Estimation of kinship between F2 and captive individuals
From the SNP set filtered by the criteria determined above, COLONY 2.0 (Jones and Wang, 2010), Sequoia (Huisman, 2017), and gRandma (Delomas and Campbell, 2021) were used to estimate relationships between F2 and captive individuals. COLONY 2.0 was used to estimate FS and HS relationships between F0 individuals and between F2 individuals. Both COLONY 2.0 and gRandma were used to estimate GG relationships between F0 and F2 individuals. Sequoia was used to estimate UANN relationships between F1 and F2individuals. Here, HS, GG, and UANN have the same relatedness (Thompson, 1975). Although these three relationships are inherently genetically indistinguishable without utilizing linkage information, it is important to note that in this study, F0 consistently represents the grandparent generation, F1 corresponds to the uncle/aunt generation (parent generation), and F2represents the grandchild generation (Fig. 1). Therefore, if a half-sibling (HS) relationship is estimated between F0and F2 individuals using COLONY 2.0, it actually indicates a GG relationship (J. Wang, personal communication). Specifically, all 62 individuals from the F0 and F2 generations were included as candidate offspring in COLONY 2.0, and if an HS relationship was estimated between F0 and F2 individuals, it was considered a GG relationship. Likewise, all 55 individuals from the F1 and F2 generations were entered as candidate offspring in COLONY 2.0, and UANN relationships were considered as HS relationships.
For the COLONY 2.0 settings, both the allele dropout rate and error rate were set to 0.01 based on the results of the preliminary considerations alongside other filtering conditions (Table S6), and inbreeding was considered. Other settings were the same as those used in the consideration of filtering parameters (random mating was assumed, estimation accuracy was very high, and run length was set to 3). From a prior study, those relationships with a support probability of 0.8 or greater in COLONY 2.0 were considered to be true kinship (Thow et al., 2022).
gRandma can provide estimates for trio relationships between paternal and maternal grandparents and one grandchild. The estimated GG trios are then output with the Mendelian incompatibility (MI: the closer to 0, the less genetic inconsistency) and the log likelihood ratio (LLR) of assuming that the trio is unrelated. It is also possible to calculate the false positive and false negative rates for each MI when the LLR threshold is set to a certain value. In this study, the false positive and false negative rates for each MI were calculated for LLR = 30, 50, 60, 70, 80, and 90 as candidates for the threshold, and LLR = 70 was adopted because at this value the false positive rate was 0.06 and the false negative rate was 0 when MI = 0 (Table S2). The results of the kinship estimation are presented as “robust” if LLR ≥ 70 and MI = 0, and “weak” if LLR < 70 and MI = 0 or LLR ≥ 70 and MI > 0. A total of 62 individuals (F0–F2) were used, the same as in the estimation of GG using COLONY 2.0, and the filtering conditions were also the same (MIN_DP = 5, MAX_DP = 50, MIN_MEAN_DP = 15, MIN_GQ ≥ 30, CR > 0.9, MAF ≥ 0.03, HWE < 0.00001, HET ≤ 0.7, LD ≤ 0.2). The allele dropout rate and error rate were both set to 0.01, the same as in COLONY 2.0.