rossibarra / phasing_tests Goto Github PK

Right now assumes no missing data in mom. Should be able to resolve with adding more 1's to the probability matrices?

With recombination (and occasionally without), some windows will make multiple phases/haplotypes equally likely. This is suggestive of recombination.

Are there good heuristics for when there has been recombination in a window?

If there are lots of equally probable haplotypes can we resolve by storing all, then traversing the list of haplotypes at the end to find the path with minimum differences (a sort of cheesy HMM)?

Switching errors may happen in the regions where contain some recombinant kids.

A total of 10 progeny (shown 5 in the figure) were simulated to phase the mom. In the above figure, 1st panel, mom real hap1, hap2, genotype, phase1 and phase2; 2-6 panels, 5 kids. Yellow color indicates switching errors.

Currently only imputes selfed progeny. Will require
- all parents imputed (and phased?) first
- second pass to use imputed parents and small modification of current code to do imputation of progeny