Leveraging Multiple Layers of Data To Predict Drosophila Complex Traits

Area of Science:

• Genomics
• Quantitative Genetics
• Systems Biology

Background:

• Accurate prediction of complex traits from genetic data is vital for personalized medicine and agriculture.
• Current prediction accuracy for complex traits using genomic data remains low.
• Understanding the genetic architecture of complex traits is a key challenge.

Purpose of the Study:

• To compare the predictive accuracy of gene expression versus genotypes for three complex traits in Drosophila melanogaster.
• To evaluate the combined predictive power of genotype and gene expression data.
• To assess the impact of incorporating gene ontology (GO) categories on prediction accuracy.

Main Methods:

• Utilized whole genome sequences, deep RNA sequencing, and phenotype data from ~200 Drosophila melanogaster Genetic Reference Panel (DGRP) lines.
• Developed prediction models using genotypes, gene expression levels, and combined data.
• Incorporated gene ontology categories as additional information layers for genomic variants and transcripts.

Main Results:

• Gene expression levels provided higher prediction accuracy than genotypes for starvation resistance.
• Prediction accuracy was similar for chill coma recovery and lower for startle response when comparing expression and genotypes.
• Models combining genotype and expression did not outperform single-component models.
• Including gene ontology categories significantly increased prediction accuracy for all three traits.

Conclusions:

• Gene expression data can be more predictive than genotypes for certain complex traits.
• Integrating gene ontology information substantially improves prediction accuracy for complex traits.
• This study elucidates the genetic architecture of complex traits and highlights the value of multi-source data integration.