Efficient evidence-based genome annotation with EviAnn.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Machine learning-based ab initio gene finding has dominated eukaryotic genome annotation.
• Historically, gene expression data and protein homology were underutilized due to cost and availability.
• Advances in sequencing technology now provide abundant gene expression data.

Purpose of the Study:

• To develop a novel evidence-based eukaryotic gene annotation system named EviAnn.
• To leverage transcript alignments and protein homology for more accurate gene structure prediction.
• To improve upon existing gene annotation packages.

Main Methods:

• EviAnn utilizes a data-driven approach, relying on transcript alignments and protein-sequence homology.
• It builds exon-intron structures based on empirical evidence rather than solely ab initio methods.
• The system was compared against state-of-the-art packages like BRAKER3, MAKER2, and FINDER.

Main Results:

• EviAnn consistently outperforms BRAKER3, MAKER2, and FINDER when given identical input data.
• The system achieves high accuracy in eukaryotic gene annotation.
• Annotation of a mammalian genome was completed in under an hour on a single multi-core server, demonstrating computational efficiency.

Conclusions:

• EviAnn represents a significant advancement in evidence-based eukaryotic gene annotation.
• Its data-driven approach and efficiency offer a superior alternative to existing methods.
• The system is freely available as an open-source tool.