Source: UNIVERSITY OF FLORIDA submitted to
ACCELERATED BREEDING BY IMPROVED ACCURACY AND MATE ALLOCATION USING GENOME-WIDE SELECTION
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
TERMINATED
Funding Source
AFRI COMPETITIVE GRANT
Reporting Frequency
Annual
Accession No.
1000611
Grant No.
2013-67013-21159
Project No.
FLAW-2013-01849
Proposal No.
2013-01849
Multistate No.
(N/A)
Program Code
A1141
Project Start Date
Sep 1, 2013
Project End Date
Aug 31, 2018
Grant Year
2013
Project Director
Kirst, M.
Recipient Organization
UNIVERSITY OF FLORIDA
G022 MCCARTY HALL
GAINESVILLE,FL 32611
Performing Department
AG-SCHL-FOREST RES / CONSERV
Non Technical Summary
Breeding of forest trees is a long, complex and costly process. Just asingle breeding cycle can take over a generation, and requires large field trials. However, breeding can be hyper-accelerated through selection based on favorable genetic markers,also known asgenome-wide selection. In genome-wide selection, genetic markers and trait data collected from a breeding population are used to develop a prediction model that is used to estimate the genetic value of individuals in future generations, based on their genetic marker composition. These models also allow breeders to identify mates to cross and generate families with optimal gene combinations. However, to utilize genome-wide selection for mate allocation, predictive models that account for additive and non-additive effects need to be developed. In this project we will create and disseminate improved methods of genome-wide selection for early identification of elite individuals and the most valuable crosses to be made from a breeding population. Our goals are to (i) increase and accelerate genetic gain by developing and applying improved genome-wide selection prediction models that incorporate non-additive effect; (ii) predict the performance of crosses to generate families with superior properties in future generations; and (iii) develop tools and resources to facilitate the use of these advanced methods by breeders. Methods developed in this proposal will be applied to forestry breeding populations, for productivity and disease resistance traits, and can be readily extended to agricultural crops.
Animal Health Component
0%
Research Effort Categories
Basic
(N/A)
Applied
50%
Developmental
50%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
2010699108150%
2010699108050%
Knowledge Area
201 - Plant Genome, Genetics, and Genetic Mechanisms;

Subject Of Investigation
0699 - Trees, forests, and forest products, general;

Field Of Science
1081 - Breeding; 1080 - Genetics;
Goals / Objectives
(1) Improve early selection of elite genotypes by developing more accurate predictive models that incorporate non-additive effects. Non-additive effects will be incorporated in methods that differ with respect to assumptions regarding distribution of marker effects and partition of variance, for productivity, development and disease resistance traits. Accuracy and bias will be compared among methods, and relative to traditional additive genome-wide selection models. (2) Predict yet-to-be-made crosses that create optimal allelic combinations for high productivity and resistance to biotic stresses. Advanced marker-based predictive models (including additive and non-additive effects) will be applied to identify mate-pairs that create superior families and cultivars for traits of commercial value (productivity and disease). Performance of these predictions will be tested by evaluating families derived from those crosses. (3) Facilitate and accelerate the use of advanced genome-wide selection tools by the community of breeders. Computational tools will be developed that will be made readily and easily accessible to traditional and molecular breeders, in agriculture and forestry. Training workshops and extension tools will be created to ensure the success of their application.
Project Methods
AIMI. Predictive models that incorporate non-additive effects will be evaluated in the context of three different approaches that differ with respect to assumptions regarding distribution of marker effects, partition of variance components and normality. In the experiments to be carried out in the first aim, we will test the hypothesis that including non-additive variation will improve accuracies relative to standard (additive) models developed previously. Methods to be tested will include linear parametric Genomic BLUP where all the markers effects are assumed to contribute with the same variance (Experiment I-1), the Bayesian Bayes Cπ where only a reduced number of markers is expected to contribute to the variance (Experiment I-2) and the semiparametric Reproducing Kernel Hilbert Space (Experiment I-3). Predictive ability, accuracy and bias will be compared among methods, and relative to traditional additive genome-wide selection models (Experiment I-4). The various methods will be evaluated across 17 growth, development and disease resistance traits previously phenotyped in a breeding population. These traits include stem diameter, stem height, total height to the base of the live crown; crown width properties; branch angle and diameter; tree stiffness; lignin, cellulose and hemicellulose content; latewood percentage; wood specific gravity; and several fusiform rust and pitch canker disease resistance parameters. These traits have distinct heritabilities and genetic architectures. Thus, they are particularly suited to evaluate different methods of developing genome-wide selection prediction models. Alternative methods are expected to perform differently depending on their specific assumptions regarding the contribution of markers to trait variation, as we recently observed. AIMII. Prediction models that incorporate non-additive effects are essential for optimal allocation of mate-pairs, because they allow the prediction of specific combining abilities. In Experiment II-1, the genetic value of all potential families that can be generated among individuals in a breeding population, and their progeny, will be estimated in silico for productivity (stem diameter and height, wood specific gravity) and disease resistance traits (fusiform rust and pitch canker), based on models selected in Aim I. Then, the method will be validated by evaluating the performance of selected families derived from a breeding population (Experiment II-2). AIMIII. Based on the information generated in the first two aims of this project we will create a free analysis package that implements the methods of developing predictive models that incorporate non-additive effect, and the use of that information for allocating optimal mate-pairs (Experiment III-1). Breeders will be trained on the use of standard and advanced genome-wide selection tools to address their needs, through workshops and newly developed online courses to be made publically available (Experiment III-2).

Progress 09/01/13 to 08/31/18

Outputs
Target Audience:1. Plant breeders (traditional and molecular): The project targetted primarily plant breeders that use traditional and advanced methods of genetic improvement, particularly strategies such as genomic selection. This audience is the primary users of the information and methods developed in this project and main beneficiaries of the advances made possible by it (methods of prediction and incorporate non-additive effects and methods for mate-pair allocation). 2. Research scientists (quantitative genetics, genomics, biotechnology): A product of this project was the development of methods to identify the presence of non-additive effects that contribute to complex traits. Thus, the advances contribute to a better understanding of the genetic factors that control traits of agricultural interest. 3. Graduate students: Graduate students were trained in the interface between quantitative genetics and genomics, a field of study in high-demand in the agricultural and forestry sectors. Students trained in this program have proceeded to careers in academia and industry, related to plant breeding and selection. 4. Farmers/Growers: These are the beneficiaries of the methods developed here, as they capture the gains generated by the development of new and improved germplasm. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?During the years of the project, the PIs organized the workshop "Phenotype Prediction using Genomic Data", which yearlyon-site and on-line. Attendance reached 500-600 participants every year. A full overview of the workshops delivered every year since they were first implemented as part of this project can be found at: http://ufgi.ufl.edu/seminars-events/phenotypic-prediction-workshop-2020/ Similarly, the project has provided Ph.D. training opportunities to at least three graduate students in the first of quantitative genetics and plant breeding. How have the results been disseminated to communities of interest?See "Opportunitites for Training and Professional Development". What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? The following was accomplished under the goals of the project: (1) We developed and published methods that incorporate non-additive effects in genomic prediction models. For a detailed outcome of the implementation of these methods, see the publication by de Almeida et al. (Heredity 2016 117:33-41) (2) We created methods to predict the outcome of yet-to-be-made crosses, and applied it toidentify superior families and cultivars for traits of commercial value in forestry. These advances were published as part of a Ph.D. dissertation by Dr. Marcio Resende. (3) As part of this and other projects, we developed tools for genome-wide selection that incorporated the methods developed in the first two aims.

Publications

  • Type: Journal Articles Status: Published Year Published: 2016 Citation: Vazquez AI, Veturi Y, Behring M, Shrestha S, Kirst M, Resende MF Jr, de Los Campos G. Increased Proportion of Variance Explained and Prediction Accuracy of Survival of Breast Cancer Patients with Use of Whole-Genome Multiomic Profiles. Genetics. 2016 Jul;203(3):1425-38. doi: 10.1534/genetics.115.185181. Epub 2016 Apr 29. PMID: 27129736; PMCID: PMC4937492.
  • Type: Journal Articles Status: Published Year Published: 2016 Citation: de Almeida Filho JE, Guimar�es JF, E Silva FF, de Resende MD, Mu�oz P, Kirst M, Resende MF Jr. The contribution of dominance to phenotype prediction in a pine breeding and simulated population. Heredity (Edinb). 2016 Jul;117(1):33-41. doi: 10.1038/hdy.2016.23. Epub 2016 Apr 27. PMID: 27118156; PMCID: PMC4901355.
  • Type: Journal Articles Status: Published Year Published: 2014 Citation: Mu�oz PR, Resende MF Jr, Gezan SA, Resende MD, de Los Campos G, Kirst M, Huber D, Peter GF. Unraveling additive from nonadditive effects using genomic relationship matrices. Genetics. 2014 Dec;198(4):1759-68. doi: 10.1534/genetics.114.171322. Epub 2014 Oct 15. PMID: 25324160; PMCID: PMC4256785.


Progress 09/01/13 to 08/31/14

Outputs
Target Audience: The target audience reached by the project durign the reported period included: 1. 577 registered plant breeders, research scientists and students participated in the workshop “Phenotype Prediction using Genomic Data” (http://ufgi.ufl.edu/seminars-events/genomic-workshop/), supported by this project. 2. Fifteen graduate students that attended journal clubs on genomic selection Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided? The project provided a workshop in genomic selection that was widely atteneded by scientists, students and other professionals. The project also provided revised curriculum in advanced plant breeding, to graduate students at the University of Florida. The PIs Kirst and Munoz are currently working to provide this graduate course through disntance learning. How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? Regarding the goals described below, the accomplishments were: (1) Improve early selection of elite genotypes by developing more accurate predictive models that incorporate non-additive effects. We completed the development of prediction models that incorporate non- additive effects, as described in the project. In order to fully evaluate their performance, we have added a simulation component (not described in the original project) that will allow us to fully assess the accuracy and bias of these models underdifferent genetic architectures. (2) Predict yet-to-be-made crosses that create optimal allelic combinations for high productivity and resistance to biotic stresses. The identification of crosses to be made, that optimize allelic combinations, is currently in progress. (3) Facilitate and accelerate the use of advanced genome-wide selection tools by the community of breeders. The first training workshop occured in August 11, and included the participation of ~80 local and over 500 participants online. A workshop that follows the same model will be organized every year of the project. An R-script aimed at facilitating the implementation of non-additive models in genomic selection is currently under development, as are the extension tools.

Publications

  • Type: Theses/Dissertations Status: Published Year Published: 2014 Citation: Resende, M.F.R. (2014) GENOMIC SELECTION IN PLANT BREEDING: PREDICTING BREEDING VALUES, NON-ADDITIVE EFFECTS AND APPLICATION TO MATE-PAIR ALLOCATION (Doctoral dissertation).
  • Type: Conference Papers and Presentations Status: Published Year Published: 2013 Citation: Resende Jr., M.F.R., S.A. Gezan, M.D.V. Resende, G. de los Campos, M. Kirst, D. Huber, G.F. Peter and P. Mu�oz. 2013. Poster presentation: Unraveling additive from non-additive effects using genomic relationship matrices. Impact of Large-scale Genomic Data on Statistical and Quantitative Genetics Conference, November 24-26, Seattle, USA.
  • Type: Conference Papers and Presentations Status: Other Year Published: 2013 Citation: Kirst, M. 2013. Empowering breeders through genomic selection and next-generation sequencing. Animal Sciences Department, University of Florida, November 6, Gainesville, FL.
  • Type: Conference Papers and Presentations Status: Other Year Published: 2013 Citation: Kirst, M. 2013. Accelerated breeding of pines using advanced methods and applications of genomic prediction. Scion New Zealand Crown Research Institute and Radiata Pine Breeding Company Meeting, October 11, Rotorua, New Zealand.