Machine learning and hypothesis driven optimization of bull semen cryopreservation media

Area of Science:

• Animal Science
• Biotechnology
• Machine Learning

Background:

• Cryopreservation is vital for dairy herd genetics management.
• Current cryoprotectant extender media show limited improvement in post-thaw sperm recovery.
• Optimizing extender media is challenging due to complex component interactions and sample sizes.

Purpose of the Study:

• To optimize bull sperm cryopreservation media using supervised machine learning.
• To identify optimal extender media formulations for improved post-thaw sperm motility.
• To establish a machine learning framework applicable to other cell cryopreservation.

Main Methods:

• Utilized supervised learning models, including artificial neural networks and Gaussian process regression (GPR).
• Identified 11 media components from existing literature for bull semen cryopreservation.
• Trained models on 200 extender-post-thaw motility pairs and tested on 32 pairs.

Main Results:

• Coupling differential evolution with GPR enhanced median post-thaw motility from 52.6% ± 6.9% to 68.3% ± 6.0%.
• Several optimized media formulations significantly outperformed control media.
• Demonstrated the effectiveness of machine learning in optimizing cryopreservation media.

Conclusions:

• Machine learning provides a novel and effective approach for optimizing bull sperm cryopreservation media.
• The developed method significantly improves post-thaw sperm motility, benefiting dairy herd genetics.
• This study presents a template for applying machine learning to optimize cryopreservation for various cell types.