biorxiv.org
Gene functional-state covariance reveals modular structure in the Arabidopsis thaliana dispensable genome
Classical multilocus theory predicts that selection acting on combinations of loci can maintain non-random associations among allelic states, causing selection to act on multilocus genotypes rather than individual alleles. Across eukaryotic species, pangenome studies have documented thousands of genes that vary in functional status between individuals, yet this variation is analyzed as if each gene evolves independently. Here we test for non-random association within the dispensable pangenome across 1,012 natural genotypes of Arabidopsis thaliana to better understand the structure and evolution of eukaryotic pangenomes. Genome-wide pairwise association testing of loss-of-function alleles across genotypes identified 150,033 significantly covarying gene pairs among 5,679 unique genes after controlling for kinship and population structure. Most associations were between genes on different chromosomes, and covarying gene pairs showed greater co-expression and functional similarity than expected by chance. Network analysis revealed a non-random higher-order architecture comprising stable functional communities with distinct biological functions and structured connectivity among communities. The enriched functions and connectivity of communities revealed a tight coupling of modules involving plant immunity, environmental response, and specialized metabolism, highlighting how these systems integrate and evolve in response to dynamic ecological pressures. Taken together, these results show that the dispensable genome has a complex higher-order structure driven by functionally interacting genes, and that variation in gene content provides a population-scale resource for investigating the organization and evolution of biological systems. ### Competing Interest Statement The authors have declared no competing interest. Division of Integrative Organismal Systems, 2208944
