TSEBRA: transcript selector for BRAKER.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• BRAKER pipelines (BRAKER1, BRAKER2) automate eukaryotic gene annotation using statistical models and extrinsic evidence.
• BRAKER1 utilizes RNA-seq data, while BRAKER2 uses cross-species protein databases.
• Current BRAKER versions do not effectively integrate both RNA-seq and protein evidence simultaneously, leading to unexploited data.

Purpose of the Study:

• To develop a software tool, TSEBRA, that integrates gene predictions from BRAKER1 and BRAKER2.
• To enhance the accuracy of eukaryotic gene annotation by leveraging both RNA-seq and protein evidence.

Main Methods:

• TSEBRA selects optimal gene predictions from BRAKER1 and BRAKER2 outputs.
• It employs a rule-based system to compare overlapping transcript scores based on RNA-seq and protein evidence support.
• Evaluated on 11 eukaryotic genomes.

Main Results:

• TSEBRA achieves higher accuracy than BRAKER1 or BRAKER2 when used independently.
• TSEBRA demonstrates competitive performance compared to the EVidenceModeler tool.
• The software effectively combines complementary evidence for improved gene prediction.

Conclusions:

• TSEBRA is a user-friendly and efficient software for eukaryotic gene annotation.
• It enhances the BRAKER pipeline by integrating both RNA-seq and homologous protein evidence.
• Facilitates more comprehensive and accurate gene prediction sets for novel genome projects.