Integrative multi-omics and AI-driven digital twin modelling for HPV-associated oropharyngeal carcinoma

Background. Over the past 30 years, the number of registered cases of squamous cell carcinoma of the oropharynx has shown a steady upward trend, with 50–70% of patients having cancer associated with human papillomavirus. Purpose – to design and validate an artificial intelligence model combining multi- omics-hierarchies (DNA/RNA(Transcriptomics)/Proteomics/Lipidomics) and Digital Pathology features for improved prognostic accuracy and personalized-treatment strategies in human papillomavirus-associated oropharyngeal squamous cell carcinoma patients. Materials and Methods. As part of our study, we performed a comprehensive multi- omic characterization of 47 human papillomavirus-associated oropharyngeal squamous cell carcinoma patients by employing high-throughput DNA and RNA sequencing, mass spectrometry-based proteomics and lipidomics, and AI-driven histopathological image analysis. Molecular and morphological markers of significance were identified and incorporated into machine learning models, including deep neural networks, Random Forest, and L1 regression (L1). External validation was made by TCGA-HNSC (n = 96) and GEO GSE65858 (n = 86) datasets. Results. Overexpression of immune-system-related genes (IFNG, PD-L1, CXCL9) and inflammation proteins (S100A9, LCN2) was associated with CD8⁺ T-cell infiltration and response to therapy. The accuracy of the model increased with morphologic features (hyperchromasia and keratin pearl density). The digital twin model achieved an area under curve on the receiver operating characteristic curve of 0.91 on internal validation and preserved predictive performance in external cohorts, with an 8.3-month median predicted survival difference between responders and non-responders (p < 0.01). Conclusions. The multi-layered artificial intelligence approach based on functional analysis is capable of constructing biologically and clinically relevant digital twins of HPV-positive oropharyngeal squamous cell carcinoma, providing a scalable framework for personalized therapy prediction and enhancing the precision of cancer care.