Calling with cancer module and -t increases runtime multifold about octopus HOT 3 CLOSED

It seems related to the data, it slows down at the chr13 range when running without the -t parameter.

from octopus.

Hi, thanks for your question. This is likely because this region is 'complex' in some way; either there is substantial true or false variation (e.g. due to miss-mapped reads or sequencing errors). It's also possible that the depth is very high in this region. The runtime of Octopus is mostly influenced by the number of candidate haplotypes that are considered, and therefore the number of proximate candidate variants. This particularly affects the cancer calling model as this is considerably more complex than other models, even more-so without a normal sample as there is generally greater uncertainty in sample genotypes.

Unless you have a strong reason to believe that these regions do not contain true variation, in which case you could skip them (e.g. using the --skip-regions or --skip-regions-file option), then I wouldn't take any further action. If the runtime is prohibitively long then you could consider ways to reduce runtime (e.g. reducing --max-haplotypes, --max-somatic-haplotypes, or --max-genotypes), although this will likely come at an accuracy cost.

from octopus.

Thanks, @dancooke. We actually looking at variants at these locations (BRCA1 and 2, mostly) and from a visual inspection there's seem to be quite a lot of insertions in this area, that might be increasing complexity. This is IonTorrent data which is know to have lower quality than Illumina, which we are testing soon.

Thanks again.

from octopus.