Learning genotype-phenotype associations from gaps in multi-species sequence alignments.

Area of Science:

• Evolutionary Biology
• Genetics
• Bioinformatics

Background:

• Understanding genotype-phenotype relationships is key to biological research.
• Existing methods often require extensive, hard-to-obtain data.
• Novel approaches are needed for efficient phenotype prediction.

Purpose of the Study:

• Introduce GAP, a machine learning framework for predicting binary phenotypes from sequence alignment gaps.
• Demonstrate GAP's utility in identifying genotype-phenotype associations and evolutionary insights.
• Provide a tool that uses readily available data for broad applicability.

Main Methods:

• Developed GAP, a neural network-based machine learning framework.
• Utilized multi-species sequence alignments, focusing on alignment gaps as input.
• Applied GAP to predict phenotypes in known and unknown species, and identify important genomic positions.

Main Results:

• GAP achieved perfect prediction accuracy for vitamin C synthesis in 34 vertebrates.
• Demonstrated high accuracy and power on simulated datasets.
• Identified novel candidate genes associated with vitamin C synthesis, enriched for immune function.

Conclusions:

• GAP is an effective tool for predicting genotype-phenotype associations using sequence alignment gaps.
• The framework offers a simple yet powerful method for evolutionary and genetic studies.
• GAP expands the possibilities for analyzing genotype-phenotype relationships across diverse species.