Which variants are used in consensus assembly?

[tavareshugo]

viralrecon outputs the variants_long_table.csv file with a lot of variants that have the PASS filter. However, not all of these variants make it to the assembly.

• For ivar consensus, looking at the workflow's config we can see only variants with at least 75% frequency and minimum depth of 10 make it to the consensus.
• The same is true if using bcftools as the caller instead. In this case, bcftools mpileup is ran first, setting a minimum depth 10 (config here). Later, there is a bcftools filter step, which in the workflow config again suggests filters for variants with >75% frequency.
• For nanopore I don't know if there is a straightforward "threshold", as I assume that medaka consensus will use its own trained model to make the decision, so different types of variants may have different "thresholds"? I'm not sure...

[tavareshugo]

It may be possible to customise the workflow. For example, include the following in a file called nextflow.config in the project directory:

process {
  withName: 'IVAR_CONSENSUS' {
    ext.args = '-t 0.90 -q 20 -m 10 -n N'
  }
}

[tavareshugo]

See clarifications about output files in #23