Genomic prediction informed by biological processes expands our understanding of the genetic architecture underlying free amino acid traits in dry Arabidopsis seeds

Abstract Amino acids are a critical component of plant growth and development, as well as human and animal nutrition. A better understanding of the genetic architecture of amino acid traits, especially in seeds, will enable researchers to use this information for plant breeding and biological discovery. Despite a collection of successfully mapped genes, a fundamental understanding of the types of genes and biological processes underlying amino acid related traits in seeds remains unresolved. In this study, we used genomic prediction with SNPs partitioned by metabolic pathways to quantify the contribution of primary, specialized, and protein metabolic processes to free amino acid (FAA) homeostasis in dry Arabidopsis seeds. First, we demonstrate that standard genomic prediction is effective for FAA traits. Next, we show that genomic partitioning by metabolic pathway annotations explains significant genetic variation and improves prediction accuracy for many FAA traits, including many trait-pathway associations that have not been previously reported. Surprisingly, SNPs related to amino acid and primary metabolism had limited effects on prediction accuracy for most FAA traits, with the largest effects observed for branched chain amino acids (BCAAs). In contrast, SNPs related to secondary and protein metabolism had a more extensive effect on prediction accuracy. The use of a genomic partitioning approach also revealed specific patterns across biochemical families, in which protein related annotations were the only category influencing serine-derived FAAs and primary and specialized metabolic pathways were the only categories contributing to aromatic FAAs. Based on these findings, we used pathway-guided association analysis to identify novel SNP associations for traits related to methionine, threonine, histidine, arginine, glycine, phenylalanine, and BCAAs. Taken together, these findings provide evidence that genomic partitioning is a viable strategy to uncover the complexity of FAA homeostasis and to identify candidate genes for future functional validation. Author summary Plant growth, development, and nutritional quality depends upon the regulation of amino acid homeostasis, especially in seeds. However, our understanding of the underlying genetics influencing amino acid content and composition remains limited, with only a few candidate genes and quantitative trait loci identified to date. As an alternative approach, we implemented multikernel genomic prediction to test whether or not genomic regions related to specific metabolic pathways contribute to free amino acid (FAA) variation in seeds of the model plant Arabidopsis thaliana. Importantly, this method successfully identifies pathways containing known variants for FAA traits, in addition to identifying new pathway associations. For several traits, the incorporation of prior biological knowledge provided substantial improvements in prediction accuracy. We present this approach as a promising framework to guide hypothesis testing and narrow the search space for candidate genes.