Of the 212 plasma specimens evaluated, 123 specimens had one or more mutations identified by one or more of the technology platforms, with no mutations identified in the remaining 89 specimens. The mutations identified included 116 EGFR, 25 KRAS, 14 PIK3CA, 10 BRAF and 50 “other gene” mutations. The input amount of cfDNA was not standardized across assays and was generally higher for the custom NGS-based assay. When mutations were limited to those detected by both OncoBEAM and SeqSensei platforms, the sensitivity and specificity of the OncoBEAM assay for detection of mutations found using SeqSensei assay were 84% and 94%, respectively. Concordance between OncoBEAM and SeqSensei assays was observed in 75% of cases, with mutations in 7% of cases missed due to the sensitivity limit of the SeqSensei assay, 5% of cases missed due to the sensitivity limit of the OncoBEAM assay, and 13% of cases because gene mutations were not detected with theOncoBEAM assay. Mutations with discordant finding often had a MAF <0.15% or were in regions with low coverage in the OncoBEAM assay. The MAF was strongly correlated between SeqSensei and OncoBEAM assays (R2=0.74, P<0.001)
When mutations were limited to those detected by both the custom NGS and SeqSensei assays, the sensitivity and specificity of the custom NGS assay for detection of mutations found using SeqSensei were 83% and 76%, respectively. Concordance between the custom NGS assay and SeqSensei was observed in 68% of cases, with discordance attributable to the limit of detection of the custom NGS assay (11% of cases), the limit of detection of the SeqSensei assay (5% of cases), and because mutations were only detected by the custom NGS assay (16% of cases). Much of the discordance occurred for mutations with a MAF<1.0%. The MAF was modestly correlated between SeqSensei and the custom NGS assay (R2=0.46, P<0.001). It is important to note that the custom NGS assay detected additional mutations in TP53, ERBB2 and MET genes not included in SeqSensei which the authors state have prognostic and/or predictive value.