DGAT1 is a major gene influencing yield in cattle with effect magnitudes that vary by breed, implying that DGAT1 interacts with other parts of the genome. This study aims to determine whether consideration of interactions between DGAT1 and single nucleotide polymorphisms (SNP) could improve the accuracy of genomic predictions for fat yield in Holsteins. The initial data set included 1,143 305-d fat yield records from 358 Holstein cows in 11 herds that were genotyped for DGAT1 and 45,187 SNPs. A series of analyses included a random animal effect, a fixed DGAT1 effect, a single fixed SNP effect, and a fixed interaction of DGAT1 with the same SNP. This analysis was repeated for each SNP, 191 of which significantly interacted with DGAT1 (P < 0.05). The significant SNP were then included in a single analysis and backward eliminated until a group of 41 significant DGAT1 × SNP interactions remained. To evaluate whether consideration of the interactions improved genetic prediction, 5 data sets were created where all data from 2 to 3 herds was eliminated. Breeding values were subsequently estimated for those cows whose data were excluded. A sixth validation data set was created where data from the youngest cows was excluded. Four breeding values were estimated for each validation data set: EBVBASE (random animal effect only), EBVDGAT1 (EBVBASE plus DGAT1 effects), EBVSNP (EBVDGAT1 plus SNP effects), and EBVI (EBVSNP plus DGAT1 × SNP effects). Validation models were more significant when 305-d yield was regressed on EBVI (F = 37.41) than on EBVSNP (F = 14.03), EBVDGAT1 (F = 14.51), or EBVBASE (F = 0.84) for the young cow validation data. Validation models were also most significant when 305-d fat yield was regressed on EBVI in 4 of the 5 validation groups where data were excluded from specific herds. Similarly, the correlation between EBVI and EBV from a full model with all data included was highest in 4 of the 5 herd validation analyses (mean = 0.48, range = 0.29 to 0.64) and for the young cow evaluation (0.54), whereas the lowest was always EBVBASE (range = −0.10 to 0.27). Consideration of SNP interactions with DGAT1 may yield more accurate genomic predictions for fat yield.