Background:
Sinonasal and salivary gland tumors represent a complex group of over 40 different types of cancer, sharing an overlapping anatomical compartment and ranging from benign to highly malignant biology. Many tumors show morphologic similarities that present significant challenges to diagnosis and result in substantial diagnostic discrepancy. Further, methods to predict treatment response or to predict tumor progression or metastasis are lacking. Epigenetic classification based on DNA methylation and machine learning has demonstrated tremendous power in unifying diagnostics in brain tumors. A single array-based test can effectively classify between many different types of cancer and allows unprecedented diagnostic harmonization.
Hypothesis:
We hypothesize that DNA methylation signatures of sinonasal and salivary gland tumors are sufficiently distinct to generate machine-learning classifiers for robust classification. We further propose that implementation of decentralized DNA methylation-based classification can outperform routine pathological workup, by improving accuracy and speed of diagnosis and may identify clinically relevant tumor subtypes. Finally, we hypothesize that DNA methylation signatures can be used to predict response to treatment and possibly propensity to progress or develop recurrence/metastasis.
Aims:
The primary aim is to provide a unifying classification system for sinonasal and salivary gland tumors that harmonizes diagnostic standards across the participating centers. Secondary aims are to explore spatial and temporal changes in DNA methylation patterns to deepen understanding of tumor progression, metastasis, epigenetic regulation of intratumoral heterogeneity and explore DNA methylation in relation to clinical outcomes to predict therapy responses.
Methods:
We have assembled a preliminary DNA methylation reference cohort for sinonasal and salivary gland tumors containing all relevant differential diagnoses of this anatomical compartment. Using this data, we will develop a DNA methylation- based classifier and will decentrally validate the algorithm in six independent laboratories on retrospective cases and three clinical trial cohorts. We will further use our collaborative network to assemble relevant numbers of inverted papilloma progressing to sinonasal cancer and tumors developing recurrence/metastases to study epigenetic changes during these processes. Our large, combined dataset will further be used to perform and validate outcome analyses.
Expected results and impact:
Harmonizing diagnostic classification tools across Europe will enable quality assurance in future clinical management, facilitating the development of transnational clinical trials and improved models for tumor prognostication. More rapid diagnostics will allow earlier administration of treatment. Finally, we will gain an insight into the role of epigenetic alterations driving metastasis and intratumoral heterogeneity.