Meta-MEME

Meta-MEME constructs motif-based Hidden Markov Models (HMMs) from MEME-derived motifs to model families of related biological sequences by focusing on conserved regions, enabling effective training and sensitive database searches with limited sequence data.

Key Features:

• Motif-based HMM construction: Creates smaller HMMs derived from motif models produced by MEME.
• MEME with EM algorithm: Uses MEME's Expectation-Maximization (EM) algorithm to identify and characterize conserved sequence motifs.
• Conserved-region focus: Concentrates on conserved regions to reduce the number of HMM parameters.
• Training on limited data: Enables effective training of HMMs with smaller datasets and few known family members.
• Improved search performance: Improves sensitivity and selectivity of database searches compared to standard linear HMMs.
• Empirical validation: Demonstrated superior recognition performance on short chain alcohol dehydrogenases and 4Fe-4S ferredoxins.

Scientific Applications:

• Modeling sequence families: Modeling families of related biological sequences using HMMs when training data are limited.
• Sensitive database searches: Conducting sensitive and selective database searches for motif-containing sequence families.
• Protein family recognition: Recognition and classification of protein families such as short chain alcohol dehydrogenases and 4Fe-4S ferredoxins.
• Genomics and proteomics analyses: Detecting conserved motifs and assigning family membership in genomics and proteomics studies.

Methodology:

Generate motif models with MEME using the Expectation-Maximization (EM) algorithm; construct smaller HMMs from those motif models; reduce HMM parameters by focusing on conserved regions; train models on the reduced parameter sets for database searches.