Pre-Filtering genomes in VarFish

[alexejknaus]

I was wondering whether it would make sense to pre-filter WGS samples before loading them into VarFish.

In most cases Geneticists would filter out variants based on population frequency. Hence, after genotype calling one could add two steps: annotate all variants with GnomAD, 1KGP and other population frequency and filter out common variants (e.g. >2%)
Thereby a majority of common variants from a genome would be filtered out prior upload to VarFish, this would make the entire filtering faster.

[holtgrewe]

Yes, this can be implemented after varfish-annotator and before importing the data into VarFish.

varfish-annotator just creates a TSV file with the annotated frequency in one of the columns. You can use, e.g., awk to exclude any file with a frequency above your threshold. For WGS data, I would rather/further exclude everything off-target as interpretation of intergenic or deeply intronic variants is generally not feasible. You could also do this for exomes to get rid of spurious calls.

Also, I would be very interested seeing benchmark results for these differences.

Also, query plans would be interesting to see for differnt cases. Before doing any optimization, you should enable logging in postgres and execute the large statements that you see varfish issuing prefixed with EXPLAIN ANALYZE. You will see the query plan and the actual times. There is a couple of visualization tools for the plans:

• https://explain.dalibo.com/plan
• https://tatiyants.com/pev/#/plans
• https://explain.depesz.com/

HTH,
Manuel