TSEBRA

TSEBRA selects transcripts from BRAKER1 and BRAKER2 outputs to produce an evidence-supported set of protein-coding gene models for eukaryotic genomes.

Key Features:

• Integration of Evidence Types: Combines RNA-seq evidence used by BRAKER1 with cross-species protein database evidence used by BRAKER2 to leverage both types of extrinsic support.
• Transcript Selection Mechanism: Applies rule-based comparisons of scores for overlapping transcripts, evaluating support from introns, start/stop codons, RNA-seq alignments, and homologous protein evidence.
• Benchmarking and Accuracy: Demonstrated higher gene prediction accuracy than BRAKER1 or BRAKER2 alone across 11 species and shown competitive performance with combiners such as EVidenceModeler.

Scientific Applications:

• Novel genome annotation: Improves gene annotations in novel eukaryotic genome projects when both RNA-seq and protein evidence are available.
• Genomic research and comparative annotation: Produces more evidence-supported gene models for downstream analyses and comparative studies of eukaryotic genomes.

Methodology:

TSEBRA takes gene predictions generated independently by BRAKER1 and BRAKER2 and applies a transcript selection algorithm that compares scores of overlapping transcripts, prioritizing support from introns, start/stop codons, RNA-seq data, and homologous protein evidence.