Genomic classification of intrapulmonary metastasis and multiple primary lung cancer

The presentation of multifocal tumors at the time of lung cancer diagnosis exhibits either intrapulmonary metastasis (IPM) or multiple primary lung cancer (MPLC), the accurate discrimination of which is clinically important. However, current genome-based approaches are prone to platform diversity, which demands a more robust method for clinical use. Here, we developed MeTel, a novel Bayesian probabilistic model for rigorous and unbiased classification, comparing normalized likelihoods based on genomic profiles. Test on 279 multifocal lung cancers (204 MPLC and 75 IPM) from six independent cohorts, MeTel showed highest accuracy (97.49%) that outperformed previous methods (82.08-95.70%), with best platform consistency (minimum accuracy 95.41% vs. 9.09-81.82%) across various panel sizes (22-605 genes and whole-exome). Application to 12 in-house patients led to re-adjucation of original diagnosis in four patients, all towards to MeTel prediction. Our results verify the utility of genomic information in the classification of multifocal lung cancers, and the applicability to current diagnostic practices.