Inverse modeling to estimate deep phenotypes of the feed utilization complex in dairy cows

Area of Science:

• Animal Science
• Genetics
• Agricultural Engineering

Background:

• Deep phenotypes offer superior measures for improving feed efficiency in cows compared to traditional traits.
• Estimating these complex traits at scale is challenging, necessitating advanced modeling techniques.

Purpose of the Study:

• To estimate deep phenotypes for feed utilization using inverse modeling.
• To evaluate the precision of these estimates and the impact of feed quality data granularity.
• To determine if estimated deep phenotypes represent biological traits.

Main Methods:

• Utilized inverse modeling combining the LiGAPS-Dairy mechanistic model and a genetic algorithm (GA).
• Integrated measured phenotypes (FPCM, DMI, BW, gross feed efficiency) with genetic algorithm optimization.
• Assessed precision through model fitness (objective function value) and reproducibility (concordance correlation coefficient).

Main Results:

• Inverse modeling successfully estimated 8 out of 9 deep phenotypes with reasonable precision.
• Model fitness was adequate, with objective function values indicating good performance across feed quality scenarios.
• Reproducibility analysis suggested that 8 deep phenotypes align with biological traits, not just model proxies.

Conclusions:

• Deep phenotypes can be reliably estimated using inverse modeling for applications in animal breeding.
• Eight specific deep phenotypes show promise as novel, biologically relevant traits for enhancing feed efficiency.
• Further research is warranted to integrate these deep phenotypes into breeding programs for improved animal performance.