Table 1.

Summary of recommendations

Issue	Recommendation
Reference material selection	A reference material (or a set of reference materials) with a high density of known variants spanning a range of low allele frequencies (e.g., from above 1 to 20%) is needed to assess the analytical performance of oncopanels (Fig. 1a). Ideally, the reference material will encompass a diversity of simple and complex variants in different cancer-associated genes.
Sensitivity	Sensitivity was found to be high (> 96.5%) for variants previously verified to have variant allele frequency (VAF) greater than 5% (Fig. 2a). We recommend that a less stringent VAF threshold is used to achieve high sensitivity for variants of VAF below 5%.
Spike-in	Utilizing a sample spiked-in at a specified amount (e.g., 5%) can provide additional variants at known allele frequencies for analytical validation of oncopanels.
Variant type	The sensitivity for detecting insertion and deletion variants (indels) is typically more variable and poorer than single nucleotide variants (SNVs), and this difference becomes more pronounced at low VAFs (Fig. 3). We recommend that analytical validation of indels is performed independently to SNVs.
Determining VAF threshold	Reference materials can be used to establish an optimal VAF threshold that reduces false positives (FPs) and retains sensitivity. Indels and SNVs may require different VAF thresholds to optimize performance (Fig. 3b).
Controlling the FP rate	In applications where a minimal FP rate is required, raising the allele frequency threshold was effective at reducing FPs. The additional restriction of analysis to the consensus targeted regions (CTR) can further reduce FP rate.
Genomic location	Genomic location can impact the rate of FPs detected, and we recommend that analytical validation of panels is independently performed inside and outside of the CTR (Fig. 4a).
Cross-lab reproducibility	Measuring the analytical performance of a panel in multiple labs is critical to establish reproducibility (Fig. 4c).
Estimating  tumor mutational burden (TMB)	TMB estimation should be confined to the CTR of each panel. Applying a minimal VAF threshold was helpful to reduce FPs and improve TMB evaluation (Fig. 5d).