|Cervantes-Martinez, C - UNIVERSITY OF MEXICO|
|Nyquist, Wyman - PURDUE UNIVERSITY|
Submitted to: Plant Breeding Reviews
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: October 1, 2002
Publication Date: January 10, 2003
Citation: Holland, J.B., Cervantes-Martinez, C.T. 2003. Estimating and interpreting heritability for plant breeding.. Plant Breeding Reviews. Vol. 22, p. 9-112. Interpretive Summary: Observed variation among plants can be due to genetic differences or to environmental causes. Heritability measures the proportion of observed variation among plants that is due to genetic differences. The main purpose of estimating heritability is to compare the expected gain from selection from alternative selection strategies. Thus, heritability is the fundamental concept of plant breeding. Newer statistical methods have recently been developed that can be applied to the estimation of heritability. This review article discusses these newer methods and how they can be applied to heritability estimation and provides examples of computing code that can be used to obtain heritability estimates and their standard errors with a commonly used statistical software package. In order to place the estimation procedures in context, the interpretation of heritability estimators obtained from different mating schemes and generations is also discussed.
Technical Abstract: Heritability was originally defined by Lush as the proportion of phenotypic variance among individuals in a population that is due to heritable genetic effects. This definition is now termed "heritability in the narrow sense" and is designated h2. Variations on this idea are often also referred to as heritability, and they include "heritability in the broad-sense," the proportion of phenotypic variance that is due to all genetic effects, and heritability of family means, the proportion of the phenotypic variance of family means that is due to family genetic effects. Whereas Lush's definition was based on his experience as an animal breeder, in which the basic unit of observation and selection is nearly always the individual animal, plant breeders deal with a great diversity of observational units and mating systems. This complicates both the procedures for estimating heritability and the meaning of heritability itself. The main purpose of estimating heritability is to compare the expected gain from selection from alternative selection strategies. Heritability estimates are useful for comparing the gain from selection under different experimental designs, and this information, combined with information about the relative costs of additional replications within each macro-environment, additional years of evaluations, and additional locations for evaluations, can be used to design optimal breeding strategies. Similarly, heritabilities based on different family structures derived from the same base population can be compared to determine which family structure is best for maximizing genetic gain over units of time. The focus of this review will be on placing mixed models analysis procedures in the context of typical plant breeding experiments and to provide examples of computing code that can be used to obtain heritability estimates and their standard errors with the commonly used SAS system. In order to place the estimation procedures in context, we first discuss the interpretation of heritability estimators obtained from different mating schemes and generations. Finally, deciding whether effects in the statistical model are fixed or random has important implications for inferring results and estimation procedures, therefore we address this issue in terms of estimating heritability as well.