Fig. 1

Study design and classification models based on variants detected in plasma cfDNA after filtering with matched WBC sample for shared variants. a Schematic view of the study design. b Pearson correlation of AF in cfDNA (x-axis, log scale) and AF in matched WBC gDNA (y-axis, log scale). Each point represents one variant detected in matched cfDNA and WBC gDNA samples from the same patient. c Oncoplot showing the 153 mutations detected in 67 out of 111 (60.36%) LC samples. Fourty-five LC samples without any mutation detected were not shown. Each column represents a sample and each row a different gene. The upper barplot represents the frequency of mutations for each sample, and the right barplot represents the frequency of mutations for each gene. Samples are ordered by the most mutated genes. d Allele fractions (x-axis, log scale) of mutations detected in plasma cfDNA of BLN patients (blue) and LC patients (red). e Oncoplot of the 28 mutations detected in 23 out of 78 (29.49%) BLN samples. 55 BLN samples without any mutation detected were not shown. f Predictive models to distinguish LC from BLN based on mutations detected. SUMAF (blue): the sum of AFs. Weighted_SUMAF (red): the weighted sum of AFs. The AUC of SUMAF model is 0.67 with 55.9% sensitivity and 76.9% specificity. The AUC of weighted_SUMAF model is 0.68 with 59.5% sensitivity and 71.8% specificity