HMMER3

HMMER3 performs protein sequence similarity searches and alignments using profile hidden Markov models (profile HMMs) to detect homology and annotate protein families.

Key Features:

• Speed and efficiency: Achieves up to a 100-fold speed improvement over earlier versions and is comparable in speed to BLAST for protein searches, enabling large-scale database queries.
• Profile HMM search and alignment: Uses profile hidden Markov models to search and align single or multiple protein sequences and to search profile HMMs against target sequence databases.
• Multiple Segment Viterbi (MSV) algorithm: The MSV algorithm accelerates profile HMM computations by efficiently computing optimal sums of multiple ungapped local alignment segments.
• Sparse rescaling and Forward/Backward enhancements: Implements sparse rescaling to enhance the standard Forward/Backward algorithms, improving speed and sensitivity.
• Heuristic filters and accelerated algorithms: Applies heuristic filters and additional acceleration strategies to reduce the number of evaluations by full probabilistic models while maintaining sensitivity.
• Database and profile HMM library support: Supports searches against sequence databases and profile HMM libraries, including resources such as Pfam.
• Statistical scoring and compositional bias handling: Produces E-value estimates for hit significance while accounting for challenges such as compositional bias in certain protein families.

Scientific Applications:

• Protein homology detection: Detects homologous relationships between protein sequences through profile HMM similarity searches.
• Functional and structural annotation: Aids prediction of protein function and inference of structural features based on sequence homology.
• Phylogenetic reconstruction: Supports reconstruction of phylogenies by identifying homologous sequences for comparative analyses.
• Large-scale family identification and database annotation: Identifies and annotates protein family members across extensive databases such as Pfam.

Methodology:

HMMER3 employs profile hidden Markov models for sequence similarity searches and alignments, using the multiple segment Viterbi (MSV) heuristic, sparse rescaling of Forward/Backward algorithms, and heuristic filters to accelerate searches of sequences and profile HMMs against target databases while reporting E-value significance estimates.