Integration of Bioinformatics and Machine Learning to characterize Fusobacterium nucleatum ’s pathogenicity

Zihan Tian, Píetro LióbioRxiv 2025 · DOI

Abstract

Fusobacterium nucleatum (Fn) has been found to be associated with cancer lesions in both oral and colon cancers. Although important studies have dissected the clinical aspects of its remarkable pathogenicity, there is a lack of molecular studies. This study aimed to computationally predict potential pathogenicity islands (PAIs) in F. nucleatum ATCC 25586, with candidate functional relationships among PAI-encoded proteins, and generate testable hypotheses regarding iron-dependent virulence mechanisms. We employed an integrative bioinformatics pipeline combining genomic island prediction (IslandViewer), promoter analysis (PePPER), codon adaptation index calculation, protein interaction prediction (STRING), co-expression network inference (bnlearn), structural modeling (AlphaFold-Multimer), and genome-scale metabolic modeling (CarveMe/COBRApy). Computational prediction was integrated with published literature to formulate mechanistic hypotheses. Our analysis identified three candidate genomic islands, with Region 2 (PAI2; 1,496,613 to 1,523,855 bp) exhibiting the characteristics most consistent with a functional pathogenicity island, including predicted mobile genetic elements, putative toxin-antitoxin systems, and strong promoter motifs. We propose a mechanistic hypothesis linking hemolysin-mediated iron acquisition to cancer promotion through oxidative stress and Hippo pathway modulation.

Introduction

Fusobacterium nucleatum is a prominent member of the human oral microbiota, typically existing as a commensal. However, its presence in extra-oral sites, particularly within colorectal cancer (CRC) tissues, has marked it as a significant opportunistic pathogen. Clinical evidence suggests that Fn not only colonizes tumor microenvironments but actively promotes tumor progression, modulates the immune response, and contributes to chemoresistance. Despite these critical associations, the molecular mechanisms underpinning its transition from commensal to pathogen remain poorly understood.

A key challenge in studying Fn's virulence is the identification of pathogenicity islands (PAIs)—large genomic regions acquired through horizontal gene transfer that harbor virulence factors. This study utilizes a multi-layered computational approach to map these regions and simulate the metabolic adaptations that allow Fn to thrive in the unique conditions of a tumor.

Methodology: An Integrated Pipeline

To overcome the limitations of single-tool analyses, we developed a comprehensive workflow that integrates genomic, proteomic, and metabolic data. This 16-step process moves from raw genome data to a comprehensive pathogenicity profile.

Integrated bioinformatics and machine learning pipeline for pathogenicity characterization.

The pipeline consists of several key stages:

  1. Genomic Signature Analysis: We utilized GC skew analysis to identify compositional asymmetries indicative of horizontal gene transfer.
  2. Pathogenicity Island Prediction: Tools like IslandViewer were used to predict genomic islands, which were then cross-referenced with Codon Adaptation Index (CAI) values to estimate the expression potential of encoded proteins.
  3. AI-Driven Virulence Prediction: The Biomni AI tool was employed to identify potential virulence genes within the predicted islands.
  4. Structural and Interaction Modeling: AlphaFold-Multimer was used to predict the 3D structures and potential interactions of uncharacterized proteins, while STRING provided a network-level view of protein-protein interactions.
  5. Metabolic Simulation: We constructed a genome-scale metabolic model using CarveMe and performed Flux Balance Analysis (FBA) to investigate iron acquisition and pH-dependent growth.

Genomic Island Identification and Characterization

Our analysis initially identified three candidate genomic regions in the F. nucleatum ATCC 25586 genome. Comparative GC skew analysis against E. coli and P. gingivalis helped validate these findings by highlighting regions of atypical nucleotide composition.

Comparative GC skew analysis and genomic island prediction for F. nucleatum.
  • PAI 1 (Region 1): Contains Hemolysin III (FN1885), a known virulence factor. It exhibits a high promoter-to-gene ratio (0.7), suggesting active regulation.
  • PAI 2 (Region 2): Located between 1,496,613 and 1,523,855 bp, this region is the most robust candidate for a functional PAI. It is rich in mobile genetic elements, such as transposases, and contains putative toxin-antitoxin systems.
  • PAI 3 (Region 3): While predicted by some methods, our integrated analysis suggests this region may be a misclassified metabolic regulator rather than a true pathogenicity island.

The Iron-Pathogenicity Axis in the Tumor Microenvironment

A central finding of this study is the critical role of iron acquisition in Fn's pathogenicity. Fn is an "iron-hungry" specialist, possessing a robust heme-utilization pathway (HmuUV). Metabolic modeling revealed that Fn has a 77.7% preference for ferric iron (Fe3+) over ferrous iron (Fe2+). This is biologically significant because Fe3+ is more prevalent in the acidic, hypoxic environments characteristic of solid tumors.

We propose a mechanistic link between Fn's iron scavenging and host oncogenesis. The aggressive acquisition of iron by Fn may trigger the Fenton reaction in host cells, leading to the production of reactive oxygen species (ROS). These ROS can inhibit the host's Hippo signaling pathway—a key regulator of cell proliferation—thereby promoting tumor cell survival and growth.

Key Results and Metabolic Insights

  • Virulence Factors: Hemolysin III (FN1885) emerged as a primary factor with a high CAI (0.72), indicating it is highly expressed and likely central to host cell lysis for nutrient release.
  • Growth Dynamics: FBA simulations showed that Fn growth is strictly dependent on iron availability. The predicted optimal growth rate of 67.42 mmol/gDW/hr drops to zero in the absence of iron, identifying iron metabolism as a potential therapeutic target.
  • Novel Protein Complexes: AlphaFold modeling identified a potential transmembrane complex in PAI 2 composed of previously uncharacterized "hypothetical" proteins, suggesting new modes of host-pathogen interaction.

Conclusion and Future Directions

This study provides a high-resolution molecular blueprint of Fusobacterium nucleatum’s pathogenicity. By integrating machine learning and metabolic modeling, we have identified specific genomic regions and metabolic pathways that drive its association with cancer. While these "in silico" findings are robust, they serve as a foundation for future "wet lab" experiments. Validating the proposed ROS-Hippo signaling mechanism and the function of the newly identified PAI 2 proteins could lead to novel diagnostic and therapeutic strategies for Fn-associated cancers.

Summary updated

Abstract

Abstract Fusobacterium nucleatum has been found to be associated with cancer lesions in both oral and colon cancers. Although important studies have dissected the clinical aspects of its remarkable pathogenicity, there is a lack of molecular studies. This study aimed to computationally predict potential pathogenicity islands in F. nucleatum ATCC 25586, with candidate functional relation-ships among PAI-encoded proteins, and generate testable hypotheses regarding iron-dependent virulence mechanisms. We employed an integrative bioinformatics pipeline combining genomic is-land prediction (IslandViewer), promoter analysis (PePPER), codon adaptation index calculation, protein interaction prediction (STRING), co-expression network inference (bnlearn), structural modeling (AlphaFold-Multimer), and genome-scale metabolic modeling (CarveMe/COBRApy). Computational predication was integrated with published literature to formulate mechanistic hy-potheses. Our analysis identified three candidate genomic islands, with Region 2 (PAI2; 1,496,613 to 1,523,855 bp) exhibiting the characteristics most consistent with a functional pathogenicity island, including predicted mobile genetic elements, putative toxin-antitoxin systems, and strong promoter motifs. We propose a mechanistic hypothesis linking hemolysin-mediated iron acquisi-tion to cancer promotion through oxidative stress and Hippo pathway modulation. Our work has two immediate and important benefits: the improved understanding of the biological processes that shape the pathogenicity and evolution of Fusobacterium nucleatum at the molecular level and the improved ability to integrate and automate the state-of-the-art bioinformatics tools and machine learning approaches in the inference of the mechanistic interpretability of a pathogenic phenotype.
BibTeX
@misc{tian2025integration,
  author = {Zihan Tian and Píetro Lió},
  title = {{Integration of Bioinformatics and Machine Learning to characterize <i>Fusobacterium nucleatum</i> ’s pathogenicity}},
  year = {2025},
  doi = {10.1101/2025.08.21.671586},
  howpublished = {bioRxiv},
}
Figure