CORAL: Accurate annotation of compact genomes using long-read RNA-seq, demonstrated in Oikopleura dioica

The burst of high-quality genome assemblies has created an urgent need to improving genome annotation tools, especially for problematic species such as those with compact, fast-evolving genomes in which standard annotation tools underperform when facing short intergenic regions, overlapping UTRs, non-canonical splicing and widespread operons. We present CORAL (Compact-genome Oriented RNA-based Annotation using Long-reads), a long-read RNA-seq-based workflow for de novo annotation of compact eukaryotic genomes, together with GAMBA (Gene Aggregation tool for Multicistronic Block Annotation), a Rust-based module that detects polycistronic transcriptional units directly from GTF annotations and is integrated within CORAL. Using the exceptionally compact chordate genome of Oikopleura dioica as a case study, CORAL increases the number of annotated genes, improves gene-model completeness and reduces chimeric predictions relative to existing annotations, while GAMBA recovers thousands of operons, most of which are independently supported by splice-leader data. Together, CORAL and GAMBA provide an accurate, scalable framework for annotating compact, structurally atypical genomes using long-read transcriptomic data alone. Availability and ImplementationCORAL and GAMBA are open source and available at https://github.com/EvoDevoGenomics-UB/CORAL and https://github.com/nurie05/gamba-tool.