Using mixed linear models and best linear unbiased predictions to predict seed yield in soybeans

Author

Donald McLean Panter

Date of Award

8-1991

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Plant, Soil and Environmental Sciences

Major Professor

Fred L. Allen

Committee Members

Bob McLean, Dennis West, Henry Fribourg, Vern Reich

Abstract

Best linear unbiased predictions (BLUP) from mixed linear models have been used to predict breeding values of dairy sires for milk yield based on information from their dams and daughters. One of the major advantages of BLUP is that predictions of individuals can be made when information on the individual per se is unavailable, but information from its relatives is available. Since BLUP methodology can utilize information from individuals per se and their relatives to predict the value of individuals, there is potential for its use in two important areas of plant breeding: i) predicting breeding values of parents from the performance of their relatives, and ii) ranking new genotypes when observed data for them are limited (e.g. early stages of performance testing). The objectives of this dissertation were to i) compare the efficiencies of BLUP and MPV in determining seed yield performances of future soybean (Glycine max (L.) Merr.) crosses from historical information about parents, ii) determine the effects of progeny and grand-progeny yield performance information on breeding values of their parents, iii) compare seed yield predictions from BLUP to a traditional approach of estimation, best linear unbiased estimations (BLUE), for ranking new genotypes from a limited number of yield tests, and iv) develop a computing strategy for utilizing BLUP in plant breeding applications. The F₄-F₆ bulks and F_5:6, lines from 24 crosses and four parents were evaluated in replicated yield trials in 11 environments to establish their relative seed yield performances. A summary of the results was; i) predictions of the 24 crosses from BLUP using only historical parental data were better indicators of performance than estimates from MPV, ii) BLUP breeding values of parents were more precise using small amounts of progeny information than when using large amounts of grand-progeny information, iii) BLUP was superior to BLUE for ranking new genotypes evaluated in a limited number of performance trials, and iv) computer software was developed using SAS/IML™ for computing BLUP values in plant breeding applications. The BLUP methodology should be considered a superior alternative to traditional approaches to genotypic performance estimation.

Recommended Citation

Panter, Donald McLean, "Using mixed linear models and best linear unbiased predictions to predict seed yield in soybeans. " PhD diss., University of Tennessee, 1991.
https://trace.tennessee.edu/utk_graddiss/7562