301 Inferring phenotypic causal networks of reproductive traits in Landrace pigs in Japan. (7th December 2018)

• Record Type: Journal Article
• Authors: Okamura, T; Nishio, M; Ishii, K; Takahashi, K; Yoshino, J; Kobashikawa, H; Jordão de Magalhães Rosa, G; Satoh, M; Sasaki, O
• Is Part Of: Journal of animal science. Volume 96(2018)Supplement 3
• Page Start: 114
• Page End: 114
• Publication Date: 2018-12-07
• Subjects: phenotypic causal network -- reprodutive trait -- pig; Livestock -- Periodicals; Livestock; Electronic journals; Periodicals; 636.005
• DOI: 10.1093/jas/sky404.250
• Languages: English
• ISSNs: 0021-8812

Abstract

Abstract: Knowledge regarding phenotypic causal network among reproductive traits in pigs is important for predicting the result of specific interventions in the system, for example the effect of cross-fostering on litter weight gain, as well as for more accurately prediction of genetic merit of females. The objective of this study was to infer the phenotypic causal network involving reproductive traits in pigs from data without cross-fostering. Phenotypic traits were first parity litter size (LS), number born alive (NBA), number of live piglets at day 2 (LP2), at day 5 (LP5) and at day 21 (LP21), and litter weight at day 0 (LW0) and at day 21 (LW21), recorded on 841 sows from three landrace populations in Japan. The phenotypic causal network among the phenotypes was estimated using the inductive causation algorithm after adjustment for genetic effects on the data. A standard multiple-trait model was fitted using a Bayesian approach to obtain posterior samples of the (co)variance matrix of the phenotypes conditional to unobservable additive genetic effects. Statistical decisions regarding partial correlations were based on different highest posterior density (HPD) interval contents. Temporal and biological information was utilized to perform additional edge orienting, overriding the algorithm output when necessary. The magnitude of the network edges was estimated using a structural equation model, in which parent nodes of a given trait were considered as covariates for that specific trait. The results suggested two branched paths from LS to LW21 with HPD 95%, i.e. LS→NBA→LP2→LP5→LP21→LW21 and LS→NBA→LW0→LW21 (Figure 1a). With HPD 90%, an additional path was added to the network, i.e. LP5→LW21 (Figure 1b). These networks were consistent with prior biological knowledge and can be useful for more accurately predicting genetic performance of sows for reproductive traits even when cross-fostering is implemented.