Predicting HPV/p16 Status and Molecular Subtypes from Histopathology Slides using Deep Learning in Vulvar Squamous Cell Carcinoma
Journal:
medRxiv
Published Date:
Oct 24, 2026
(1)
Abstract
Vulvar squamous cell carcinoma (VSCC) comprises three prognostically distinct molecular subtypes: human papillomavirus (HPV)-associated, HPV-independent/p53-wildtype, and HPV-independent/p53-abnormal. Classification requires p16/p53 immunohistochemistry, which consumes tissue and is often incomplete in retrospective cohorts. We investigated whether deep learning could predict HPV/p16 status and molecular subtypes directly from routine hematoxylin and eosin-stained whole-slide images. Using features extracted with H-optimus-1, we developed two attention-based multiple-instance learning models: HERA, predicting HPV/p16 status, and HERMES, predicting the three molecular subtypes. Models were developed on 644 patients from three Dutch VSCC cohorts and tested on 286 patients from two independent cohorts. In independent testing, HERA achieved an area under the curve (AUC) of 0.84 (95% CI 0.75-0.91), with specificity 0.93, negative predictive value 0.94, and sensitivity 0.68. HERA-predicted groups reproduced the prognostic separation of HPV testing (log-rank p<0.001). HERMES achieved a macro-average AUC of 0.70 (0.66-0.75), with good discrimination for HPV-associated (0.80) but limited separation of HPV-independent/p53-abnormal (AUC 0.68) and HPV-independent/p53-wildtype (AUC 0.63). HERMES-predicted classes reproduced the prognostic hierarchy, with 5-year recurrence rates of 14.0% (HPV-associated), 31.1% (HPV-independent/p53-wildtype) and 43.3% (HPV-independent/p53-abnormal) (p=0.0051). Highly attended tiles showed basaloid cells and koilocytotic-like atypia in HPV-associated tumors, keratinizing nests with desmoplastic stroma in HPV-independent/p53-abnormal tumors, and well-differentiated keratinocytes with spongiosis in HPV-independent/p53-wildtype tumors. HERA could assign HPV/p16 status to large retrospective VSCC cohorts lacking ancillary testing, without consuming tissue, enabling HPV-stratified outcome analyses. Before such use, external validation of its discrimination and calibration across scanners and populations is required. HERMES suggests that molecular subtype information is encoded in hematoxylin and eosin-stained slides, but requires improved distinction between the HPV-independent subtypes.
