RefineHMM

RefineHMM: Automated Hidden Markov Model Refinement for Protein Families

RefineHMM refines hidden Markov models (HMMs) through iterative database searches and incremental adaptation of seed sets to generate stable, evolutionarily consistent models for protein families.

Key Features:

• Automated HMM Refinement Algorithm: Constructs confident seed sets from detected sequence relationships and iteratively updates HMMs to produce stable, high-quality models without manual curation.
• Modeling of MDR Superfamily: Generated HMMs for 86 families and 34 subfamilies within the Medium-chain Dehydrogenases/Reductases (MDR) superfamily, enabling automated sequence annotation.
• Zn²⁺-Dependent Functional Classification: Distinguished MDR forms containing two Zn²⁺ ions, typically dehydrogenases, from Zn²⁺-lacking forms, typically reductases, including differences between bacterial and eukaryotic MDRs.

Scientific Applications:

• Protein Superfamily Resolution: Produces HMMs corresponding to evolutionary entities, resolves overlapping models, and improves granularity and automation of protein family annotation from sequence data.

Methodology:

Refinement is performed by iterative database searches that expand and update seed sets based on confidently inferred sequence relationships. Updated seed sets are used to retrain HMMs until stable models are obtained, enabling scalable analysis of large protein superfamilies.