Fig. 6

Validation of NSCLC and HCC OCEANS panels on clinical tumor tissue samples. a Summary of NSCLC panel results on 23 clinical samples (5 FF, 18 FFPE). Input DNA quantities ranged from 10 ng to 50 ng (Additional file 2). The X-axis shows the VRF based on a standard NGS analysis, and the Y-axis shows the OCEANS VRF. The horizontal line shows the 20% VRF cutoff for OCEANS variant calls, and the vertical line shows the 5% VRF cutoff for NGS variant calls. The numbers in quadrants display the number of loci in each group; 110 of the 136 putative variants in the top-left quadrant also had Clair scores of above 180 (purple dots). 2 of the 11 NGS confirmed variants in the top-right quadrant had Clair scores below 180 (yellow dots); both variants were insertions that typically have lower Clair scores. b Summary of HCC panel results on 21 clinical samples (5 FF, 16 FFPE). Input DNA quantities ranged from 10 ng to 50 ng (Additional file 2). 118 of the 167 putative variants in the top-left quadrant also had Clair scores of above 180 (purple dots). 3 of the 17 NGS confirmed variants in the top-right quadrant had Clair scores below 180 (yellow dots); all 3 variants were in TERT amplicon within a homopolymer region that resulted in lower Clair scores (Additional file 1: Section S5). c, & d Comparison of OCEANS calculated VAF and NGS VRF for NSCLC and HCC panel. Note: For NSCLC, 136 amino acid mutations (Additional file 2) from 136 clair called nucleotide changes in panel (a) (pink dots) are plotted here. For HCC, 135 amino acid mutations (Additional file 2)from 132 clair called nucleotide changes in panel (b) (pink dots) plus the 3 TERT mutations are plotted here. A insertion mutation in NSCLC and mutations in homopolymer regions in HCC are indicated red. VAF underestimation at these positions could be due to higher error rate. e, & f Precision-recall curve for data in panel (c) and (d), based on changing the OCEANS calculated VAF cutoff. g Panel (c) showed a significant number of loci wherein NGS results had 0% variant read frequency (VRF). Here, we show a histogram showing NGS read depth distribution for the 0% VRF loci, as well as the fraction of loci for each read depth group where OCEANS VRF was greater than 20%. Only loci with less than 300x NGS read depth showed discordance with OCEANS (mutation call based on 20% VRF)