HHPred

HHPred is a web-based server and software package that enables sensitive detection of distant protein homology by comparing hidden Markov models (HMMs). It is part of the HH-suite and is widely used for functional annotation, domain identification, and structure-guided analysis. HHPred operates by building a query HMM from an input sequence or multiple sequence alignment and then searching a database of HMMs derived from known protein families, domains, or structural templates.

The method relies on HMM-HMM comparison. The input sequence or alignment is used to generate a query HMM, typically through iterative profile construction. This query is then compared against a database of HMMs using the HHsearch algorithm, producing hits with statistical measures such as probability, E-value, and score. The output includes the aligned regions, domain boundaries, and links to potential structural templates, enabling researchers to infer function or make structural inferences for the query protein.

Databases and usage options are flexible. HHPred supports multiple curated HMM databases, including those derived from Pfam, PDB-derived structures, and other domain or family resources, and can also accommodate custom databases. Users can submit single sequences or alignments, review ranked hits, and examine alignments in detail. The tool is commonly used to assign domains, propose fold-level classifications, and guide downstream modeling or experimental design, contributing to a broader understanding of protein function and evolution.