Glimmer-MG

Glimmer-MG predicts protein-coding genes in environmental shotgun metagenomic DNA sequences using interpolated Markov models (IMMs) to enable assessment of microbial community functional potential.

Key Features:

• Interpolated Markov models (IMMs): Uses IMMs to distinguish coding regions from noncoding DNA in metagenomic reads.
• Phylogenetic parameterization: Incorporates phylogenetic classification of sequences to parameterize models according to evolutionary relationships.
• Sequence clustering and iterative retraining: Clusters sequences by likelihood of originating from the same organism and iteratively retrains models within each cluster using initial gene predictions to refine accuracy.
• Error modeling of indels and substitutions: Models insertion/deletion and substitution sequencing errors to adjust coding frames or pass through stop codons when errors are predicted.
• Read length and error-rate robustness: Handles varying read lengths and sequencing error rates for both simulated and real metagenomic datasets.

Scientific Applications:

• Metagenomic gene prediction: Identifying protein-coding genes in metagenomes to survey the functional potential of microbial communities across diverse ecosystems, including the human body.
• Functional genomics and comparative analyses: Supporting functional genomics studies by providing high sensitivity and precision across read lengths and error rates in comparative evaluations.

Methodology:

Applies interpolated Markov models (IMMs); performs phylogenetic classification for model parameterization; clusters sequences by likelihood of common origin and iteratively retrains models within clusters using initial gene predictions; models insertion/deletion and substitution sequencing errors to permit frame adjustments or passing of stop codons.