AI solutions for evolutionary genomics of nonmodel species

Area of Science:

• Evolutionary Biology
• Genomics
• Bioinformatics

Background:

• Large-scale genomic datasets enable new evolutionary biology questions.
• Studying nonmodel species presents technical challenges for traditional methods.
• Artificial intelligence (AI), particularly deep neural networks, shows promise for analyzing genomic data from nonmodel organisms.

Purpose of the Study:

• Highlight recent deep learning trends for evolutionary inference in nonmodel species.
• Propose novel research directions for AI algorithm development in this field.
• Address challenges like data missingness, uncertainty, and unknown parameters.

Main Methods:

• Reviewing current AI applications in evolutionary genomics.
• Identifying strategies for handling data imperfections (missingness, uncertainty).
• Developing AI to infer selection with unknown demographic and genomic parameters.
• Implementing AI to detect selective sweeps in low-sample, uncertain data.

Main Results:

• Demonstrated an original AI implementation for detecting selective sweeps in nonmodel species.
• Showcased AI's capability to handle low sample sizes and uncertain sequencing data.
• Validated AI's potential for evolutionary inferences under challenging conditions typical of nonmodel organisms.

Conclusions:

• Nonmodel organisms offer unique opportunities for developing general-purpose, data-driven evolutionary inference methods.
• AI and deep learning are crucial for advancing evolutionary studies in diverse species.
• Resource sharing and inclusive frameworks are essential for leveraging AI in evolutionary biology.