Machine Learning For Evolutionary Genetics And Molecular Evolution

Over the past decade, the rapid expansion of large-scale data and advances in computational power have allowed machine learning (ML), especially deep learning, to reshape many areas of biological research. Evolutionary genetics and molecular evolution are also poised for a similar transformation. In this review, we discuss key advances and ongoing challenges in applying ML to the study of genetics and evolution, and we highlight the potential of artificial intelligence to connect genotype, phenotype, and evolutionary history.

Highlights

Recent advances in machine learning (ML) especially deep learning have enabled breakthroughs in biology and are poised to play an important role in evolutionary biology.

• ML models are increasingly capable of connecting genetic variation to molecular function, integrating structural, regulatory, and phenotypic data at multiple scales.
• ML models can now be used directly on raw SNPs, haplotypes, or allele frequency spectra, reducing reliance on summary statistics and enabling the discovery of previously unrecognized evolutionary patterns.
• Emerging approaches move beyond correlation, using causal inference to generate mechanistic insights into evolutionary processes.
• Integrating multiomics data through ML is revealing hidden layers of complexity, such as epistasis, regulatory evolution, and phenotypic plasticity, that shape adaptation and diversification.

Machine learning for evolutionary genetics and molecular evolution, Trends In Genetics (open access)

Astrobiology, genomics, evolution, Machine Learning,