Assessment and refinement of eukaryotic gene structure prediction with gene-structure-aware multiple protein sequence...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Accurate eukaryotic gene identification is crucial for understanding genomes.
• Existing gene prediction methods lack robust evaluation, hindering accuracy assessment.
• Scarcity of proper assessment methods limits critical evaluation of predicted gene structures.

Purpose of the Study:

• To develop and evaluate a novel gene-structure-aware multiple sequence alignment method for eukaryotic gene prediction.
• To devise an iterative refinement strategy for improving inaccurately predicted gene structures.
• To assess the accuracy of annotated gene structures in plant genomes.

Main Methods:

• Gene-structure-aware multiple sequence alignment using amino acid sequences of homologous genes.
• Iterative refinement of predicted gene structures using spliced alignment with consensus sequences or homologs.
• Application to cytochrome P450 and ribosomal protein genes across 47 plant genomes.

Main Results:

• The method provides reliability information for each predicted gene structure.
• Analysis of 47 plant genomes revealed defects in 50-60% of annotated gene structures.
• A majority of defect-containing gene candidates were improved using the iterative refinement method.

Conclusions:

• Gene-structure-aware multiple protein sequence alignment significantly enhances eukaryotic gene structure prediction quality.
• The method is applicable to evolutionarily stable protein-coding gene families across all kingdoms of life.
• This strategy offers a valuable approach for refining gene annotations in newly sequenced genomes.