search

KHanghoj / NGSremix

GNU General Public License v2.0
6 stars 1 forks source link

Handling of missing data with lcWGS #7

Open K-ayers opened 1 year ago

K-ayers commented 1 year ago

Hi! I am using lcWGS and some of my sites have zero read depth and thus the GL is not computed during calling. I am unsure how to handle these sites when converting to Beagle format. How is NGSremix expecting missing data as input? Because I am doing pairwise comparisons of samples, do I need to filter out the union of these sites for each pair (and thus filter the reference input frequencies as well), or can the GLs be input as 'missing' in some way, eg 0,0,0 or something similar? Thanks for any help.

GenisGE commented 7 months ago

Hi, sorry for the late reply. The way to code missing data in beagle GL format is to give the same genotype likelihood to each genotype, i.e.e code it as 0.333333,0.333333,0.333333 (in the normalized version).

Genis