Missing variants with high coverage WGS samples. How to capture the missed variants?

[ChristopheLegendre]

@ctsa

Hi Chris,

background

We are performing somatic calls [constitutional versus tumor samples] with Manta on Whole Genome Sequencing (WGS). The average difference of coverage between constitutional and tumor samples is of factor 4 on average (let's say ~25X for constitutional and ~110X-200X for tumors) Some valid somatic variants present in the tumor samples are not called by Manta using the default configuration file; These variants definitely does exist in the tumor and not in the constitutional sample.

tests performed

We performed some tests by modifying some variables in the config file, variables that seem most relevant to capturing these missed variants. None of the modifications but one allowed us to capture the variants: graphNodeMaxEdgeCount. We tried to set the value to 1, 5, 10, 20, 30, 50, 100, 200. None of these values worked. The only value that worked and allowed us to capture some of the variants was 0 (Zero), which means we ran our WGS as if it were RNA or Exome as mentioned in the config file. (solution was also suggested by you in #8, but it seems the case was for tumor only call)

issue with graphNodeMaxEdgeCount = 0 when dealing with high coverage WGS samples

Unfortunately setting that value to zero increases the runtime for a very very longtime, much longer than the average 2-5 hours when using default ini configuration.

Based on that, we assumed that the issue could come from the coverage difference between the constitutional sample and the tumor sample.

code digging

By digging into the code, we found that variable called maxDepthFactor which is hard-coded to 3.0f and which is related to somatic calls specifically it seems. https://github.com/Illumina/manta/blob/27ba65e788572abb2b3d8afcb6aea1f5e8e98ae6/src/c%2B%2B/lib/options/CallOptionsSomatic.hpp#L38

Questions:

• what settings recommendation would you advise to use in the case of WGS somatic calls with ~25X Normal and ~110X tumor?
• Can we modify the variable maxDepthFactor at runtime ; if yes, How? adding it to config file? what value will you recommend?
• Any other variables we could modify at runtime to allow capturing the missed somatic variants? such as variable in hat function called CallOptionsSomatic there: https://github.com/Illumina/manta/blob/27ba65e788572abb2b3d8afcb6aea1f5e8e98ae6/src/c%2B%2B/lib/options/CallOptionsSomatic.hpp#L24-L49

Thanks

Thanks for your help and recommendations Best Regards,

Christophe