Genomic selection (GS) consists in predicting breeding values of selection candidates, using a large number of genetic markers. An important question in GS is the determination of the number of markers required for a good prediction. When the genetic map is too sparse, it is likely to observe some imperfect linkage disequilibrium: the alleles at a gene location and at a marker located nearby vary. We tackle here the problem of imperfect linkage disequilibrium and we present theoretical results regarding the accuracy criteria, the correlation between predicted value and true value. Illustrations on simulated data and on rice real data are proposed.