Glimmer-HMM

Glimmer-HMM predicts eukaryotic gene structures ab initio using Generalized Hidden Markov Models (GHMMs) to model exons, introns, and intergenic regions.

Key Features:

• Generalized Hidden Markov Model framework: Models introns of each phase, intergenic regions, and four types of exons within a GHMM.
• Splice site models: Integrates splice site models adapted from GeneSplicer to improve exon–intron boundary prediction.
• Decision tree adaptation: Utilizes a decision tree approach derived from GlimmerM to assist identification of coding regions.
• Interpolated Markov models: Employs interpolated Markov models for both coding and noncoding regions.
• Probabilistic submodels: Supports combinable probabilistic submodels, including Maximal Dependence Decomposition trees and interpolated Markov models.
• Modular C/C++ architecture: Implemented in C/C++ with a modular, reusable, and extensible codebase.
• Retrainable and reconfigurable models: Allows model retraining and reconfiguration using new training data.

Scientific Applications:

• Eukaryotic genome annotation: Applied to annotate complex eukaryotic genomes, including Aspergillus fumigatus and Toxoplasma gondii.
• Gene structure analysis: Predicts exon–intron structures to assist studies of gene organization and function in eukaryotes.

Methodology:

Integrates probabilistic submodels within a GHMM framework by combining splice-site predictions from GeneSplicer, decision-tree analysis derived from GlimmerM, interpolated Markov models, and Maximal Dependence Decomposition trees.