Thread by @ShainLab

Wanted to put out a brief thread on calculating tumor mutation burden because few papers explain how they do this, and I suspect most are not calculating it accurately. Mutation burden is typically expressed as mutations per megabase.

To calculate the numerator, first and foremost, you need a high quality set of mutation calls. Most studies meet this [bare minimum] standard, however, they fail to consider…

You must also define a cutoff for clonality. If you take the same sample and sequence it to 10X, 100X, 1000X, and 10000X coverage, you will find more mutations with more coverage b/c you pick up more subclonal mutations.

I typically count mutations that are predicted to be in greater than 50% of tumor cells. This is because we are ordinarily powered to detect mutations at this threshold in most samples we sequence. The biological questions being asked might factor into this decision, too.

To calculate the denominator, one must know how much sequencing footprint has sufficient coverage to call a mutation. Most labs just divide mutation count by the footprint of their baits – but can you really detect a mutation at 100% of sites?

In our lab, we ask: What threshold of coverage is needed to detect a mutation? Answer: depends on tumor cellularity and the variant caller. After determining an appropriate threshold, we use the Footprints software to count the number of basepairs that meet this requirement.

By the way, here is a video of how to run Footprints: . Described here: biorxiv.org/content/10.110…

To emphasize the importance of the denominator calculation, we recently downloaded tumor exomes from various studies. Our ability to detect mutations varied dramatically depending on overall coverage, variability in coverage, and tumor cellularity.

Overall, the considerations described above would permit more standardized and accurate measurements of tumor mutation burden (TMB) – a clinically actionable variable if measured correctly.