Traits related to body maintenance processes, such as feed intake and residual feed intake, show low influence of selection at the nucleus level on the rate of genetic change in the production level. Therefore, optimal selection strategies for feed efficiency should be based on feed intake data recorded in commercial conditions measured on crossbred pig. When using crossbred animal as reference population we have to take into account that QTL allele in breed A can behave different that in breed B. To find these differences, and employ them in genomic prediction, a model with line origin of allele will be implemented. First SNP genotypes for crossbred pigs will be phased and compared to haplotypes observed in the parental purebred lines to derive which allele originates from each parental line. Second, this information will be used to construct breed-specific partial relationship matrices. With this approach the effect of a SNP allele from breed A when it is present in a crossbred AB animal can be differentiated from the effect of a SNP allele from breed B. The data used to test this methodology included 1.3k three-way-crossbred pigs (A(BC)), with individual daily feed intake observations from 25 kg until 120 kg and genotyped with a 10k SNP chip and imputed to 60k. In addition, we had 1.3k two-way-crossbred pigs (BC) and 9.7k purebred pigs (1.7k A line, 3.5k B line and 4.5k C line), all genotyped with a 60k SNP chip. We expect that this methodology will allow us to better estimate SNP effects and therefore, provide a tool to better select purebred animals based on crossbred performance for feed efficiency.