Site-specific oral microbiome patterns in Puerto Rican head and neck cancer patients

Introduction: Head and neck cancers (HNCs) are the seventh most common cancer globally and the sixth most prevalent in Puerto Rico. They are also the eighth leading cause of cancer deaths among Puerto Rican men. Understanding HNC development and prognosis in this population is vital. The oral microbiome plays a significant role in HNC, as previous studies have shown that microbial dysbiosis can affect diagnosis and treatment. However, little is known about the oral microbiome patterns in Puerto Rican HNC patients and their stability in response to clinical factors. Therefore, the objective of this study is to determine the bacterial communities associated with HNC in Puerto Rican patients. Methods: Genomic DNA was extracted from 79 histologically confirmed HNC tumor samples from Puerto Rican patients, followed by HPV genotyping and microbiome analyses using 16S rRNA gene amplicon sequencing. Data were deposited in Qiita/Deblur for quality control and bioinformatics analyses. Downstream analyses, including alpha and beta diversity, taxonomic characterization, and biomarker analyses, were conducted using QIIME2, MicrobiomeAnalyst, and Random Forest. Results: Although beta diversity was similar across different anatomic sites, a significant difference was observed between the larynx and the oropharynx (P = 0.041, FDR = 0.355). The larynx consistently demonstrated lower richness and Shannon diversity compared to the oropharynx, hypopharynx, and oral cavity. Furthermore, Random Forest models showed weak discriminatory power (MDA < 0.05) for all six metadata variables analyzed: anatomic location, surgery, HPV status, radiotherapy, and chemotherapy. This was observed despite the contributions from various phyla, including Bacillota_A , Deinococcota, Pseudomonadota , and Bacteroidota , as well as genera such as COE1, Duncaniella, Escherichia, Paramuribaculum , and Aggregatibacter . Conclusions: The oral microbiome of Puerto Rican HNC patients appears highly resilient, with significant variations in richness and diversity primarily linked to the laryngeal anatomical site. To enhance predictive accuracy for overall, cancer-specific, and disease-free survival, future studies will systematically correlate the microbiome's composition with clinical HNC characteristics. This will be accomplished by employing an array of survival-based machine learning methodologies, including penalized Cox regression (Elastic Net), Random Survival Forests, and Gradient Boosted Survival models (XGBoost-Cox). This work promises to unlock the microbiome's potential as a powerful tool for improving patient prognoses.