Abstract Genetic analysis of transcriptional profiling experiments is emerging as a promising approach for unraveling genes and pathways that underlie variation of complex biological traits. However, these genetical genomics approaches are currently limited by the high cost of microarrays. We studied five different strategies to optimally select subsets of individuals for transcriptional profiling, including (1) maximizing genetic dissimilarity between selected individuals, (2) maximizing the number of recombination events in selected individuals, (3) selecting phenotypic extremes within inferred genotypes of a previously identified quantitative trait locus (QTL), (4) purely random selection, and (5) profiling animals with the highest and lowest phenotypic values within each family–gender subclass. A simulation study was conducted on the basis of a linkage map and marker genotypes were derived from data on chromosome 6 for 510 F2 animals from an existing pig resource population and on a simulated biallelic QTL with pleiotropic effects on performance and gene expression traits. Bivariate analyses were conducted for selected subset sample sizes of 80, 160, and 240 individuals under three different correlation scenarios between the two traits. The genetic dissimilarity and phenotypic extremes within genotype methods had the smallest mean square error on QTL effects and maximum sensitivity on QTL detection, thereby outperforming all other selection strategies, particularly at the smallest proportion of samples selected for gene expression profiling (80/510).