The interaction between sire and f was a significant term when fitted in the MANOVA of the nine morphometric traits (Fthirty-six,2208=1.451, P=0.041) but f fitted as a main effect was not (F9,549=0.903, P=0.523). MLH was not a significant term either as a main effect (F9,549=1.5, P=0.144) or as an interaction with sire (Fthirty six,2208=0.715, P=0.896). Note that f and MLH were not fitted in the same model for either the univariate or the multivariate analyses.
Forecasts for other vertebrate communities
And the Coopworth sheep inhabitants, bottom line analytics in accordance with f and marker heterozygosity were collected to possess eleven other communities. These studies was basically next always guess the new relationship coefficient between f and you may MLH (a) with the markers which were typed in the study population to date, and you may (b) in the event the 100 markers out-of indicate heterozygosity 0.7 were published. Rates is actually exhibited in Desk 1. The people in which MLH try the best predictor out of f is actually Scandinavian wolves with an asked r(H, f)=?0.71 in case your 29 recorded microsatellites had been authored and you may an expected r(H, f)= ?0.ninety if the 100 loci was indeed blogged. The people where MLH try terrible in the anticipating f try the fresh new collared flycatchers (Ficedula albicollis) toward Swedish Island regarding Gotland, that have an expected r(H, f)=?0.08 when your around three noted microsatellites had been authored and a supposed r(H, f)=?0.32 if the 100 loci were published. Generally, heterozygosity would not render powerful estimates out of f, in the event one hundred loci was typed. Eg, the latest expected roentgen(H, f) is actually weakened than simply –0.5 for five of your own twelve populations and weaker than just ?0.eight to have 9 of communities.
In seven of the populations, r(H, f) had actually been estimated, enabling a comparison between expected and seen correlation coefficients (Table 1). In Scandinavian wolves and Large Ground Finches, the observed and expected correlation coefficients were almost identical. In four of the five other populations, r(H, f)observed was weaker than r(H, f)expected, perhaps due to errors in estimation of f (see Conversation).
Discussion
The primary objective of this study was to establish if and when MLH can be used as a robust surrogate for individual f. A theoretical model and empirical data both suggest that the correlation between MLH and f is weak unless the study population exhibits unusually high variance in f. The Coopworth sheep data set used in this study comprised a considerably larger number of genotypes (590 individuals typed at 138 loci) than any similar study, yet MLH was only weakly correlated to individual f. Furthermore, f explained significant variation in a number of morphometric traits (typically 1–2% of the overall trait variance), but heterozygosity did not. From equation (5), it can be seen that the expected correlation between trait value and MLH is the product of the correlation coefficient between f and the trait (hereafter r(W, f)) and r(H, f). Estimates of the proportion of phenotypic trait variation explained by f are scarce, although from the limited available data 2% seems a typical value (see for example Kruuk et al, 2002; this paper, Table 2). Assuming r(W, f) 2 =0.02, and given the median value of r(H, f)=?0.21 reported in Table 1, a crude estimate of average r(W, H) is 0.03, which is equivalent to MLH explaining <0.1% of trait variance. These findings are consistent with a recent meta-analysis that reported a mean r(W, H) of 0.09 for life history traits and 0.01 for morphometric traits (Coltman and Slate, 2003). In summary, MLH is a poor replacement for f, such that very large sample sizes are required to detect variance in inbreeding in most populations.