Exploring the statistical nature of independent chromosome segments

Previous research in genomics developed the concept of independent chromosome segments and suggested that genomic prediction works on capturing the segment effects rather than on LD to individual QTL. The number of independent chromosome segments (M-e) was posited as 4N(e)L, a function of effective population size (N-e) and genome length (L) in Morgans (M). The objective of this study was to determine whether the physical segments are approximately consecutive haplotype blocks of length 1/4 M, with the number of haplotype blocks for each physical location equal to Ne. In a simulated population, the number of animals randomly selected as reference animals is represented as Na. For all animals, the genome was split into equal-sized segments represented as Ns. For each specific location, segments of non-reference animals were assigned the most similar segment of one reference animal. Genomic analyses estimated the value of the segment effects, and breeding values were a sum of all segment effects for a specific animal. Accuracies of segment effects were calculated by correlating the true breeding values (TBV) and the breeding values based on segments. Segment effect accuracies were compared with the true accuracy calculated by the correlation of TBV and genomic estimated breeding values (GEBV) computed using GBLUP. Accuracies were maximized at N-a=N-e, N-s=4 L, but they were not as high as in GBLUP. Accuracies may be smaller using the statistical concept of segments due to approximations based on computing limitations, as the origin of each segment and the recombination sites were unknown in the simulation. Therefore, random animals served as reference animals, and each non-reference animal received the most similar segment of a reference animal instead of a linear combination of such segments. Genomic selection acts partially on 1/4 M long chromosome segments, and using the statistical definition of segments moderately explains the accuracy. More complex simulations are needed to investigate the issue thoroughly.