The simulation aimed to reduce the variant density and reduce low VAF variants to better mimic a “real-world” clinical sample. While the high variant density and high percentage of variants with VAF below 5% in the test samples provided an advantage in estimating detection sensitivity with narrow confidence intervals, they presented obstacles here for the evaluation of TMB technical variability. The high variant density (equivalent to a TMB value about 1000) in the test samples will overshadow the effect of the FP calls and thus greatly underestimate TMB technical variability. The high percentage (over 75% in sample C) of variants with VAF below 5% may lead to an overestimate of TMB technical variability due to the variability in detecting these low VAF variants. The simulation was applied to the CTR. The resultant median VAF for low-frequency (< 20%) KPs would be between 5 and 10%. Sample C was chosen for the simulation analysis as there are fewer uncharacterized true variants in sample C with VAF greater than 2.5% in comparison to sample A. All the known variants in sample B were excluded first from our KP list and from all the variant calling results (VCF files of library replicates) of each panel. The KPs with VAF < 5% or VAF > 40% in sample A were also removed. In theory, no KP with VAF < 2.5% or VAF > 20% in sample C was left. To mimic multiple samples with the constant TMB at a similar level with a clinical sample, we kept a fixed portion of KPs that resulted in the same number of KPs across the CTR for panel testing in the simulated samples. Specifically, based on their VAF in sample A, k% (k = 0, 1, 2, 5, 10, 20) of KPs were retained for VAF between 10 to 40% and 0.5 × k% of KPs were kept for VAF between 5 and 10%. Given each value, (100 − k)% of KPs were randomly excluded from CTR to generate a base-report-region for all the pan-cancer panels. Therefore, the variants not excluded from a VCF file were counted as the reported calls by the library of a pan-cancer panel in a simulation run. It is worth pointing out that the reported variants include FP calls and possibly uncharacterized true variants in additional to the retained KPs. Finally, 5000 rounds of simulation were performed for each based-report-region with a k value greater than 0.

For each one of 5000 base-report-region with a k value (k > 0), the reported variants were then counted for these above a chosen VAF threshold for each library replicate. The variance and mean across all technical replicates for each panel were calculated. Then, we computed the overall average of variance and mean over the 5000 runs of simulation. Finally, the CV for TMB were calculated as:

The mean TMB for each chosen k value was calculated as the overall average count of variants divided by the report region sizes of six panels. This CV is adopted to measure the technical run-to-run variability and examine its dependence on TMB and VAF cutoff.

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.