The introduction of molecular information into genetic evaluation systems is currently under research. Based on an equivalent method, we developed from existing theory a new alternative strategy for the prediction of gene effects and especially their smooth integration into genetic evaluations. Underlying hypothesis were based on the idea that knowledge of genotypes will not affect overall additive genetic variance but only change expected values of genetic effects for animals with known genotypes. However, all animals could not be genotyped. Thus, the developed equations were modified to allow the integration of the known genotype for a portion of the population. This strategy was tested for the mh locus (responsible for the double-muscling phenotype) in dual-purpose Belgian Blue cattle. The genotype was determined for 123 bulls and 1940 cows (+/+ 19.5, mh/+ 39.3 and mh/mh 41.2%). These animals had 11,150 daughters with test-day (TD) records. The genotypes were incorporated into a modified genetic evaluation based on the current routine multi-trait multibreed test-day model used in the Walloon Region (Belgium). Data used included 12,829,309 TD records for 689,057 dairy cows in production. The pedigree file contained 1,606,024 animals (cows with TD records and ancestors). Computation of the modified mixed model equations was done solving iteratively two systems of equations, one for the polygenic effects and one for the gene effect until the relative differences in the gene solutions were below 10-5. A linear extrapolation was also used to speed up the convergence of gene effects. As expected, the mh locus exerts negative effects on milk production traits. For the first three lactations, the average estimated allelic substitution effects were -158.7 kg milk, -8.93 kg fat and -5.64 kg protein per lactation (305 days).