Fig 4. Genetic variation and correlations. The posterior mode and 95% HPD interval of the genetic correlation between the log-contrast of Z11-16:OAc on 14:Ald and each of the other three log-contrasts is shown for the starting generation (S, in black), the low (blue) and high (red) lines. 111 females were phenotyped for the starting generation. For the low line, 443, 284, and 330 phenotyped females were in generation blocks 1-3, 4-6, and 7-9, respectively. For the high line 465, 324, and 344 phenotyped females were in generation block 1-3, 4-6, and 7-9, respectively. Pedigree data were obtained for 1,065 mating pairs prior to the first generation of selection and for an additional 2,250 and 2,236 pairs for the low and high line, respectively.
Second, the partitioning of genetic variance among gPC1 and gPC2 evolved. The first two genetic principle components changed in magnitude of their eigenvalue (i.e. the amount of genetic variance they describe) and in their loadings (i.e. in their correlation with the pheromone components; Fig 5). More of the total genetic variance across the four components was in the direction of gPC1 in the high and low lines (~80%) compared to the starting generation (59.1%). This means that more of the genetic variance of the sex pheromone was biased towards a single dimensions in phenotypic space. Moreover, variation in the component under selection, Z11-16:OAc, was more strongly associated with gPC2 and less with gPC1 in generations after compared to before selection, while the other traits showed a trend in the opposite direction (Fig 5). These changes in the G matrix resulted in a more modular pheromone blend, because Z11-16:OAc, which was under selection here, became uncoupled from the components that were not under selection.