search

Illumina / manta

Structural variant and indel caller for mapped sequencing data
GNU General Public License v3.0
408 stars 154 forks source link

Apply min-candidate-spanning-count per sample #17

Open ctsa opened 9 years ago

ctsa commented 9 years ago

via @hguturu

re: the other discussion on min-candidate-spanning-count Possibility of feature add of applying it at the sample level rather than at the cohort level? Since before the multiple input bams were one person it made sense to apply it across bams, but I am seeing high error due to a single bad read pair mapping that gets called since you get 1 or 2 spurious calls in each individual and it adds up to be sufficient to pass the filters

  • Test whether this should be option or default
  • In theory scoring function should filter the above described noise with low GQ scores, maybe the germline spurious chimera rate parameter isn't high enough? Maybe it needs to be estimated per sample?
hguturu commented 9 years ago

Additionally, I believe this gives you control over filtering PR, exposing a flag to filter over SR would be nice too.

Related, I have found sometimes with these spurious mappings is one side of the PR maps with high mapq, but the other side maps with a low one (not sure how the --min-mapq flag operates if it keeps the pair if one satisfies or requires both to satisfy). Similar comments regarding the AS score vs. MAPQ, but not sure if the algorithm accounts for --min-alignment-score internally.