FragGeneScan

FragGeneScan predicts protein-coding regions from short and error-prone sequencing reads, with emphasis on metagenomic samples.

Key Features:

• Hidden Markov Model (HMM): Uses an HMM framework that integrates sequencing error models and codon usage patterns for gene prediction.
• Error Tolerance: Maintains high prediction accuracy in the presence of sequencing errors and shows improved performance over MetaGene as error rates increase.
• Short Read Optimization: Optimized for short reads and outperforms Glimmer and MetaGene on fragmented metagenomic sequences.
• Enhanced Accuracy: For 400-base reads with a 1% sequencing error rate it improves prediction accuracy by approximately 62%, and for 100-base error-free reads it enhances accuracy by about 18%.
• Gene Recovery: Recovers a substantially higher number of genes compared to MetaGene and identifies over 90% of genes detected through homology searches.

Scientific Applications:

• Metagenomic gene prediction: Predicts protein-coding regions directly from short reads without relying on genome assembly, enabling analysis of complex environmental samples.
• Gene recovery and discovery: Recovers more genes than MetaGene and uncovers novel genes with no known homologs in protein databases, supporting studies of microbial diversity and function.

Methodology:

Integrates sequencing error models and codon usage statistics within a hidden Markov model framework.