Fig. 4

Bioinformatics and limit of detection of somatic mutation calls from multi-gene OCEANS panels. a Summary of variant read frequencies (VRF) observed for a 7-plex OCEANS panel covering recurrent mutations in the KIT, IDH1, FLT3, NPM1, IDH2, and DNMT3A genes, which are present in high frequency in acute myeloid leukemia (AML). The input samples used here were internal reference samples constructed by spiking known quantities of synthetic DNA bearing a mutation of interest into NA18562 gDNA; see Adiitional file 1: Section S4 for summary of Nanopore Sequencing results on these experiments. Based on calibration experiments, we tentatively propose a 20% VRF cutoff for calling a variant. The limit of detection for different mutations varied between ≤0.05% VAF (for KIT, IDH1, FLT3, and NPM1) and 1% VAF (for DNMT3A). b Qualitative de novo variant calls using the Clair [21] bioinformatics pipeline, on the same data as in panel (a). Based on Oxford Nanopore internal calibration, Clair scores ≥180 are reliable. Although the Clair variant call differed from the VRF-based variant call for the NPM1 mutation, all other mutations were detected with similar VAF limits of detection. To be conservative and reduce false positive variant calls, we typically make a somatic variant call only when both VRF ≥20% and Clair ≥180. c Results of the AML 7-plex OCEANS panel on 50ng of a third-party reference DNA sample (Horizon HD829). All mutations in the reference sample covered by our panel were at 5% VAF, and were detected. d Summary of VRF observed for different internal reference samples using a 15-plex OCEANS panel covering 7 genes frequently mutated in melanoma (AKT1, AKT3, KRAS, MAP2K1, MAP2K2, NRAS, and PIK3CA). For this panel, the limits of detection varied between ≤0.05% VAF and 0.5% VAF. e Results of the melanoma 15-plex OCEANS panel on 50ng of a 99%:1% mixture of NA18562 gDNA and a third-party reference DNA sample (Horizon HD238). The HD238 sample is nominally a reference sample that is 50% VAF in BRAF-V600E (c. 1799T >A), so the input sample should be positive only for BRAF-V600E at 0.5% VAF. Upon our discovery of the KRAS-c.38G >A and PIK3CA-c.3140A >G mutations, we checked with the manufacturer and confirmed that these mutations are also present in the reference sample at low VAFs