Integrating pan-genome analysis, GWAS, and interpretable machine learning to prioritize trait-associated structural...

Area of Science:

• Genomics
• Plant Science
• Crop Improvement

Background:

• Structural variations (SVs), particularly presence-absence variations (PAVs), significantly influence crop domestication and trait development.
• Pan-genome analysis offers comprehensive PAV insights but faces cost and sample size limitations.
• Genome-wide association studies (GWAS) identify trait-marker associations in large populations but often miss PAVs and struggle with linkage disequilibrium.

Purpose of the Study:

• To systematically explore PAVs in foxtail millet (Setaria italica) using a graph-based pan-genome.
• To identify genomic regions and variants associated with leaf, leaf sheath, and leaf pulvinus color in a large population.
• To develop a cost-efficient framework integrating pan-genome, GWAS, and machine learning for crop trait studies.

Main Methods:

• De novo assembly of eight reference-quality foxtail millet genomes.
• Construction of a graph-based pan-genome to analyze PAVs.
• Genome-wide association study (GWAS) on 344 millet accessions for leaf color traits.
• Application of interpretable machine-learning models for variant prioritization.

Main Results:

• Identification of a 5002-bp Copia element insertion and other key variants in chromosome 7 (26.84–26.94 Mb).
• Association of these variants with phenotypic variations in leaf color traits.
• Demonstration of an integrative approach combining pan-genome, GWAS, and machine learning.

Conclusions:

• The study presents an effective and cost-efficient framework for investigating agronomic traits in crops.
• The identified variants provide insights into the genetic basis of leaf color in foxtail millet.
• This integrative strategy enhances variant detection and prioritization for crop improvement.