Exploring genotype-by-environment interactions in American Angus cattle

Data import

Documented in notebooks/import_regions.Rmd

Resulting data

• data/derived_data/import_regions/animal_regions.rds: direct and maternal weaning weight breeding values, actual weaning weights, associated regional metadata
• data/derived_data/import_regions/cg_regions.rds: weaning weight contemporary group solutions

Geographic QC & filtration

Removed:

• Rows from outside continental U.S.
• Rows where zip does not reside inside herd_state
• Rows with zip that do not assign to a region
• Regions 4 & 6
• Rows where region == 8 but K31 fescue not actually found in herd_state (herd_state %in% c("CA", "WA", "OR", "PA", "NY", "WI", "NJ", "MA", "MD", "DE", "MI", "CT", "RI", "NH", "NE") & region == 8)
• Rows where region == 9 but herd_state not in Upper Midwest/Northeast (herd_state %in% c("CA", "OR", "ID", "WA") & region == 9)

Other QC & filtration

• Converted weights, EBVs, and contemporary group solutions from lbs. to kg.

Removed:

• Rows with no recorded weight
• Rows prior to 1990
• Creep-fed calves
• Calves born via embryo transfer
• Single-sire and single-dam contemporary groups
• Records from calves with unknown parentage

Then after filtration, removed:

• Contemporary groups with fewer than 5 animals
• Contemporary groups containing phenotypic outliers (weights 3 SD +/- national mean)

Exploratory analyses

• Summarize phenotypic differences as well as differences in breeding values & CG solutions resulting from NCE weekly growth run in notebooks/regions_summary.Rmd
  • Summarize the same data as function of time in notebooks/trends_summary.Rmd
• Summarize phenotypic variance in AI sire progeny across regions in notebooks/progeny_variance.Rmd
• notebooks/genetic_trends.Rmd
• notebooks/environmental_variance.Rmd
• notebooks/weather.Rmd

Genetic correlations between regions

Data sampling

• In source_functions/setup.gibbs_varcomp.R
• Remove records from dams with calves in both the High Plains and one of the other regions (to avoid between-region MPE covariance issues)
• For 5 separate iterations, sample 50,000 +/- 500 records by zip code from the High Plains and each of the 6 other comparison ecoregions

Variance components estimation

• BLUPF90 Gibbs sampling ran in source_functions/gibbs_varcomp.snakefile
  • 1,000,000 total samples with burn in of 50,000 samples and every 100th sample retained
• Post-Gibbs analysis ran in source_functions/post_gibbs.snakefile and evaluated in notebooks/post_gibbs.Rmd
• Variance component estimation results in notebooks/gibbs_varcomp.Rmd

GWAS

• For each comparison region, extract