Genomic Selection on Novel Variation

A key property of genomic selection is that it can accelerate the breeding cycle by enabling the selection of individuals prior to phenotyping them. Assuming this practice, genomic selection is "blind" to new causal mutations occuring among selection candidates, since these mutations do not appear in the training population. To explore the extent to which this blindness hurts genomic relative to phenotypic selection on novel (as opposed to standing) variation, I simulated selection under mutation-drift-selection balance.  Under a variety of settings, genomic selection under-performed phenotypic selection, suggesting that caution over the long term is warranted in contemplating a shift to genomic selection.  For a few generations after the transition from historic phenotypic selection to genomic selection, however, genomic selection often strongly out-performed phenotypic selection.  This observation suggests that phenotypic selection more effectively lifts new favorable mutations to a frequency at which they are unlikely to be lost, whereas, once they have reached that frequency genomic selection more rapidly brings them to fixation.