diff options
| author | Pjotr <wrk@bosa.lan> | 2026-11-10 22:55:59 +0000 |
|---|---|---|
| committer | Pjotr <wrk@bosa.lan> | 2026-11-10 22:55:59 +0000 |
| commit | e2386cdf37196f7571afda4cfb9d1ec70b194cf7 (patch) | |
| tree | 88517f78c96452474fba3ad2cbbbbb15f9e5c24c /test/python/test_version.py | |
| parent | b3811de4f2fb0e1d65ade757ad3ea992eb3f9f67 (diff) | |
| download | genenetwork4-e2386cdf37196f7571afda4cfb9d1ec70b194cf7.tar.gz | |
qtl2: an option to keep heterozygous calls, to test what dropping them costs
R/qtl's convert2riself drops the third genotype code (see the note below),
and so does this converter -- but qtl2 then never sees the call, the HMM
re-imputes it from flanking markers, and on a marker that is 14% H that is
a different predictor from the one R/qtl fits.
{"group":"BXD","genotype_version":"BXD.8","het":"keep"} -> crosstype "f2"
On BXD trait 16627 (BXD.8, 35 phenotyped animals):
marker Mb R/qtl qtl2 drop qtl2 keep
rsm10000000889 5.652 9.59 3.79 9.98
rs33299937 5.653 9.59 3.73 9.68
rs13476332 6.311 9.59 3.70 ...
rsm10000000890 6.312 9.59 3.67 ...
So dropping the H calls is what costs qtl2 the peak: kept as a third
genotype it recovers R/qtl's number (9.68 against 9.59). The option is a
DIAGNOSTIC: an inbred line has no heterozygotes, so an f2-shaped cross is
wrong for inference even when it reproduces the other engine -- it is how
one finds out that the conversion, not the model, is the difference.
The policy is part of the conversion cache key (v3), and the reply says
when it is not the default. `het` applies only to a panel this server
converts: a shipped control-file cross carries its own coding and is
refused rather than silently ignored.
Diffstat (limited to 'test/python/test_version.py')
0 files changed, 0 insertions, 0 deletions
