Comparison of marker- and pedigree-based methods for estimating heritability in an agroforestry population of Vitellaria paradoxa C.F. Gaertn. (shea tree)
- First Online:
- Cite this article as:
- Bouvet, J.M., Kelly, B., Sanou, H. et al. Genet Resour Crop Evol (2008) 55: 1291. doi:10.1007/s10722-008-9328-8
- 101 Downloads
We implemented a regression-based method between pairwise relatedness estimated from markers and phenotypic similarity to estimate heritability of traits related to leaf size and morphology in a wild tree population (Vitellaria paradoxa C.F. Gaertn.: shea tree). We then compared the results with heritability estimated with a classical pedigree-based method. We tested both approaches in an agroforestry population of this tree species, a very important one and abundant in the Sudano-Sahelian zone of Africa. Twelve microsatellite loci were used to estimate pairwise relatedness after selection of estimator coefficients based on Monte Carlo simulation. The regression-based method applied to 200 individuals did not display a significant trend with physical distance between trees for relatedness as well as for actual variance of relatedness. In consequence, estimates of narrow-sense heritability of traits related to leaf size were not significant. The pedigree-based method using a progeny test with 39 families and 15 individuals per family from the same population showed high and significant estimates of narrow-sense heritability for the same traits (h2 = 0.36–0.95), demonstrating a marked genetic variation within the population. This discrepancy between methods stresses the poor performance of the molecular marker-based method. This can be explained by the absence of fine-scale structure within the agroforestry population of shea trees, other parameters being consistent with recommended values. The regression-based method does not seem well adapted to the agroforestry tree population. New experiments in tree populations and theoretical approaches are needed to evaluate the real potential of the marker-based methods.