HH-suite

HH-suite performs sensitive sequence similarity searches and protein fold recognition by pairwise alignment of profile Hidden Markov Models (HMMs) to detect remote homologs and support protein structure and function prediction.

Key Features:

• Pairwise profile HMM alignment: Performs pairwise alignment of profile Hidden Markov Models (HMMs) constructed from multiple sequence alignments of homologous proteins.
• Profile HMM construction: Builds profile HMMs from multiple sequence alignments of homologous proteins to capture position-specific residue conservation and variation.
• SIMD-vectorized Viterbi algorithm: Implements a single-instruction multiple-data (SIMD) vectorized version of the Viterbi algorithm for profile HMM alignment.
• Performance improvements: Accelerates HHsearch by ~4× and HHblits by ~2×, with HHblits3 outperforming PSI-BLAST by ~10× and HMMER3 by ~20×.
• Parallelization: Supports parallel execution using OpenMP and the Message Passing Interface (MPI) across multiple cores and cluster servers.
• High-throughput query handling: Enables simultaneous processing of numerous query profile HMMs for large-scale analyses.
• High sensitivity: Provides highly sensitive detection of remote homologies between protein sequences.

Scientific Applications:

• Remote homology detection: Detects remote homologs between protein sequences to reveal distant evolutionary relationships.
• Protein fold recognition: Identifies protein folds and supports protein structure prediction via HMM-HMM comparisons.
• Functional annotation: Facilitates functional annotation of proteins through sensitive homology detection.
• High-throughput genomics and metagenomics: Applied in large-scale genomics and metagenomics projects for rapid sequence analysis of extensive datasets.
• Evolutionary studies: Supports analysis of evolutionary relationships among homologous proteins.

Methodology:

Constructs profile HMMs from multiple sequence alignments and performs pairwise HMM–HMM alignment using a SIMD-vectorized Viterbi algorithm, with parallel execution supported via OpenMP and MPI.