Predicting Fitness-Related Traits Using Gene Expression and Machine Learning.

Area of Science:

• Evolutionary Biology
• Genomics
• Transcriptomics

Background:

• Connecting genomic elements to observable traits is key in evolutionary biology.
• Gene expression serves as an intermediate phenotype between genes and macroscopic traits.
• High-dimensional transcriptomic data presents statistical challenges.

Purpose of the Study:

• To investigate the association between fitness components and gene expression data.
• To explore the utility of machine learning (ML) in analyzing transcriptomic data for natural selection studies.
• To identify specific genes or gene categories under selection in Ipomoea hederacea.

Main Methods:

• Utilized two distinct machine learning approaches.
• Analyzed gene expression data from Ipomoea hederacea grown under field conditions.
• Correlated gene expression patterns with fitness components.

Main Results:

• Identified evidence that the expression of photosynthesis-related genes is likely under selection.
• Found that genes involved in stress and light responses were significant predictors of fitness.
• Demonstrated the effectiveness of ML models for analyzing smaller sample sizes in genetic studies.

Conclusions:

• Machine learning models are valuable tools for understanding natural selection from transcriptomic data, even with limited sample sizes.
• Gene expression patterns, particularly for photosynthesis, stress, and light response genes, are associated with fitness in natural populations.
• This study highlights the potential of integrating ML with transcriptomics to advance evolutionary biology research.