P59.08 Identifying the Origin of Lung-Specific Cancer of Unknown Primary Based on Comprehensive Genomic Profiling Optimized With DNA Methylation

Abstract

Precise diagnosis of the tissue origin for metastatic carcinoma is essential to decide the treatment scheme and improve outcome. The purpose of this study is to predict the primary site for lung-specific cancer of unknown primary (CUP) through genomic and DNA methylation profiles. Comprehensive genomic profiling (CGP) was performed to determine the genomic status within individual paired samples, and targeted bisulfite sequencing was exploited to interrogate their methylation status. Model training subjects were consisted of lung cancer and another type of cancer to compute the prediction signature weights for each DMR feature. A total of 300 patients with multi-site malignancies including lung and other organ lesions were screened retrospectively. Excluding individuals clearly diagnosed with imaging modalities and IHC staining, 40 patients needed further differential diagnosis. Among them, 26 cases were identified with novel molecular events like EGFR or ALK alterations; the remaining 14 lung-specific CUPs were located with another type of caners in stomach (n=5), intestine (n=5) and cervix (n=4). Of the 14 CUP cases, mutation spectrum analyses could distinguish metastatic disease from multiple primary tumors, while it was hard to determine the origin of metastatic cancers. However, methylation profiles not only predicted the consistent results with mutation spectrum, but also identified the origin of tumors through our train classification models (AUC>0.98), which can be confirmed with patients’ clinical efficacy who received treatments guided by prediction signature. Collectively, compared with genomic profiles, DNA methylation demonstrated better tumor traceability to determine the primary site of CUPs. The signature based on within-sample CGP testing optimized with DNA methylation could exactly identify the origin of lung-specific CUPs.