Pan-genome-wide association analysis reveals complementary variant-class associations with key agronomic traits in oil palm

Abstract Graph-based pan-genomes can improve the representation of genomic variation underlying complex agronomic traits. We conducted a pan-genome-wide association study of 384 oil palm accessions comprising wild African dura populations and elite tenera-derived breeding materials. Mapping short-read sequences to a pan-genome graph identified approximately 32.8 million variants. Structural indels differentiated the germplasm groups and were proportionally enriched in promoter-proximal and intronic regions. Four primary GWAS panels comprising 1,292,996 unique SNPs, multi-nucleotide polymorphisms (MNPs), micro-indels, and larger indels were evaluated, with intermediate- and structural-indel subsets additionally analyzed. The framework recovered the established major-effect SHELL locus controlling shell-to-fruit ratio, providing an internal positive control. Palm-height analysis identified five panel–model-specific FDR-significant candidates, with a structural indel showing the strongest cross-model support. This indel was located near a gene encoding a GATA transcription factor, and a two-marker combination was prioritized for independent validation. Multiple variant classes identified candidate associations with oil-to-dry mesocarp ratio, and MNP analysis detected association signals for rachis length. These findings support integrating multiple variant classes in oil palm association mapping and provide candidates for validation in independent populations and functional studies.