### Table of Contents

# PREDF90

## Summary

Predicts direct genomic value (DGV) for young animals based on only genotypes, i.e., $\mathbf{\hat{u} = Z\hat{a}}$ where $\mathbf{\hat{u}}$ is DGV, $\mathbf{\hat{a}}$ is the estimated SNP effect, and $\mathbf{Z}$ is the matrix of SNP content that will be centered by twice the current allele frequency (based on the animals used to compute SNP effects).

The prediction is based on SNP effects obtained from POSTGSF90. For young animals that were not included in the previous analysis, DGV can be calculated using the “snp_pred” file from POSTGSF90.

PREDF90 requires some output files from POSTGSF90 and a genotype file for the animals to be predicted. It does not accept a parameter file but takes command-line options.

## Command-line options

--snpfile <name>

Provides the SNP file for animals to be indirectly predicted. PREDF90 will ask for the SNP file name if this command is not present. The SNP file has the same format as for PREGSF90.

--acc

Computes reliability for indirect predictions.

It requires `OPTION snp_p_value`

in BLUPF90+ and `OPTION snp_var`

in POSTGSf90. It reads “snp_var”, a file with SNP PEC created by POSTGSF90

--acc_type

Select 1.0 for dairy cattle (Reliability) or 0.5 for beef cattle (BIF accuracy) (default 1.0).

--use_diagG_acc

Uses inbreeding (F) from **G** in the denominator of the reliability formula: reliability = 1-PEV/($\sigma\mathbf{}_{u}^{2}$(1 + F)) Aguilar et al. (2020).

--use_mu_hat

Adds the base ($\mathbf{\hat{\mu}}$) for DGV, so the values are comparable to GEBV. See Legarra et al. (2021) and Lourenco et al. (2018) for more details.

--use_var_mu_hat

Considers the variance of $\mathbf{\hat{\mu}}$ when calculating the reliability of DGV and is automatically turned on if `--use_mu_hat`

and `--acc`

are present.

--help

Shows the main options.

### Usage

predf90 --snpfile new_genotypes.txt --use_mu_hat --acc --use_diagG_acc

With these commands, predf90 will compute indirect predictions for the animals in new_genotypes.txt, including $\mathbf{\hat{\mu}}$ (i.e., DGV = $\mathbf{\hat{\mu}}$ + $\mathbf{Z\hat{a}}$), computing reliabilities adjusted for inbreeding in **G**.

### Output

It creates a file named “SNP_predictions” that contains ID, call rate, DGV, reliability (if --acc is present)