Whole Genome HPV Liquid Biopsy for HPV-Associated Cancer Detection and Viral Physical State Classification

Abstract

Purpose: HPV-associated carcinomas(HPV+cancers) account for 5% of all cancers. Circulating tumor HPV DNA(ctHPVDNA) assays for HPV+ cancer surveillance have limited prognostic utility at the time of cancer diagnosis. While HPV integration into the host genome is a proven tissue-based biomarker predicting poor clinical outcomes, existing clinically utilized ctHPVDNA assays cannot classify the viral physical state. Methods: We previously developed HPV-DeepSeek, a multi-feature HPV whole-genome sequencing liquid biopsy with 99% diagnostic accuracy at the time of HPV+ oropharynx cancer diagnosis. We test the diagnostic accuracy of HPV-DeepSeek in a cohort of 235 HPV+ cancers across nine anatomic sites and employ a novel blood-based computational classifier to infer HPV genome physical state from plasma, termed HPV-SIGNAL, to assess its prognostic potential. Results: HPV-DeepSeek demonstrated a sensitivity and specificity of 99%. In 189 eligible samples, HPV-SIGNAL identified four viral physical states: episomal-only(N=71), episomal-rearranged(N=52), integrated-mixed(N=57), and integrated-clonal(N=9), which were confirmed and further elucidated via three orthogonal tissue and blood approaches. Patients harboring integrated viral states in the blood exhibited significantly worse age- and AJCC stage-adjusted progression-free survival(HR 2.78, 95% CI1.30–5.99, p=0.009) and overall survival(HR2.95, 95% CI 1.01–8.61, p=0.048) compared to patients with episomal states. Conclusion: HPV whole-genome sequencing liquid biopsy has high diagnostic accuracy across HPV+ cancer types and can be used to identify and classify HPV physical state from blood. Patients with integrated viral states detected in the blood demonstrated worse progression-free and overall survival, suggesting blood-based HPV physical state classification could be used as a prognostic tool at the time of cancer diagnosis.