A new approach for applied nutritional models: Computing parameters of dynamic mechanistic growth models using genome-wide prediction

Nutritional models have long been used as decision support tools by the livestock industry. Despite the advance of genomic prediction, these two disciplines have evolved separately. Because model parameters are responsible to describe between-animal variability, we propose an integration of nutritional models with genomics by means of such parameters. Two dynamic mechanistic models of cattle growth were used: Cornell Cattle Value Discovery System (CVDS) and Davis Growth Model (DGM). We estimated SNP marker effects for their parameters and also for observed phenotypes. Then, we compared what would be the best prediction scenario - model simulation with parameters computed from genomic data or genomic prediction directly on higher phenotypes. We found that genomic prediction on dry matter intake (DMI) and average daily gain (ADG) are still a better approach than using CVDS for predictions. Dry matter required (DMR), a CVDS-predicted value for DMI had higher correlation (r=0.253) with observed DMI than results from genomic prediction (r=0.07). DGM had better predictive ability (r=0.38) than genomic prediction on ADG (r=0.098). This is also the case for whole-body protein (r=0.496) and fat at slaughter (r=0.505) whose predictions were better with DGM than genomic prediction performed on the observed traits (r=0.194 and r=0.183, respectively). When contrasting simulations with genomically predicted parameters to those with regularly computed ones, CVDS showed moderate correlation and low bias between simulations of DMR (r=0.966; b=0.9%) and ADG (r=0.645; b=5.5%). Although further model development is necessary, the DGM with subject-specific parameters computed from genotypic data was a better option for predicting phenotypes than genomic prediction alone. In addition, simulations with genomically and regularly computed parameters match at a reasonable extend. This is the main argument to call attention from the research community that our approach may pave the way for the development of a new generation of applied nutritional models, especially towards individual-based simulations with subject-specific parameters computed from genomic information.