Assessing inbreeding networks from partial correlations and information theory in rabbits

Accurate estimates of inbreeding (F) are crucial for the management of populations in selection schemes. Traditionally, inbreeding has been estimated from pedigree-based information. Currently, genome-based estimates of inbreeding can be obtained from SNP chip data. The correlation between pedigree and different genome-based estimates of inbreeding is usually discussed in the literature. However, some of these correlations could be spurious. Using partial correlations and information theory (PCIT) it is possible to distinguish a significant association between two variables which is independent from associations with a third variable. The objective of this study is to implement PCIT to assess the relationship between different estimates of inbreeding and phenotypes using a selected population of rabbits. Pedigree data and genomic information from a 200K chip were available. After applying filtering criteria, the data set constituted 490 animals genotyped for 114,604 autosomal SNPs. Phenotypic records for average dairy gain (ADG) were also accessible for full (f) and restricted (r) feeding. Pedigree-based inbreeding estimates (ped) were decomposed in ancient (A), intermediate (I) and recent (R) inbreeding. Accordingly, genomic estimates of inbreeding based on runs of homozygosity (roh) were also decomposed in A, I and R. Other genomic estimates of inbreeding were also evaluated: VanRaden (van), proportion of homozygous SNPs (snp), proportion of homozygous SNPs for the minor allele frequency (PHoMA) and the compression efficiency (CE). The results indicate that FrohR is an essential node with 6 edges, one of which is negative with FrohA (- 0.31). CE is a balanced node with 2 positive (Fsnp and FrohR) and 2 negative (FrohA and FpedI) edges. Fvan is the less related node with only one positive edge with PHoMA (0.70). When including phenotypic information, FpedI is correlated negatively with ADGf (-0.11) and positively with ADGr (0.10). In addition, ADGr has a negative edge with FrohA (-0.11). These results indicate that PCIT approach allows inferring meaningful associations between estimates of inbreeding.