When does accounting for gene-environment interactions improve complex trait prediction? A case study with Drosophila...

Area of Science:

• Genetics
• Quantitative Biology
• Evolutionary Biology

Background:

• Gene-environment interactions (G×E) contribute significantly to complex trait variation across species.
• Previous studies show G×E improves prediction accuracy in agriculture but not in human genetics.

Purpose of the Study:

• To investigate specific scenarios where incorporating G×E enhances prediction accuracy.
• To understand why G×E benefits agricultural predictions more than human predictions.

Main Methods:

• Utilized Drosophila melanogaster lifespan data across varied environments and genotypes.
• Employed three cross-validation (CV) scenarios simulating different population relationships.
• Compared statistical models with and without G×E terms.

Main Results:

• G×E accounted for 8% of the variance in lifespan.
• Prediction accuracy improved with G×E inclusion only when genotypes overlapped between reference and test sets in CV.
• This overlap is common in agricultural populations but rare in human studies.

Conclusions:

• The utility of G×E in prediction models is context-dependent on the cross-validation scenario.
• The lack of overlapping genotypes in human studies may explain the limited observed benefits of G×E in prediction accuracy.
• Findings provide insight into the application of G×E in different biological and agricultural contexts.