Users can easily implement simple linear model (LM), BLUP, PBLUP, GBLUP (rrBLUP) and SSGBLUP, HIBLUP can automatically switch to one of them according to the data type provided by users. The following table details the required or non-required flags for different model in HIBLUP:
| Model | –qcovar/–dcovcar | –rand | –pedigree | –bfile |
|---|---|---|---|---|
| LM | √ | × | × | × |
| BLUP | O | √ | O | O |
| PBLUP | O | O | √ | × |
| GBLUP | O | O | × | √ |
| SSGBLUP | O | O | √ | √ |
√: required; ×: not required; O: optional.
rrBLUP model is theoretically the same with GBLUP model, users can just add a flag --snp-effect on the base of GBLUP model to obtain the SNP effect for rrBLUP model.
IMPORTANT
It should be noted that, the PBLUP model can be pretty efficient owing to the high performance of FSPAK technology on the very sparse ‘LHS’ of mixed model equation. However, HIBLUP is designed for genome-based genetic evaluation, in which the ‘LHS’ is no longer sparse, the ‘V’ matrix based strategy that is achieved in HIBLUP would be more competent for this case. Although users can fit PBLUP model by using HIBLUP, it may be slower and memory-costing compared with those of the software, e.g., DMU, BLUPF90. Therefore, we recommend using DMU or BLUPF90 when fitting a model only involves pedigree information.