Genome annotation was carried out using a standardized DOE Joint Genome Institute and Phytozome pipeline in which each pangenome genome first received independently predicted gene models based on multiple lines of transcript and protein evidence. In a second round, high-confidence models were used to hard-mask each assembly and were projected across the pangenome with BLASTX and EXONERATE to generate additional predictions, followed by filtering and manual curation to retain only well-supported, non-repetitive gene models.