FlowerPower

FlowerPower identifies global homologs and clusters protein sequences by shared domain architecture to support structural phylogenomic analysis.

Key Features:

• Iterative Clustering Approach: Employs an iterative strategy similar to PSI-BLAST that emphasizes global rather than local homology detection.
• Subfamily Identification via SCI-PHY: Uses the SCI-PHY algorithm to identify subfamilies within protein families.
• Subfamily HMMs for Homolog Alignment: Employs subfamily-specific hidden Markov models to select and align homologous sequences for clustering based on domain architecture.
• Performance Comparison: Demonstrated to outperform BLAST, PSI-BLAST, and the UCSC SAM-Target 2K in discriminating proteins that share the same domain architecture from those with different overall structures.

Scientific Applications:

• Structural phylogenomics: Clusters sequences by shared domain architecture to support phylogenomic analyses of protein families.
• Annotation error mitigation: Reduces systematic annotation-transfer errors arising from top-hit homology searches by prioritizing global architecture consistency.
• Function prediction for multi-domain proteins: Improves functional inference within an evolutionary context for proteins with multiple domains.

Methodology:

Uses SCI-PHY to identify subfamilies and then employs subfamily-specific HMMs to select, align, and cluster homologous sequences according to consistent domain architecture.