Utilizing plasma biochemical indicators to improve prediction of economic traits in crossbred duck population

Plasma biochemical indicators are commonly considered as direct indicators of metabolism and health in both animals and humans. However, genomic predictions for biochemical traits and their downstream utility in predicting economic traits remain unclear. This study explores the genetic parameters of 18 plasma biochemical indicators in a population of 1,059 ducks and evaluates their potential to enhance genomic predictions for 53 economic traits. High heritability was observed for cholinesterase (CHE, 0.57), while traits like lactate dehydrogenase (LDH) and creatine kinase (CK) showed negligible heritability (< 0.1). LDH exhibited strong positive correlations with hydroxybutyrate dehydrogenase (HBDH, 0.95), aspartate aminotransferase (AST, 0.88), and CK (0.83), while high-density lipoprotein cholesterol (HDLC) demonstrated moderate negative correlations with CK (-0.81), LDH (-0.7), and HBDH (-0.58). Using data from 941 genotyped ducks, we estimated predictive reliabilities of biochemical traits under pedigree-based BLUP, genomic BLUP (GBLUP), and Bayesian models. GBLUP outperformed pedigree BLUP, with an average reliability improvement of 0.024, though Bayesian models offered incremental gains for specific traits (e.g., +0.165 for CHE under BayesN). Hierarchical clustering and principal component analysis revealed distinct metabolic networks: Cluster 1 (e.g., triglycerides, uric acid) correlated with leg muscle and viscera traits, while Cluster 2 (e.g., cholesterol, albumin) associated with breast muscle and fat deposition. Integrating plasma indicators into multi-trait GBLUP models improved reliability for key economic traits, notably feed conversion ratio (FCR, +0.068 with glucose) and residual feed intake (RFI, +0.019 with direct bilirubin). These findings highlight the potential of plasma biomarkers as auxiliary traits for genomic selection, particularly in optimizing feed efficiency and carcass composition, while underscoring the need for trait-specific model strategies.