Computational Genomic and Systems-Level Analysis of Fusobacterium nucleatum and Bacillus amyloliquefaciens in Oral Squamous Cell Carcinoma and their Associations with TP53 and NOTCH1 Pathways
Year: 2026
Type: preprint
Abstract: Background A significant cancer burden, especially in South and Southeast Asia, is oral squamous cell carcinoma (OSCC). It is unclear how TP53 and NOTCH1 mutations relate to the oral microbiota, although they represent different genetic subgroups of OSCC. Methods Comparative pan-genome/ortholog analysis (DIAMOND reciprocal best hit), 16S rRNA amplicon metagenomic profiling of matched OSCC tumour and adjacent-normal tissue (PRJNA822685; n = 81 patient pairs), DIAMOND-based protein mimicry screening, PSORTb subcellular localisation prediction, HMMER/Pfam-A domain annotation, DESeq2 differential abundance analysis, protein-protein docking with cross-validation across two independent platforms (HDOCK and ClusPro), PICRUSt2 functional prediction, and bipartite host-microbe network construction are integrated into this multi-omics computational framework. As a non-pathogenic comparator, Fusobacterium nucleatum (Fn) was profiled against Bacillus amyloliquefaciens (Ba, strain FZB42). Results Reciprocal-best-hit analysis identified 603 shared orthologs between Fn and Ba (11.7% of the combined gene-cluster set), with 1,413 Fn-unique (27.5%) and 3,121 Ba-unique (60.8%) genes. 16S rRNA sequencing of 81 matched OSCC patient pairs (162 samples; PRJNA822685) showed that Fusobacterium was the most significantly enriched genus in tumour tissue (log₂FC = + 2.30, Wald stat = 8.85, p = 8.62×10⁻¹⁹, padj = 1.01×10⁻¹⁶ by DESeq2 with paired design). Weighted UniFrac PERMANOVA confirmed significant community-level separation between tumour and normal tissue (pseudo-F = 9.59, p = 0.001). Mimicry screening found no hits meeting the pre-specified criteria (≥ 20% identity, ≥ 30% query coverage, E ≤ 0.001); an unfiltered sensitive-mode search recovered four sub-threshold ankyrin-repeat hits to NF-κB/NOTCH1. PSORTb predicted 57.5% cytoplasmic, 19.7% cytoplasmic-membrane, 18.9% unknown, 2.3% outer-membrane, 1.4% periplasmic, and 0.1% extracellular localisation (χ²=2894.2, p < 0.0001). Pfam-A domain analysis identified 1,474 TPR, 1,471 AAA-ATPase, 499 HTH, 235 MORN_2, and 150 ankyrin-repeat hits, with zero PKD or FN3 domains detected. HDOCK docking predicted confident interactions (confidence > 0.83) for CABKKDON_00786 vs NF-κB p50 and NOTCH1, and CABKKDON_01826 vs NOTCH1, independently confirmed by ClusPro (Cluster 0: 56–94 members). Network analysis revealed NOTCH1 and NF-κB p50 as hub nodes each receiving predicted interactions from two candidate mimicry proteins (5 nodes, 4 edges). Conclusions These findings, drawn from real patient sequencing data across nine integrated analytical layers, support a mechanistic model in which Fn contributes to OSCC pathobiology via tumour-selective colonisation, predicted protein mimicry of NF-κB and NOTCH1 signalling components, and measurable community-level dysbiosis, providing a defensible computational foundation for targeted biochemical validation.
Source: Research Square
Institution: Vellore Institute of Technology University
Language: English
Subfield: Periodontics
Field: Dentistry
Domain: Health Sciences
Open Access status: green
Funder: VIT University
OpenAlex ID: W7225920849
