Genotyping within trios

[vinzenzmay]

Hello, I have been able to produce very promising CNV calls with ExomeDepth, but now I would like to use the results to check the consistency in given trios (parents and child). Some CNV calls might show small differences due to differences in the hidden states and consequently lack a proper comparability. Q: (How) can I use ExomeDepth to give it a list of possible CNVs (from previous CNV calls) and then receive this list aligned with the probabilities (or any other score) of each CNV to be present in a given sample? That way, I could make the CNV calls of two different but related samples comparable and thereby check mendelian consistency within a trio.

Thank you very much and best regards, Vinzenz

[vplagnol]

This goes way back, but I did have a routine for this a long time ago. SOrry my memory is fuzzy but this piece of code I think does something like what you want. It computes the probabilities across all the provided exons. Object is an ExomeDepth object here. That may be enough of an inspiration to get it done?

load('debug.RData')


print(.Object@phi)
print(.Object@expected[1:100])
print(as.integer(.Object@reference + .Object@test)[1:100])
print(as.integer(.Object@test)[1:100])

.Object@likelihood <- .Call("get_loglike_matrix",
                              phi = .Object@phi,
                              expected = .Object@expected[1:10],
                              total = as.integer(.Object@reference + .Object@test)[1:10],
                              observed = as.integer(.Object@test)[1:10])