Project Details:

Increasing the rate of genetic gain for yield in soybean breeding programs

Parent Project: Increasing the rate of genetic gain for yield in soybean breeding programs
Checkoff Organization:North Central Soybean Research Program
Categories:Breeding & genetics
Organization Project Code:NCSRP
Project Year:2018
Lead Principal Investigator:Leah McHale (The Ohio State University)
Co-Principal Investigators:
William Beavis (Iowa State University)
Silvia Cianzio (Iowa State University)
Asheesh Singh (Iowa State University)
William Schapaugh (Kansas State University)
Dechun Wang (Michigan State University)
Jianxin Ma (Purdue University)
Katy M Rainey (Purdue University)
Leah McHale (The Ohio State University)
Brian Diers (University of Illinois at Urbana-Champaign)
Matthew Hudson (University of Illinois at Urbana-Champaign)
Aaron Lorenz (University of Minnesota)
Andrew Scaboo (University of Missouri)
Grover Shannon (University of Missouri)
George Graef (University of Nebraska)
David Hyten (University of Nebraska at Lincoln)
Steven Clough (USDA/ARS-University of Illinois)
Show more

Contributing Organizations

Funding Institutions

Information and Results

Click a section heading to display its contents.

Project Summary

Increases in soybean yield through breeding have been slower than growers expect, with the rate of yield increases for soybean substantially less than that for corn. A collaborative study led by Diers of a historic set of MG II-IV varieties released from 1923 to 2008 revealed a recent rate of genetic gain of 0.43 bu/ac/yr, whereas reports of genetic gain in corn generally range from 1.0 to 1.2 bu/ac/yr.
There are several possible targets for improving the rate of gain in soybean grain yield if the equation of genetic gain is considered. The rate of genetic gain from a single cycle of selection can be predicted by the breeders’ equation (?G = S[sA/sP]), where ?G is the change in yield from the parental generation to the progeny of selected parents in a single breeding cycle, S is the selection differential between parents and progeny, sA is additive genetic variance, and sP is total phenotypic variance, including variance due to the environment and error. Accordingly, there are several possible targets for improving the rate of genetic gain in soybean grain yield, including increasing the selection intensity (S), increasing measurement accuracy, increasing genetic diversity and additive genetic variance, and decreasing the amount of time required for each breeding cycle. Through coordinated activities to be carried out across twelve breeding programs in the North Central region, Objectives 1, 2 and 3 will each address one or more of these target areas, with the overall goal of increasing the rate of genetic gain for yield in soybean. Objective 4 is aimed at developing a metric to accurately assess realized genetic gain for yield on an annual basis.
The proposed activities build on and encompass an earlier proposal led by Dr. Aaron Lorenz, “Initiation of a genomic selection pipeline for public soybean breeders in the North Central Region” as well as the concluded NCSRP funded research led by Drs. George Graef, Brian Diers, and Randall Nelson.

Project Objectives

OBJECTIVE 1: Increasing selection intensity and decreasing non-genetic sources of variability through improved progeny row testing
OBJECTIVE 2: Increasing selection coefficient and decreasing length of breeding cycle through genomic selection
OBJECTIVE 3: Increasing additive genetic variance
OBJECTIVE 4: Development of a metric to estimate genetic gains on an annual basis

Project Deliverables

Within the existing pipelines of the co-investigators’ breeding programs, progeny row selection will take place as usual; however, breeders will also participate in the selection experiment to compare the agronomic performance of lines selected by breeders using their usual selection methods to lines selected through prediction of yield performance using new sources of data and information. The selection experiment will be conducted in all states, with equal numbers of lines selected from each breeding program and testing coordinated across locations. Specifically, in years 1 & 2, approximately 5000 progeny rows in each breeding program will be considered for selection using at least three selection categories. Each breeding program is responsible for the general management and experimental design of a typical progeny row test. In years 2 & 3, coordinated preliminary yield trials will be planted by each program to test the performance of lines within the selection categories. The germplasm will advance over years and forward breeding progress will be made.
Purpose: Evaluate increases in the rate of genetic gains attained by selecting in early stages of the breeding pipeline using data integrated from various sources.
Task 1: Collection of additional data in all progeny rows.
The 11 breeders participating in objective 1 will grow ~5000 progeny rows in years 1 & 2. Each breeder will submit block-range-row information of each line for spatial correction of field variation, parental information of each line will be included to enhance selection accuracy among families, and breeder selection information (yes/no), along with any other data used by the breeder to make selections will be included. Additional performance data collected may include yield, flower and maturity dates (R1 plus R8), or canopy coverage, or a combination of some or all of these data as determined by each breeder. The reproductive period i.e., the time between R1 and R8, and canopy coverage are targeted because we have evidence that these measurements are associated with yield. Also, canopy coverage is a new opportunity for soybean breeders because it can be measured with drones. Pedigrees can also be divided over environments.
Task 2: Selections from progeny rows.
In the 5000 progeny rows considered for each breeder at least three selection categories will be applied: 1) breeders' selections, 2) breeding value calculated using pedigree and spatial adjustments, and 3) breeding value plus other phenotypic data (minimum yes/no selection if no other data are collected). For selection categories 2 and 3, yield for each line will be predicted with a statistical model that incorporates all available information, adjusts all observations for local field variation, and infers trait values for all lines based on the pedigree relationships between the lines. Selection category 2 tests the increase in selection accuracy attainable without investing in additional resources. This selection category takes information that breeders already have and applies it in new ways. Selection category 3 tests further increases in selection accuracy gained when resources are applied to collecting additional performance data (see Task 1 and section 1.a, above).
An 8% selection intensity will be applied to select approximately 400 lines for each of the three categories, totaling ~1200 lines from each breeding program. The selected lines will very likely overlap between the three categories, so we can add more than three selection categories- the final numbers are determined empirically. For validation of methods, a small number of total plots will be allocated to selection of lines with poor performance.
Task 3: Preliminary yield trials to evaluate the increase in the rate of genetic gains.
All selections will be divided into preliminary yield trial experiments, and trials will be organized by maturity with lines randomly assigned to environments, standard checks, and an experimental design that adjust for field variation. Plots will be unbordered and at least 12’ in length. The accuracy of progeny rows selection schemes and change in genetic gain will be calculated based on the yield rank information from the preliminary yield trials. Genetic gain will be estimated as the difference in the expected breeding values from the progeny row test of the first generation and the preliminary yield trial in the second generation.
Expected outcome: All breeders’ lines ranked simultaneously for yield breeding value, maturity prediction and a metric of diversity. Lines selected using additional sources of information may provide higher rank-order correlation with the performance of preliminary yield trails. Lines in preliminary yield trials can be submitted for collection of genomic information as part of Objective 2, Task 6.

Progress of Work

Updated April 17, 2018:
The SOYGEN group had two project meetings this Spring, the first was at the Soybean Breeders’ Workshop on February 13, 2018. During this meeting we discussed updates and action points for each main project objective. Specific attention was paid to potential modifications for FY19. It was suggested that additional, high impact phenotyping of the PIs grown as part of objective 3 would be extremely valuable and maximize the use of the multi-location field trials that are unlikely to be grown again outside of this project. Present in St. Louis for this meeting: Aaron Lorenz, George Graef, Danny Singh, Katy Martin Rainey, Ed Anderson, Brian Diers, Dechun Wang, Pengyin Chen, David Hyten, Steven Clough. Present on the conference line: Leah McHale, Matthew Hudson, Jianxin Ma, Andrew Scaboo, Bill Beavis.

The second meeting was web-based Zoom meeting on March 30, 2018. The primary objective for this meeting was to ensure that plans were in place and understood for the collaborative field trials (Objective 1 and 3). Important points of discussion included the addition of in-season selections made by Rainey’s group in 2018 in order to ease harvest of the ~5000 progeny rows and reduce the number of lines actually needed for harvest. In addition, future potential changes to the project for FY19 were discussed, these included additional phenotyping for Objective 3 (R1, R3, R5 dates, V4 image data, canopy temperatures, and weather data). Participating breeders will attempt to gather this data in FY18 as well, though it was not budgeted for this granting period. For FY19, we also proposed to collect functional genotype data (maturity genes and branching genes) for the germplasm panel. Work with tomentella-derived lines has been re-focused to the identification of introgression of tomentella DNA into soybean. Attending Zoom meeting: McHale, Beavis, Clough, Graef, Hudson, Ma, Lorenz, Martin Rainey, Scaboo, Schapaugh, Wang.

A full progress report is attached as a PDF.

View uploaded report PDF file

Updated October 31, 2018:
In the second year of our objective 1, aimed at improving progeny row testing, we are evaluating the selections from progeny row testing made in year 1 as well as carrying out a second year of the methods of progeny row testing from year 1. Preliminary results are pending the report of yield from this harvest.

Objective 2, aimed at decreasing the time for each cycle of selection by implementing a genomic selection strategy using public resources is making good progress. The genomic selection study from the SoyNAM population has been completed and in the final stages of manuscript preparation. Databasing and genotyping from the Uniform Regional Trials is continuing on an ongoing basis as yearly entries are submitted and yield data is collected. Preliminary tests of predictive ability using the URT training set has been exploring using cross validation on adjusted means of URT lines. Further refinement of the analysis indicates respectable prediction accuracies ranging from 0.50 to 0.68 across maturity groups. Finally, a cost effective genotyping strategy has been developed.

Objective 3 aims to implement several strategies for increasing genetic diversity in breeding programs. During the 2018 season, we conducted the first year of our 2-year evaluations of the 250 PI accession selected for the validation set from across the distribution of predictions for yield of ~9,700 untested PI accessions. As this is likely the only time that such a collection of PIs will be evaluated in replicated trials across many environments, we are taking the opportunity to collect as much phenotypic data as reasonable. In addition, this objective has successfully taken several approaches to identify potential yield loci through QTL analysis, GWAS, as well identification of signatures of selection through population diversity statistics.

Finally, in Objective 4, an introductory video about genetic gain has been developed and delivered to the NCSRP. Several steps have been made in an effort to develop metrics to accurately estimate realized genetic gains. Syngenta has delivered phenotypic data (yield, maturity, planting date, longitude x latitude) for lines grown in annual field trials for maturity groups II, III and IV for years 2009 – 2017. Software has been written that will merge genotypic and phenotypic data from the URT’s. For the actual development of methods, a recent PNAS publication (Li et al, 2018) provides a novel method for removing the GxE contribution from the non-genetic (environmental) effects, thus leaving only genetic and an indexed genotypic value for specific environments as a means to calculate realized genetic gains.

Details for each objective and task are in the report attached below.

View uploaded report PDF file

Final Project Results

Benefit to Soybean Farmers

Performance Metrics

Project Years

YearProject Title (each year)
2021SOYGEN2: Increasing SB genetic gain for yield & seed composition by developing tools, know-how & community among public breeders in the NC US
2021SOYGEN2: Increasing SB genetic gain for yield & seed composition by developing tools, know-how & community among public breeders in the NC US
2020Inceasing soybean genetic gain for yield by developing tools, know-how and community among public breeders in the north central US
2020Inceasing soybean genetic gain for yield by developing tools, know-how and community among public breeders in the north central US
2019Increasing the rate of genetic gain for yield in soybean breeding programs
2018Increasing the rate of genetic gain for yield in soybean breeding programs
2017Increasing the rate of genetic gain for yield in soybean breeding programs