CodingQuarry: highly accurate hidden Markov model gene prediction in fungal genomes using RNA-seq transcripts

Area of Science:

• Genomics
• Bioinformatics
• Fungal Biology

Background:

• Accurate gene annotation is crucial for functional and comparative genomics, especially in non-model fungi.
• Traditional homology-based methods often fail due to incomplete or unreliable protein sequence data.
• Generalised hidden Markov models (GHMM) are valuable, and RNA-seq offers a cost-effective way to improve automated gene annotation.

Purpose of the Study:

• To develop and evaluate a novel fungal gene predictor that integrates RNA-seq data during the prediction phase.
• To improve gene prediction accuracy in fungi, particularly when protein homology data is limited.
• To leverage the strengths of GHMMs and RNA-seq for enhanced fungal genome annotation.

Main Methods:

• Developed CodingQuarry, a self-training GHMM fungal gene predictor.
• Incorporated assembled, aligned RNA-seq transcripts for both training and prediction.
• Utilized predictions directly from fungal transcript sequences to overcome assembly issues.

Main Results:

• CodingQuarry achieved high accuracy, perfectly predicting 91.3% of Schizosaccharomyces pombe genes and 90.4% of Saccharomyces cerevisiae genes.
• Demonstrated a 4-5% improvement over AUGUSTUS, the next best RNA-seq driven predictor.
• Validated improvements through comparisons against whole-genome annotations.

Conclusions:

• Successfully demonstrated a novel method for incorporating RNA-seq data into GHMM fungal gene prediction.
• Achieved high-quality fungal gene annotation without relying on protein homology or pre-existing gene sets.
• CodingQuarry is freely available and suitable for integration into genome annotation pipelines.