Evidence-based gene predictions in plant genomes

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Automated gene building provides cost-effective genome annotations.
• Evidence-based gene builders require species-specific transcriptional data (proteins, cDNAs, ESTs).
• Plant genome sequencing outpaces EST/cDNA data generation, posing a challenge for gene annotation.

Purpose of the Study:

• To develop an evidence-based gene build system utilizing cross-species transcriptional evidence.
• To address the limitation of lacking species-specific ESTs/cDNAs in plant genome annotation.
• To assess the efficacy of using cross-species ESTs for gene prediction in plants.

Main Methods:

• Developed the Gramene pipeline, an evidence-based gene build system.
• Utilized Ensembl computing infrastructure and a novel data processing scheme.
• Tested the pipeline on annotated genomes of Arabidopsis thaliana (dicot) and Oryza sativa (monocot).

Main Results:

• Cross-species ESTs within monocot or dicot classes serve as valuable evidence for gene predictions.
• The Gramene pipeline, using only EST and cross-species evidence, generated a plant gene set comparable in quality to human gene sets.
• The Gramene pipeline outperformed widely used ab initio gene prediction programs in rice when using cross-species gene products.

Conclusions:

• The Gramene pipeline effectively uses cross-species ESTs for robust plant gene annotation.
• This approach significantly enhances gene building for plant genomes with limited species-specific data.
• The pipeline demonstrates potential for application to other newly sequenced genomes, improving annotation efficiency and accuracy.