search

hassansaei / VNtyper

A tool to genotype MUC1 coding VNTR in ADTKD
Other
3 stars 4 forks source link

Explain Kestrel postprocessing heuristic better? #6

Open holtgrewe opened 3 weeks ago

holtgrewe commented 3 weeks ago

:wave: @hassansaei @berntpopp - the V2 update is greatly appreciated as it allowed me to run this successfully.

I wonder, however, whether the heuristics around here could be explained in a better fashion?

https://github.com/hassansaei/VNtyper/blob/37f88e94e38b4089d1b5b1a7a50891a10b926245/vntyper/scripts/kestrel_genotyping.py#L491

In particular, the black magic around pandas data frame wrangling could use some polishing. IMO, we are not dealing with relevant amounts of data and explicit modeling with Python dataclasses or so could make things more transparent. If you are interested, I could contribute some patches that make things more explicit without dataframes as I have been working on reproducing the method without data frames (which is now obsolete, luckily, because of your V2 updates).

berntpopp commented 3 weeks ago

Hey Manuel! Er are working on that. Let's write by mail for Carstens cohort. Best, Bernt