Abstract
Abstract only 3083 Background: The diagnosis of a malignancy is typically informed by clinical presentation and tumor tissue features including cell morphology, immunohistochemistry, cytogenetics, and molecular markers. However, in approximately 5-10% of cancers, ambiguity is high enough that no tissue of origin can be determined and the specimen is labeled as a Cancer of Occult Primary (CUP). Lack of reliable classification of a tumor poses a significant treatment dilemma for the oncologist leading to inappropriate and/or delayed treatment. Methods: 40,000 tumor patients with NGS data were used to construct a multiple parameter lineage-specific classification system using an advanced machine learning approach. The dataset for each classifier was split 50% for training and the other 50% for testing. The training task for each classifier was to identify the cases that were similar to the cases it was trained on against a backdrop of randomly selected cases of other histological origins. Results: Tumor lineage classifiers predicted the correct classifications where the primary site was known with accuracies ranging between 85% and 95%. When applied to CUP cases (n = 500), an unequivocal result could be obtained 100% of the time. Conclusions: Lineage predictors can render a histologic diagnosis to CUP cases that can inform treatment and potentially improve outcomes.