New Models Combine AI and Molecular Data to Predict HPV Outcomes

Infection with human papillomavirus (HPV) is extremely common and, in the vast majority of cases, clears up on its own. In some people, however, it persists and can develop into precancerous lesions or even cancer. Identifying these high-risk cases at an early stage is the focus of a new study led by researchers at the University of Liège. 

For several decades, researchers have tried to identify biomarkers capable of accurately predicting the progression of infection with high-risk oncogenic HPV. Despite significant progress, no viral, immune, or molecular marker currently allows for reliable prediction of which infections will resolve on their own and which will progress to more severe disease.

This difficulty stems in part from the complexity and dynamic nature of carcinogenesis, the biological process by which a healthy cell transforms into a cancer cell, associated with HPV. The course of an infection depends on multiple factors, including viral characteristics, the host’s immune response, genetic and epigenetic factors, the microbiota, hormonal influences, and the environment. Random events, notably the accumulation of mutations linked to virus-induced genomic instability, add further complexity. “Perhaps the main challenge is no longer to discover a new biomarker, but to understand how all the biological mechanisms interact to determine the outcome of an HPV infection,” says Michael Herfs, an oncologist at GIGA-ULiège.

Faced with this complexity, a new study led by Professor Herfs asserts that approaches relying on a single biomarker have probably reached their limits. Recent advances in high-throughput molecular analysis and artificial intelligence are opening up new avenues, making it possible to simultaneously integrate vast amounts of biological data to build more comprehensive predictive models. The aim is to better distinguish between the various possible trajectories of an infection or a low-grade precancerous lesion: spontaneous regression, persistence, progression to a more serious lesion, or invasive cancer. These integrative approaches are now regarded as one of the most promising avenues for improving risk stratification.

“Artificial intelligence and integrative models could help us better understand the diversity of biological trajectories observed in patients and, ultimately, offer a more personalized risk assessment,” explains Diane Bruyère, first author of the study published in Trends in Molecular Medicine.

While these new models offer great hope, they have not yet been sufficiently validated for use in routine clinical practice. At present, care still relies on HPV testing, cytology, and histopathological analysis, supplemented by monitoring that is not highly personalized.

For the researchers, the future lies in a better overall understanding of the biology of HPV infections rather than in the search for a single indicator. Ultimately, a more accurate prediction of individual risk could make it possible to tailor patient follow-up, reduce unnecessary interventions, and improve the management of those infected.