PlantGeneAnn: a high-accuracy ab initio plant gene annotation tool fine-tuned from a genome foundation model

PlantGeneAnn achieves a new state-of-the-art F1 score of 0.856 at the exon level in plant gene annotation across 20 species, outperforming existing tools.

• Why it matters: Accurate structural annotation is essential for understanding plant genomes, but current methods depend on RNA-seq and homologous evidence often unavailable for non-model species, creating a significant gap.
• What they did: The authors developed PlantGeneAnn, a plant-optimized ab initio gene annotation tool trained on 40 angiosperm species, combining a DNA encoder, embedding decoder, and semi-Markov pipeline, with fine-tuning capabilities.
• The result: PlantGeneAnn's high accuracy, lower memory requirements, and effective fine-tuning enable reliable annotation of large, diverse, and non-model plant genomes on standard workstations, supporting broad genomic research.