Analysis of relationships by gRandma
By utilizing kinship estimates from COLONY 2.0 or Sequoia, including GG,
UANN, and FS relationships, we were able to identify families for 36 out
of 47 F2 individuals. However, the families for the
remaining 11 F2 individuals were still unknown. In
contrast, gRandma, a specialized software for estimating GG trio
relationships, was able to estimate grandparents for almost all
F2 individuals. gRandma failed to infer any GG
relationship for only two of the F2 individuals, but
Sequoia estimated that they had a UANN relationship with an
F1 individual (mA1-4) of the A1 family, which supports
the hypothesis that these F2 individuals were produced
by breeding between F1 individuals of the A1 family.
The algorithm used by gRandma is unique because it considers GG trios
instead of just GG pairs. Originally developed to estimate GG
relationships between wild-caught grandchildren and hatchery
grandparents, gRandma was designed to detect crossbreeding when adult
sampling in the wild is limited (Delomas and Campbell, 2021). As a
result, it was initially intended for cases where only one set of
grandparents, paternal or maternal, was sampled. However, in this study,
where both paternal and maternal grandparents were sampled, gRandma was
able to estimate both sets of grandparents (T. A. Delomas, pers. comm.).
However, because Delomas and
Campbell (2021) initially assumed a low probability of false positives
from trios other than GG, it is possible for a false-positive trio to
yield a high LLR when, for example, the potential grandparents include
siblings (Delomas, pers. comm.). As the COLONY 2.0 analysis revealed
that the target population in this study contained many siblings in the
grandparents’ generation, we incorporated a Mendelian incompatibilities
(MI) threshold in addition to the LLR value. This is because a trio with
a high LLR and low MI is more likely to be reliable, as a true GG trio
would have both a high LLR and low MI, whereas a false-positive trio
would have a high LLR and high MI. Moreover, we have information on the
relationships between grandparents, such as if they are FS and HS, and
which individuals were possibly interbred, so we were able to identify
whether the inferred trios were false positives.