family-based p value

[ygao940]

Dear authors,

We want to compute family-based p value using RV-NPL. The idea is to look if a particular subset of families drives the P value so that we can look at their clinical data to reduce heterogeneity.

I am not sure if RV-NPL has this option. Is there any way that I can do this on RV-NPL?

Best,
Yizhe

[gaow]

@percylinhai would it be possible to output that in the log file if not the main result output? If you dont have time to code it up you can point me to the correct lines to output , by responding to this ticket. I'll take care from there. I just want to make sure I output the correct quantity.

@ygao940 a dirty solution is that you write some pipeline to break up one large FAM file to multiple small ones and run rv-npl on each of them ...

[ygao940]

I see. Splinting input files into small families does work. Thank you!

[percylinhai]

@ygao940 Instead of splitting the original input files, I think it's better that you split the ped file with the CHP-coded markers.

@gaow We don't have p-value for each individual family. Alternatively, Z scores are available for each family (but they are not great indicators for p-value which is why we did not use analytic p-value and chose not to output this information). They are referenced below:

rvnpl/src/npl.py

Line 191 in 22053ca

z,o = fam_npl.stat() #get Z-score and observed IBD among affected individuals for each family

[gaow]

@ygao940 I see. What @percylinhai said makes sense -- when permutation testing is involved it does not help to save the original z-scores. Perhaps the "dirty" solution is still the most feasible.

[ygao940]

@gaow @percylinhai Ok. I will try the "dirty" solution. Thank you so much for the help.