Machine Learning Approach for Muscovy Duck (Cairina moschata) Semen Quality Assessment.

Area of Science:

• Animal Science
• Reproductive Biology
• Biotechnology

Background:

• Artificial insemination in Muscovy ducks (Cairina moschata) requires accurate assessment of ejaculate quality.
• Traditional sperm kinetic analysis may not fully capture the functional capacity of sperm for breeding.

Purpose of the Study:

• To develop a comprehensive machine learning (ML) approach for assessing Muscovy duck ejaculate quality.
• To integrate sperm kinetics with non-kinetic parameters like enzyme activities and DNA methylation for enhanced predictive capacity.

Main Methods:

• Sperm kinetics were assessed using Computer-Assisted Sperm Analysis (CASA).
• Non-kinetic parameters included vitality, enzyme activities (AP, CK, LDH, GGT), and total DNA methylation.
• Machine learning models were trained using these features to classify sperm sample quality.

Main Results:

• Samples were classified based on motility and DNA methylation, showing significant differences in kinetic parameters and live normal sperm cells.
• Enzyme activities (AP, CK) differed significantly and correlated with LDH and GGT.
• Machine learning models, particularly neural network and gradient boosting, highlighted DNA methylation's importance for accurate classification, with parameters like ALH, VCL, LDH, and VAP being key predictors.

Conclusions:

• Integrating non-kinetic parameters, especially DNA methylation, with ML models significantly enhances the accuracy of duck sperm quality assessment.
• This approach facilitates the selection of kinetically and morphologically superior sperm samples for artificial insemination.
• The study provides a promising method for improving breeding efficiency in Muscovy ducks.