摘要：Diagnosis of lung cancer can sometimes be challenging and is of major interest since effective molecular-guided therapies are available. Compounds of tobacco smoke may generate a specific substitutional signature in lung, which is the most exposed organ. To predict whether a tumor is of lung origin or not, we developed and validated the EASILUNG (Exome And SIgnature LUNG) test based on the relative frequencies of somatic substitutions on coding non-transcribed DNA strands from whole-exome sequenced tumors. Data from 7,796 frozen tumor samples (prior to any treatment) from 32 TCGA solid cancer groups were used for its development. External validation was carried out on a local dataset of 196 consecutive routine exome results. Eight out of the 12 classes of substitutions were required to compute the EASILUNG signature that demonstrated good calibration and good discriminative power with a sensitivity of 83% and a specificity of 72% after recalibration on the external validation dataset. This innovative test may be helpful in medical decision-making in patients with unknown primary tumors potentially of lung origin and in the diagnosis of lung cancer in smokers.
