GGE Biplot Analysis: A Graphical Tool for Breeders, Geneticists, and Agronomists

287 Pages·2002·9.316 MB·\287

by  Weikai Yan (Author), Manjit S. Kang (Author)

GGE
Biplot Analysis
A Graphical Tool for Breeders,
Geneticists, and Agronomists

GGE
Biplot Analysis
A Graphical Tool for Breeders,
Geneticists, and Agronomists
Weikai Yan • Manjit S. Kang
CRC PR ESS
Boca Raton London New York Washington, D.C.
Library of Congress Cataloging-in-Publication Data
Yan, Weikai.
GGE biplot analysis : a graphical tool for breeders, geneticists, and agronomists /
Weikai Yan and Manjit S. Kang.
p. cm.
Includes bibliographical references (p. ).
ISBN 0-8493-1338-4 (alk. paper)
1. Plant breeding--Statistical methods. 2. Crops--Genetics--Statistical methods. 3.
Genotype-environmental interaction. I. Kang, Manjit S. II. Title.
SB123 .Y364 2002
631.5′2′072--dc21 2002067091
This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with
permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts have been made to publish
reliable data and information, but the author and the publisher cannot assume responsibility for the validity of all materials
or for the consequences of their use.
Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic or mechanical,
including photocopying, microfilming, and recording, or by any information storage or retrieval system, without prior
permission in writing from the publisher.
The consent of CRC Press LLC does not extend to copying for general distribution, for promotion, for creating new works,
or for resale. Specific permission must be obtained in writing from CRC Press LLC for such copying.
Direct all inquiries to CRC Press LLC, 2000 N.W. Corporate Blvd., Boca Raton, Florida 33431.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation, without intent to infringe.
Visit the CRC Press Web site at www.crcpress.com
© 2003 by CRC Press LLC
No claim to original U.S. Government works
International Standard Book Number 0-8493-1338-4
Library of Congress Card Number 2002067091
Printed in the United States of America 1 2 3 4 5 6 7 8 9 0
Printed on acid-free paper
Foreword
"One picture is worth ten thousand words."
One of the most well-developed and acute faculties of the human brain is the ability to perceive,
analyze, and interpret complex visual information. Patterns, or visual relationships, are the easiest
to understand quickly. Comprehension of visual information is much better and faster than linear
or tabular numerical information.
Extensive databases of complex information are being meticulously accumulated and main-
tained in many research programs, but they are not being utilized to their full potential because
interpretation of them is too complicated. The principal component methodology and the GGE
biplot software described in this book can change that by making interpretation of complex data
sets as easy as a few keystrokes and a few minutes of viewing easy-to-understand graphic displays.
Using GGE biplot software requires minimal preparation of data — simple two-way data
spreadsheets are the basic input format. The initial graphic output can be easily modified from
drop-down menus. The data set being analyzed can be manipulated by removing various parameters
and entries during the real-time massaging of the data set and different statistical analyses can be
performed. Several alternative ways of viewing the output are possible to assist in interpretation
of the relationships among the parameters, among the entries, and their interactions, and it is all
done “live” in real time. Information can be analyzed faster, and can be understood more completely,
using the GGE biplot software visual display than the data tables and graphic displays generated
by most spreadsheets and database systems. The graphic output can be printed to produce a
hardcopy version, and/or saved as an image and used in presentations or publications. A log file
of the relevant statistics and correlation matrices is also produced and automatically saved for later
detailed reference.
The GGE biplot can help anyone learn more about the interactions among a multitude of
traits/parameters in their own data than ever before. GGE software has tremendous potential value
to plant breeders, agronomists, pathologists, physiologists, nutritionists, and anyone working in an
applied science field. It is currently being used in the cereal breeding program at the University
of Guelph to evaluate overall agronomic merit, quality, genotype X environment interaction for
numerous traits, genotype X trait interactions, and trait X environment interactions in breeding lines
being advanced through the testing system, to select parents and parental combinations for crossing,
to evaluate relationships among traits (especially quality), to identify determinants of yield and
quality factors in the populations, and to assess the discriminating value and stability of various
testing locations.
GGE biplots have been utilized in chromosome mapping of morphological and molecular
markers, and to identify and locate QTL markers in doubled haploid barley mapping populations.
The biplot concept is also used to determine the physiological associations of morphological traits
and kernel characteristics with quality factors. This is especially valuable when both negative and
positive correlations exist among several desired traits. Biplots can also be used to identify and
quantify the effects of environment on one or more traits across a range of genotypes, and the
effects the environment may be having on the relationships among those traits. This is particularly
useful when selecting among elite lines to advance to more extensive and intensive trials, and in
choosing parents for the next cycle of improvement in complex traits with interactions among the
components.
Biplots have been used in evaluating students' performance in a course to determine which
assignments and exercises are most likely to contribute to their understanding of concepts, based
on association with performance on later examinations.
Essentially, any two-way data set with multiple entries and multiple parameters can be analyzed
using GGE biplot; the result will be a better understanding of the interactions among entries, among
parameters, and interactions between entries and parameters. The GGE biplot software offers
convenience, the resolving power of principal component analysis, speed, and simplicity of inter-
pretation. The potential applications of the principal component analysis technique using this
amazingly friendly and informative software are limited only by the imagination of the researcher.
GGE biplot is simple and fun to use; it is much like a useful and productive video game for scientists.
Duane E. Falk, Ph.D.
Associate Professor and Cereal Breeder
Department of Plant Agriculture
University of Guelph
Guelph, Ontario, Canada
and
Adjunct Associate Professor
Department of Plant Biology
Faculty of Natural and Agricultural Sciences
University of Western Australia
Crawley, Western Australia
Preface
The human population of the world is currently growing at the rate of 1 billion people per 10 to
12 years. Projections are that, by 2050, world population will increase from the current 6 billion
to about 10 billion. During the past 50 years, agricultural research and technology transfer have
helped increase the output of world crops two and a half-fold. Ruttan (1998), while summarizing
the world’s future food situation, referred to the “2–4–6–8” scenario, which means a doubling of
population, a quadrupling of agricultural production, a sextupling of energy production, and an
octupling of the size of the global economy by 2050. Currently, more than a billion people can be
categorized as the world’s absolute poor, subsisting on less than $1 of income per day, and 800
million of these do not have secure access to food (McCalla, 2001). The challenge for agricultural
researchers to meet the food demand is astounding.
From the perspective of food security, the stability of agricultural production is as important
as, if not more important than, the magnitude of output (Wittwer, 1998). Food production is very
much a function of climate, which in itself is unpredictable; the principal characteristic of climate
is variability (Wittwer, 1998). The consultative group on international agricultural research
(CGIAR) warns, “Agricultural growth has to be achieved with methods that preserve the productivity
of natural resources, without further damage to the Earth’s previous life support systems — land,
water, flora, and fauna — that are already under stress” (Harsch, 2001).
Agricultural production may be increased through increased efficiency in utilization of resources
such as increased productivity per unit of land and of money, and through a better understanding
and utilization of genotype-by-environment interaction (GEI). Stuber et al. (1999) considered GEI
as one of the factors that gave impetus to research in the application of molecular-marker technology
and genomics to plant breeding. The GEI and stability of crop performance across environments
are expected to become more relevant issues in the 21st century as greater emphasis is placed on
sustainable agricultural systems.
GEI is a major concern among breeders, geneticists, production agronomists, and farmers
because of its universal presence and consequences. The occurrence of GEI necessitates multi-
environment trials (MET) and has resulted in the development and use of the numerous measures
of stability. Understanding and management of GEI has gone through several phases. The first
phase was the realization of GEI as a confounding factor in cultivar selection and plant breeding
early in the 20th century. The second phase was the concentrated study of GEI, which led to the
development of numerous measures of stability (reviewed in Lin and Binns, 1994; Kang and Gauch,
1996; Kang, 1998). The third phase was the integration of the genotype main effect (G) and GEI.
An integration of the two was needed because in practical breeding, selection only for stability is
inconsequential if production level is ignored. Concepts such as crossover interaction (Baker,
1988a), stability-modified yield (Kang, 1993), rank-based statistics (Hühn, 1996), statistics to
differentiate crossover from noncrossover interactions (Crossa and Cornelius, 1997), and methods
of identifying the “which-won-where” pattern of MET data (Gauch and Zobel, 1997) emerged; all
reflect the contemporary understanding and use of G and GEI in selection. The most recent
development along this line is the development of the genotype and the genotype-by-environment
(GGE) biplot methodology (Yan et al., 2000).
Since 2000, Yan has received tremendous support for his GGE biplot methodology from
colleagues worldwide. The GGE biplot methodology drew the attention of many plant breeders
and other researchers for two reasons. First, it explicitly and necessarily requires that genotype (G)
and (GE) interaction, i.e., GGE, be regarded as integral parts in cultivar evaluation and plant
breeding. Second, it presents GGE using the biplot technique (Gabriel, 1971) in a way that many
important questions, such as the “which-won-where” pattern, mean performance and stability of
genotypes, discriminating ability and representativeness of environments, etc., can be addressed
graphically.
Although GGE biplot analysis was initially used only for dissecting GGE and visualizing
MET data, its application has been extended to any data set that has a two-way structure. In the
area of plant breeding in particular, the GGE biplot has been used to address important questions
a breeder or researcher is likely to ask. Thus far, it has been applied to genotype-by-trait data,
genotype-by-marker data, quantitive trait loci (QTL)-mapping data, diallel-cross data, and host
genotype-by-pathogen strain data. Undoubtedly, with the fertile imagination of researchers
engaged in crop breeding and production, additional applications to other types of two-way data
will be found in time.
To facilitate the use of the GGE biplot methodology by researchers with only limited familiarity
with computer applications and statistics, a Windows application called GGEbiplot has been
developed (Yan, 2001). The GGEbiplot software has evolved into a comprehensive and convenient
tool in quantitative genetics and plant breeding.
This book was envisioned during a meeting between Manjit Kang and Weikai Yan at the
annual American Society of Agronomy meeting in Minneapolis, MN in November 2000, to make
this useful technology available on a wider scale to plant breeders, geneticists, college teachers,
and graduate students. The book is organized into three sections: Section I. GEI and stability
analysis (Chapters 1 and 2); Section II. GGE biplot and MET data analysis (Chapters 3 to 5);
and Section III. GGEbiplot software and applications in analyzing other types of two-way data
(Chapters 6 through 11).
In preparing the book, we have been cognizant of the needs of researchers, teachers, and students
of plant breeding, quantitative genetics, and genomics. We trust that readers will find the book
stimulating and useful, as we do. The book is extensively illustrated and a person with a few courses
in genetics and statistics should be able to comprehend easily the concepts and applications. It
should also be useful to all researchers in other areas who must deal with large two-way data sets
with complex patterns. We trust that this book will provide a powerful tool to breeders and
production agronomists, and make a significant contribution toward helping meet the challenges
of food production and food security that the world faces today.
Weikai Yan acknowledges that he benefited from his association and interactions with Professor
L. A. Hunt at the University of Guelph and Professor D. H. Wallace at Cornell University, and
from stimulating communications with Drs. Hugh Gauch and Rich Zobel (both at Cornell University
at the time) — two of the major advocates of the additive main effects and multiplicative interaction
effects (AMMI) model for analyzing MET data. Professor Paul L. Cornelius at the University of
Kentucky and Dr. Jose Crossa at CIMMYT provided valuable comments and suggestions during
the pre-publication phase of Yan et al. (2000). We thank John Sulzycki of CRC Press for his role
in making this project a reality.
Weikai Yan
Guelph, Ontario
Manjit S. Kang
Baton Rouge, Louisiana
Contents
SECTION I Genotype-by-Environment Interaction
and Stability Analysis...........................................1
Chapter 1 Genotype-by-Environment Interaction.........................................................................3
1.1 Heredity and Environment.......................................................................................................3
1.1.1 Qualitative Traits..........................................................................................................4
1.1.2 Quantitative Traits........................................................................................................4
1.2 Genotype-by-Environment Interaction.....................................................................................5
1.3 Implications of GEI in Crop Breeding....................................................................................7
1.3.1 The Breeding Phase......................................................................................................7
1.3.1.1 Environmental (E), Genetic (G), and GE Interaction Contributions...........7
1.3.1.2 Additional Breeding Programs.....................................................................8
1.3.1.3 Genetic Gain Reduction................................................................................8
1.3.1.4 Increased Field Evaluation Cost...................................................................8
1.3.1.5 Early- and Advanced-Performance Testing..................................................8
1.3.2 The Performance Evaluation Phase.............................................................................8
1.3.2.1 Problem in Identifying Superior Cultivars...................................................8
1.3.2.2 Increased Cost of Variety Testing.................................................................9
1.4 Causes of Genotype-by-Environment Interaction....................................................................9
Chapter 2 Stability Analyses in Plant Breeding and Performance Trials...................................11
2.1 Stability Concepts and Statistics............................................................................................11
2.1.1 Static vs. Dynamic Concept.......................................................................................11
2.1.2 Stability Statistics.......................................................................................................12
2.1.3 Simultaneous Selection for Yield and Stability.........................................................14
2.1.4 Contributions of Environmental Variables to Stability..............................................14
2.1.5 Stability Variance for Unbalanced Data.....................................................................15
2.2 Dealing with Genotype-by-Environment Interaction.............................................................16
2.2.1 Correct Genetic Cause(s) of GE Interaction..............................................................17
2.2.2 Characterize Genotypes and Environments...............................................................17
2.2.3 QTL-by-Environment Interaction...............................................................................17
2.2.4 Breeding for Stability/Reliability of Performance.....................................................18
2.2.5 Early Multi-Environment Testing...............................................................................18
2.2.6 Resource Allocation....................................................................................................19
2.3 GGE Biplot: Genotype + GE Interaction..............................................................................19