DeepGenomeScan of 15 Worldwide Bovine Populations Detects Spatially Varying Positive Selection Signals

Area of Science:

• Genomics
• Evolutionary Biology
• Animal Science

Background:

• Identifying genomic regions under selection is crucial for understanding evolution and adaptation.
• Traditional methods struggle with complex, spatially varying selection signals.
• Deep learning offers advanced capabilities for detecting subtle selection signatures.

Purpose of the Study:

• To apply deep learning for detecting spatially varying selection signatures in bovine populations.
• To uncover novel genes and genomic regions under selection in cattle.
• To enhance understanding of genetic mechanisms driving bovine adaptation and traits.

Main Methods:

• Utilized the DeepGenomeScan deep learning framework.
• Analyzed selection signatures across 15 global bovine populations.
• Compared findings with the Bovine Genome Variation Database.

Main Results:

• Detected novel selective sweep hotspots in the bovine genome.
• Identified key genes linked to physiological and adaptive traits, including milk protein and fat percentages.
• Discovered 38 previously undetected genes under selection, primarily related to milk and meat yield and quality.

Conclusions:

• Deep learning provides new insights into spatially varying selection in cattle.
• Findings contribute to understanding bovine genomic diversity and adaptation.
• This research lays groundwork for genetic improvement and conservation in livestock genomics.