Find, Label, Annotate Genomes: FLAG, a fully automated tool for structural and functional gene annotation

Abstract

Recent advances in long-read sequencing technologies and the efforts of projects aimed at increasing the universe sequenced reference genomes, such as the Earth Biogenome project have led to a growth in the number of whole genome sequencing projects for non-model organisms. Still, the vast majority (6,880) of the 28,727 unique publicly available eukaryotic genomes lack gene structure annotations. Annotating genes in a genome of interest is a multistep process, and while there are many tools available for each step of the annotation, there are few end-to-end workflows that (i) run on multiple computing environments, (ii) automatically run without initial data training and (iii) perform accurately with fragmented genomes. Here we present "Find, Label, Annotate Genomes" (FLAG), a fully automated genome annotation workflow that does not require species-specific extrinsic evidence to create highly complete and accurate gene annotations of even highly fragmented genomes. Using FLAG, various eukaryotic genome assemblies, including 1 plant, 5 living animals, and 1 extinct animal were annotated. In all living species, FLAG annotations provided an average of 18% increase in complete BUSCO scores when compared to BRAKER2. Furthermore, FLAG has more accurately predicted the number of protein-coding genes, with an average error rate 15x lower than BRAKER2.