This Read Me file was generated on 2024-11-15 by Charlotte Brault

Title of Dataset

Supporting data for Leveraging historical trials to predict Fusarium head blight resistance in wheat breeding programs.

Author Information

Principal Investigator Contact Information

James A. Anderson

Department of Agronomy and Plant Genetics, University of Minnesota 411 Borlaug Hall, 1991 Buford Circle St. Paul, MN 55108

Email: ander319@umn.edu{.email}

ORCID: 0000-0003-4655-6517

Associate or Co-investigator Contact Information

Jason D. Fiedler

USDA-ARS Cereal Crops Research Unit Edward T. Schafer Agricultural Research Center, Fargo, ND, USA

Email: jason.fiedler@usda.gov{.email}

ORCID: 0000-0001-7736-4484

Associate or Co-investigator Contact Information

Charlotte Brault

Department of Agronomy and Plant Genetics, University of Minnesota 411 Borlaug Hall, 1991 Buford Circle St. Paul, MN 55108

Email: cbrault@umn.edu{.email} / charlotte.brault@live.com{.email}

ORCID: 0000-0001-7892-4236

• Date of data collection (single date, range, approximate date): 1995-2023
• Geographic location of data collection (where was data collected?): Northern Great Plains-Midwest (MN, ND, SD, Manitoba)
• Information about funding sources that supported the collection of the data: U.S Wheat and Barley Scab Initiative, USDA-ARS.
• Overview of the data (abstract): Genomic prediction was used as a tool to improve the genetic resistance of spring wheat to Fusarium head blight, using data from a historical nursery and from two current breeding programs. Parameters related to genomic predictive ability were thoroughly studied. Predictive ability within the historical nursery was medium to high for all the eight traits studied and all methods gave similar results. Marker density did not affect predictive ability compared to training set size. Training set optimization had mixed results depending on the trait. We tested several prediction scenarios useful in a breeding context by harnessing the historical dataset in the training set for predicting breeding lines. While the lack of genetic relatedness decreased the accuracy of genomic prediction, we showed that breeding programs could benefit from this historical data by incorporating its information into training models, thus reducing the phenotyping effort.

Sharing/Access Information

Licenses/restrictions placed on the data: CC0

Data and file description

Data

URSN and URN populations

Phenotypic data

• URN_cleaned_table_phenot_1992_2023.tsv Curated phenotypic data for the URN population phenotyped for FHB related traits.
• URSN_cleaned_table_phenot_1995_2023.tsv Curated phenotypic data for the URSN population.

Genotypic data

• URN_URSN_90K_Filt_geno_formatted_imputed_mafFilt_gdose.rds input genotypic file for the URSN population for the 90K array
• URN_URSN_3K_geno_formatted_imputed_mafFilt_gdose.rds for the 3K genotypic array.
• Table of genomic information for the URN and URSN populations: genotype_info_URN_URSN.tsv.

NDSU breeding program

• NDSU_URSN_across-pop_list_inputs_v2.rds Combined phenotypic and genomic data from NDSU and URSN populations.

UMN breeding program

• geno_combined_URSN_UMN-BP_imputed_Beagle_v2.rds Genotypic data from UMN breeding program and URSN population.
• pheno_combined_URSN_UMN-BP_5traits_v2.rds Phenotypic data from UMN breeding program and URSN population.
• 2022 PY Entry list.xlsx List of PY entries for the UMN breeding program for 2022 phenotypic data, with the genotype name, locations, and pedigree.
• 2023 PY Entry list.xlsx List of PY entries for the UMN breeding program for 2023 phenotypic data, with the genotype name, locations, and pedigree.
• UMN_parent_compo.csv Table of the composition of the genotypes used from the UMN breeding program.

Code

• useful_functions.R Script which wraps custom useful functions for pre-processing, analyzing data and producing results
• subsetMrkCV.R Script to subset markers multiple times for testing the effect of marker density.
• subsetIndsCV.R Script to subset individuals multiple times to estimate the imputation accuracy.
• imputationQuality.R Script to estimate the imputation accuracy based on predicted genotypes.
• beagle.01Mar24.d36.jar Beagle software for imputation.

Bash

This folder contains the bash scripts used to run some of the analyses on the cluster.

• launch_array.sh Script to launch the array job for marker selection (with subset_MrkCV.R) and genomic prediction within the URSN population.
• launch_Rmdscript.sh Script to launch various Rmd scripts for genomic prediction on the cluster.
• loop_imputation_3k_to_90k.sh Script to launch the genomic imputation from the 3K to the 90K genotypic array in the URN-URSN population.

Analysis scripts

Prepare inputs

Phenotypic data

• extract_blups_URSN_data.Rmd. In this script, we extract the BLUPs from the phenotypic data for the URSN population, from URSN_cleaned_table_phenot_1995_2023.tsv table. It outputs a table of BLUPs: URSN_geno_blups_lme4-values.tsv, and a table of fitting information: URSN_lme4_fit-info_all_traits.tsv.
• extract_blups_URN_FHB.Rmd. In this script, we extract the BLUPs from the phenotypic data for the URN population, from URN_cleaned_table_phenot_1992_2023.tsv table. It outputs a table of BLUPs: URN_geno_blups_lme4-values_FHBtraits.tsv, and a table of fitting information: URN_lme4_fit-info_all_traits_FHBtraits.tsv.
• analyze_BLUPs_URN_URSN.Rmd. In this script, we analyze the BLUPs from the URSN and URN populations. It outputs plots of phenotypic structure and genetic correlations among traits.

Genomic data URSN

• explo_geno_URN_URSN.Rmd. In this script, we explore the genomic data for the URSN population. It outputs plots of the genetic structure, additive relatedness and linkage disequilibrium.
• URN_URSN_measure_imput_quality.Rmd. In this script, we load the imputation results for the URSN population from the 3K to the 90K array.

URSN and UMN data

• explore_UMN_BP_URSN.Rmd. In this script, we explore the genomic and phenotypic data for the UMN breeding program and the URSN population. It outputs plots of the genetic structure (PCA) and the genetic correlation and distribution between populations.

Genomic prediction within URSN

• genom_pred_URN_URSN.Rmd. In this script, we perform genomic prediction within the URSN population, using various inputs.
• genom_pred_URN_URSN_exclude_orga.Rmd. In this script, we perform genomic prediction by excluding in turn each organization.
• genom_pred_URN_URSN_subsetGeno.Rmd. In this script, we perform genomic prediction by subsetting genotypes from the 3K genotypic array to match the number of individuals in the 90K genotypic array.
• genom_pred_URSN_TSoptSFSI.Rmd. In this script, we perform genomic prediction within the URSN population, using training set optimization with sparse selection index.
• analyze_genom_pred_results_URSN.Rmd. In this script, we combine and analyze the results of the genomic prediction within the URSN population. All results were combined in a table: URSN_genomic_prediction_all_combined_results.tsv.

Genomic prediction between URSN and breeding programs

• genom_pred_acrosspop_URN_URSN.Rmd. In this script, we perform genomic prediction between the URSN population and the URN population.
• genom_pred_across-pop_UMN-BP_URSN.Rmd. In this script, we perform genomic prediction between the URSN population and the UMN breeding program.
• genom_pred_across-pop_UMN-BP_subset_UMN.Rmd. In this script, we perform genomic prediction between the URSN population and a subset of the UMN lines.
• genom_pred_across-pop_NDSU_URSN.Rmd. In this script, we perform genomic prediction between the URSN population and the NDSU breeding program.
• genom_pred_across-pop_NDSU_URSN_subset_NDSU.Rmd. In this script, we perform genomic prediction between the URSN population and a subset of the NDSU lines.
• analyze_across_pop_GP.Rmd. In this script, we gather and analyze the results of the genomic prediction between the URSN population and the URN/UMN/NDSU population. All results were combined in a table: URN_UMN_NDSU_genomic_prediction_all_combined_results.tsv.

Genomic prediction with training set optimization

• genom_pred_acrosspop_TSopt_URN_URSN.Rmd. In this script, we perform genomic prediction between the URSN population and the URN population, using training set optimization.
• genom_pred_across-pop_TSopt_UMN-BP_URSN.Rmd. In this script, we perform genomic prediction between the URSN population and the UMN breeding program, using training set optimization.
• genom_pred_across-pop_TSopt_NDSU_URSN.Rmd. In this script, we perform genomic prediction between the URSN population and the NDSU breeding program, using training set optimization.

Other

• plot_figures.Rmd. In this script, we generate the figures for the manuscript, specifically, we load the PCA genetic structure plots for the NDSU and UMN breeding programs and the URSN population and plot it together.
• index.Rmd. This script is the index of the project, it contains the description of the project and the links to the different scripts.
• about.Rmd. This script contains the information about the project.
• license.Rmd. This script contains the license information.

Output

• genotype_info_URN_URSN.tsv Table of genomic information for the URN and URSN populations.
• URSN_geno_blups_lme4-values_95-23.tsv Table of BLUPs for the URSN population, URSN_df_deregressed_blups_8traits_95-23.tsv the table of deregressed BLUPs, and URSN_geno_blups_var_lme4-values_95-23.Rdata Rdata file of the variance of BLUPs.
• URSN_lme4_fit-info_all_traits_95-23.tsv Table of fitting information for the URSN population.
• URN_geno_blups_lme4-values_FHBtraits.tsv Table of BLUPs for the URN population and URN_geno_blups_var_lme4-values_FHBtraits.Rdata Rdata file of the variance of BLUPs.
• URN_lme4_fit-info_all_traits_FHBtraits.tsv Table of fitting information for the URN population.
• res_GP_TSopt_within_across-pop_URN_URSN_7traits.tsv Table of results for the training set optimization genomic prediction within the URN population and between the URSN and URN populations.
• URN_UMN_NDSU_genomic_prediction_all_combined_results.tsv Table of results for the genomic prediction between the URSN population and the URN/UMN/NDSU population, generated in analyze_across_pop_GP.Rmd.
• URSN_genomic_prediction_all_combined_results.tsv Table of results for the genomic prediction within the URSN population.
• fig_repro_URSN_UMN_BP_PCA_inds_geno_orga.rds. This file contains the PCA outputs of the genetic structure of the UMN breeding program and the URSN population, it has been generated in explore_UMN_BP_URSN.Rmd and is used in plot_figures.Rmd.
• output/figures folder. This folder contains the figures generated in all scripts.
• output/genom_pred folder. This folder contains the outputs of genomic predictions.
• output/lme4 folder. This folder contains the outputs BLUPs prediction and fitting information.
• output/imputation folder. This folder contains the outputs of the imputation quality.

The docs folder contains the rendered Rmd scripts in html format.

Software and versions

• R version v4.4.0 and RStudio v2024.04.01+748
• vcftools 0.1.17
• Beagle v5.4

Version changes

November 2024

Update following manuscript revision:

• Update genotype names with released cultivar names
• Add genotype random subset in URSN genomic selection
• Add genomic imputation accuracy assessment:
  • Bash script and R-scripts to subset individuals and markers
  • Imputation for each chromosome, save imputed genotypes
  • Measure of imputation accuracy
  • Gather all imputation results