, comprehensive datasets detailing the composition of microbial communities across multiple body sites in healthy individuals, serving as an invaluable resource for exploring these inter-site ecological relationships (The Human Microbiome Project Consortium, 2012; Methé et al., 2012; Caporaso et al., 2011). By applying FEAST to HMP data, we aim to map the natural flow and establishment patterns of microbial populations within the human host.

Furthermore, emerging evidence increasingly highlights the significant impact of biological sex on the human microbiome, leading to the conceptualization of the 'microgenderome'. This encompasses sex-specific differences in microbial community composition, diversity, and function, which are profoundly intertwined with host endocrine and immune system biology (Markle et al., 2013; He et al., 2022; Yoon & Kim, 2021). Consequently, evaluating microbiome exchange patterns through a sex-stratified lens is essential for a holistic understanding of human microbial ecology.

This study aims to perform a comprehensive, pairwise source tracking analysis of microbial exchange patterns among distinct human body sites using 16S rRNA gene amplicon sequencing data derived from the HMP. Specifically, we sought to: 1) quantitatively estimate microbial contributions between major anatomical sites; 2) identify prominent source-sink relationships and potential dominant microbial exchange routes; and 3) investigate sex-specific differences in these microbial interaction networks. The findings will provide deeper insights into the spatial dynamics of the human microbiome and establish a foundation for future research.

2. Materials and Methods

2.1. Data Source and Preprocessing The study utilized publicly available 16S rRNA gene amplicon sequencing data from the Human Microbiome Project (The Human Microbiome Project Consortium, 2012). Two datasets representing distinct cohorts were analyzed: a Male HMP subset encompassing four body sites (Saliva, Skin, Nose, Stool) and a Female HMP subset encompassing five body sites (Skin, Nose, Saliva, Stool, Vagina). All samples were sequenced targeting the V3-V5 variable regions of the 16S rRNA gene utilizing the Illumina sequencing platform.

The raw Operational Taxonomic Unit (OTU) tables were subjected to a standardized preprocessing pipeline optimized for microbiome data (Bokulich et al., 2013; Callahan et al., 2016). This included rigorous quality control to remove low-quality reads and singletons, transposition to format the data appropriately, and rarefaction to the minimum sequencing depth across all included samples. Finally, OTU relative abundances were converted to integer counts using the ceiling function.

2.2. FEAST Algorithm for Microbial Source Tracking FEAST (Fast Expectation-maximization microbial Source Tracking) was employed to quantify source tracking dynamics. FEAST is an efficient algorithm designed to estimate the proportional contributions of multiple known sources to a given sink community (Shenhav et al., 2019). The core principle involves iteratively calculating the posterior probability of each observed OTU originating from a specific defined source and subsequently updating the estimated mixture proportions of each source contributing to the sink.

2.3. Pairwise Cross-Source Tracking Analysis Strategy To systematically evaluate inter-site relationships, a pairwise cross-source tracking analysis was performed independently for the Male and Female HMP datasets. In this analytical design, each specific body site was iteratively designated as the target "Sink" community. Concurrently, all other body sites within the same dataset were treated as potential contributing "Sources".

2.4. Computational Environment and Software All statistical analyses and source tracking computations were conducted within the R statistical computing environment (R Core Team, 2021). The FEAST R package was utilized alongside community ecology packages such as vegan (Oksanen et al., 2020) and phyloseq (McMurdie & Holmes, 2013). Python scripts were employed for the subsequent visualization and formatting of the FEAST output matrices. An opencode agent provided substantial computational assistance in orchestrating the scripts and managing the data workflow.

3. Results

3.1. Overview of Analysis Results This study conducted rigorous pairwise cross-microbial source tracking analyses on Male and Female Human Microbiome Project (HMP) datasets, facilitating a granular assessment of microbial exchange patterns across distinct anatomical niches.

3.2. Male HMP Source Tracking Results The FEAST analysis of male participants unveiled distinct, quantifiable patterns of microbial exchange among the Saliva, Skin, Nose, and Stool communities.

The proportions matrix (Table 1) details the estimated percentage contributions of each anatomical source to the corresponding sink.

Table 1: Male HMP Microbial Source Contribution Matrix (%)

Key Findings:

• High Reciprocal Similarity between Skin and Nose: The most prominent feature was a strong, bidirectional similarity between the skin and nasal microbiomes (~58% reciprocal).
• Substantial Nasal Contribution to Saliva: Approximately one-third (34.74%) of the salivary microbiome composition was explained by the nasal microbiome.
• Moderate Stool and Skin Similarity: The stool microbiome exhibited a moderate compositional similarity with the skin (20.73%).
• Minimal Interaction Between Distant Sites: Contributions between anatomically distinct sites were minimal (<1%).

Figure 1: Visualization of source tracking results for Male participants.

3.3. Female HMP Source Tracking Results The analysis of female participants incorporated an additional critical body site, the Vagina.

This inclusion provided a broader perspective on female-specific microbial exchange networks (Table 2).

Table 2: Female HMP Microbial Source Contribution Matrix (%)

Key Findings:

• Vagina and Skin: The Highest Observed Similarity: The most striking finding was the exceptionally high estimated contribution from the skin to the vaginal microbiome (76.12%).
• Conserved Nose and Skin Interconnectedness: Consistent with the male cohort, the nasal and skin microbiomes demonstrated substantial bidirectional similarity.
• Skin as a Central Multi-Site Source Node: The skin microbiome emerged as a highly prominent, central source node in females, contributing significantly to the stool microbiome (31.44%).

Figure 2: Visualization of source tracking results for Female participants, highlighting the strong skin-to-vagina contribution.

3.4. Comparison Between Male and Female HMP Results While fundamental patterns of microbial exchange were conserved across both male and female HMP datasets, highly notable sex-specific divergence was evident. Both sexes exhibited robust bidirectional similarity between the skin and nasal microbiomes and a strong oral-nasal axis. However, female microbiomes generally presented a more highly connected network, largely driven by the dominant role of the skin, functioning as a multi-site source impacting vaginal, nasal, and gut communities.

4. Discussion This study leveraged the robust FEAST algorithm to perform a comprehensive, quantitative source tracking analysis of human microbiome exchange patterns across distinct anatomical niches. Our findings illuminate the profound interconnectedness of microbial communities within the human host and reveal critical, previously underappreciated sex-specific divergences.

The observation of consistent, high-level bidirectional similarity between the skin and nasal microbiomes in both sexes is a foundational finding. This strong inter-site relationship is logically driven by their immediate anatomical proximity and shared continuous exposure to the external environment, validating established concepts of topographical diversity (Grice et al., 2009; Byrd et al., 2018; Proctor & Relman, 2017; Findley et al., 2013).

The substantial, consistent contribution of the nasal microbiome to the salivary microbiome provides compelling quantitative evidence for a robust oral-nasal microbial axis. This dynamic is primarily facilitated by the anatomical continuity of the mucosal surfaces, allowing physiological drainage of nasal secretions directly into the oral cavity (Dewhirst et al., 2010; Man et al., 2017; Segata et al., 2012).

Perhaps the most striking and clinically relevant discovery is the exceptionally high estimated contribution from the skin to the vaginal microbiome exclusively observed in female participants. This predominantly unidirectional influence strongly implies that the skin microbiome serves as the primary reservoir for the microbial communities establishing within the vaginal tract, expanding upon previous understandings of vaginal microbiota dynamics (Ravel et al., 2011; Gajer et al., 2012; Srinivasan et al., 2012).

Furthermore, the skin in female participants emerged as a highly connected, multi-site source hub, even demonstrating an intermediate contribution to the gut microbiome. While the gut has its own strong biogeography (Donaldson et al., 2016), the influence of the skin suggests broader systemic ecological dispersal mechanisms. Such wide-ranging seeding is often seen in early life events (Dominguez-Bello et al., 2010; Oh et al., 2014) but our data highlights its persistence or signature in adults.

While the application of the FEAST algorithm yields powerful quantitative insights, it is crucial to interpret these findings within the context of inherent methodological limitations. FEAST quantifies compositional similarity based on shared OTU profiles; it does not directly prove actual, physical microbial flow.

Future research directions should focus on incorporating robust longitudinal sampling designs to track how these microbial exchange networks shift over time, during illness, or in response to interventions (Zhou et al., 2024). Advancing beyond 16S amplicon sequencing to employ shotgun metagenomics will provide critical functional insights.

5. Conclusion This study applied the FEAST source tracking algorithm to quantitatively map the intricate microbial exchange networks operating across distinct human body sites. We identified highly conserved, robust bidirectional microbial exchange between the skin and nasal microbiomes across both sexes, alongside a significant contribution from the nasal cavity to the salivary microbiome. A pivotal, sex-specific finding was the overwhelming dominance of the skin microbiome as a compositional source for the vaginal microbiome in females. These findings quantitatively demonstrate the profound interconnectedness of the human microbiome and emphasize that microbial communities are participants in continuous, notably gender-specific ecological exchange networks.

Acknowledgements: The authors express their gratitude to the Human Microbiome Project (HMP) Consortium for generating these invaluable datasets. We also acknowledge the crucial computational assistance provided by the opencode agent.

References:

1. Belkaid, Y., & Hand, T. W. (2014). Role of the microbiota in immunity and inflammation. Cell, 157(1), 121-141.
2. Bokulich, N. A., et al. (2013). Quality-filtering vastly improves diversity estimates from Illumina amplicon sequencing. Nature Methods, 10(1), 57-59.
3. Byrd, A. L., Belkaid, Y., & Segre, J. A. (2018). The human skin microbiome. Nature Reviews Microbiology, 16(3), 143-155.
4. Callahan, B. J., et al. (2016). DADA2: High-resolution sample inference from Illumina amplicon data. Nature Methods, 13(7), 581-583.
5. Caporaso, J. G., et al. (2011). Moving pictures of the human microbiome. Genome Biology, 12(5), R50.
6. Cho, I., & Blaser, M. J. (2012). The human microbiome: at the interface of health and disease. Nature Reviews Genetics, 13(4), 260-270.
7. Costello, E. K., et al. (2009). Bacterial community variation in human body habitats across space and time. Science, 326(5960), 1694-1697.
8. Dewhirst, F. E., et al. (2010). The human oral microbiome. Journal of Bacteriology, 192(19), 5002-5017.
9. Dominguez-Bello, M. G., et al. (2010). Delivery mode shapes the acquisition and structure of the initial infant microbiome across multiple body habitats. PNAS, 107(26), 11971-11975.
10. Donaldson, G. P., Lee, S. M., & Mazmanian, S. K. (2016). Gut biogeography of the bacterial microbiota. Nature Reviews Microbiology, 14(1), 20-32.
11. Findley, K., et al. (2013). Topographic diversity of fungal and bacterial communities in human skin. Nature, 498(7454), 367-370.
12. Gajer, P., et al. (2012). Temporal dynamics of the human vaginal microbiota. Science Translational Medicine, 4(132), 132ra52.
13. Grice, E. A., et al. (2009). Topographical and temporal diversity of the human skin microbiome. Science, 324(5931), 1190-1192.
14. He, Y., et al. (2022). Shedding light on biological sex differences and microbiota–gut–brain axis. Microbiome, 10(1), 1-17.
15. Knights, D., et al. (2011). Bayesian community-wide culture-independent microbial source tracking. Nature Methods, 8(9), 761-763.
16. Lax, S., et al. (2014). Longitudinal analysis of microbial interaction between humans and the indoor environment. Science, 345(6200), 1048-1052.
17. Ley, R. E., et al. (2008). Worlds within worlds: evolution of the vertebrate gut microbiota. Nature Reviews Microbiology, 6(10), 776-788.
18. Lloyd-Price, J., et al. (2017). Strains, functions and dynamics in the expanded Human Microbiome Project. Nature, 550(7674), 61-66.
19. Man, W. H., de Steenhuijsen Piters, W. A., & Bogaert, D. (2017). The microbiota of the respiratory tract: gatekeeper to respiratory health. Nature Reviews Microbiology, 15(5), 259-270.
20. Markle, J. G., et al. (2013). Sex differences in the gut microbiome drive hormone-dependent regulation of autoimmunity. Science, 339(6123), 1084-1088.
21. McMurdie, P. J., & Holmes, S. (2013). phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PLoS ONE, 8(4), e61217.
22. Methé, B. A., et al. (2012). A framework for human microbiome research. Nature, 486(7402), 215-221.
23. Oh, J., et al. (2014). Biogeography and individuality shape function in the human skin metagenome. Nature, 514(7520), 59-64.
24. Oksanen, J., et al. (2020). vegan: Community Ecology Package. R package version 2.5-7.
25. Proctor, D. M., & Relman, D. A. (2017). The landscape ecology and microbiota of the human nose, mouth, and throat. Cell Host & Microbe, 21(4), 421-432.
26. R Core Team (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.
27. Ravel, J., et al. (2011). Vaginal microbiome of reproductive-age women. PNAS, 108(Supplement 1), 4680-4687.
28. Segata, N., et al. (2012). Composition of the adult digestive tract bacterial microbiome based on seven mouth surfaces, tonsils, throat and stool samples. Genome Biology, 13(6), R42.
29. Sender, R., Fuchs, S., & Milo, R. (2016). Revised estimates for the number of human and bacteria cells in the body. PLoS Biology, 14(8), e1002533.
30. Shenhav, L., et al. (2019). FEAST: fast expectation-maximization microbial source tracking. Nature Methods, 16(7), 627-632.
31. Srinivasan, S., et al. (2012). Bacterial communities in women with bacterial vaginosis: high resolution phylogenetic characterization and core microbiome analysis. PLoS ONE, 7(6), e37818.
32. The Human Microbiome Project Consortium. (2012). Structure, function and diversity of the healthy human microbiome. Nature, 486(7402), 207-214.
33. The Integrative HMP (iHMP) Research Network Consortium. (2019). The Integrative Human Microbiome Project. Nature, 569(7758), 641-648.
34. Turnbaugh, P. J., et al. (2007). The human microbiome project. Nature, 449(7164), 804-810.
35. Yoon, K., & Kim, N. (2021). Roles of sex hormones and gender in the gut microbiota. Journal of Neurogastroenterology and Motility, 27(3), 314-325.
36. Zhou, W., et al. (2024). Longitudinal profiling of the microbiome at four body sites reveals core stability and individualized dynamics during health and disease. Cell Host & Microbe, 32(4), 506-526.e9.