Whole-Genome Annotation with BRAKER

• Authors: Katharina Hoff, Ernst Moritz Arndt Universitat Greifswald; Alexandre Lomsadze, Georgia Tech; Mark Borodovsky, Emory University; Mario Stanke, Ernst Moritz Arndt Universitat Greifswald
• Journal Title: GENE PREDICTION: METHODS AND PROTOCOLS
• Volume: Volume 1962
• Publisher: HUMANA PRESS INC | 2019-01-01, Pages 65-95
• Type of Work: Article | Post-print: After Peer Review
• Publisher DOI: http://doi.org/10.1007/978-1-4939-9173-0_5
• Subjects: Biology, Genetics; Chemistry, Biochemistry; Biology, Cell; Health Sciences, Epidemiology
• Keywords: Science & Technology; Life Sciences & Biomedicine; Biochemical Research Methods; Biochemistry & Molecular Biology; Protein-coding genes; Gene prediction; AUGUSTUS; GeneMark-ES/ET; RNA-Seq reads; Protein mapping to genome; Genome annotation pipeline; BRAKER; ALIGNMENT; IDENTIFICATION; PREDICTION; SEQUENCES; FUNGAL

Abstract

BRAKER is a pipeline for highly accurate and fully automated gene prediction in novel eukaryotic genomes. It combines two major tools: GeneMark-ES/ET and AUGUSTUS. GeneMark-ES/ET learns its parameters from a novel genomic sequence in a fully automated fashion; if available, it uses extrinsic evidence for model refinement. From the protein-coding genes predicted by GeneMark-ES/ET, we select a set for training AUGUSTUS, one of the most accurate gene finding tools that, in contrast to GeneMark-ES/ET, integrates extrinsic evidence already into the gene prediction step. The first published version, BRAKER1, integrated genomic footprints of unassembled RNA-Seq reads into the training as well as into the prediction steps. The pipeline has since been extended to the integration of data on mapped cross-species proteins, and to the usage of heterogeneous extrinsic evidence, both RNA-Seq and protein alignments. In this book chapter, we briefly summarize the pipeline methodology and describe how to apply BRAKER in environments characterized by various combinations of external evidence.