Skip to main content
HHmake converts a multiple sequence alignment (MSA) into a Hidden Markov Model (HMM) profile in HH-suite format. It’s an essential preprocessing tool for preparing custom HMM databases or converting between different HMM formats.

​
Overview

HHmake reads an alignment in A2M, A3M, or FASTA format and generates an HMM file (.hhm) that can be used with other HH-suite tools. It can also convert between HMMER format (.hmm) and HHsearch format (.hhm).

​
Key Features

  • MSA to HMM conversion: Build HMM profiles from alignments
  • Format conversion: Convert between HMMER and HHsearch formats
  • Filtering: Apply sequence filters during HMM construction
  • Pseudocounts: Add context-specific or independent pseudocounts

​
When to Use HHmake

Use HHmake when you need to:
  • Build custom HMM databases: Convert alignments to searchable HMMs
  • Preprocess alignments: Generate HMMs for use with hhsearch or hhalign
  • Convert HMM formats: Transform HMMER HMMs to HH-suite format
  • Control HMM building: Fine-tune pseudocounts and filtering parameters
Most users won’t need HHmake directly, as tools like hhblits and hhsearch build HMMs automatically. Use HHmake when you need custom HMM databases or format conversion.

​
Basic Usage

1

Convert alignment to HMM

Basic conversion from A3M to HHM format:
hhmake -i alignment.a3m -o alignment.hhm
2

Build HMM and apply filters

Filter sequences during HMM construction:
hhmake -i alignment.a3m -o alignment.hhm -id 90 -cov 50
3

Read from stdin, write to stdout

Use in pipelines:
cat alignment.a3m | hhmake -i stdin -o stdout > alignment.hhm

​
Common Use Cases

​
Build HMM from FASTA Alignment

Convert a FASTA multiple sequence alignment: 
hhmake -i sequences.fas -o sequences.hhm

​
Create Filtered HMM

Build HMM while filtering redundant sequences: 
hhmake -i alignment.a3m -o filtered.hhm \
  -id 90 \      # Max 90% pairwise identity
  -cov 75 \     # Min 75% coverage with master
  -qid 30       # Min 30% identity with master

​
Build HMM Database

Create multiple HMMs for a database: 
for file in *.a3m; do
  hhmake -i "$file" -o "${file%.a3m}.hhm"
done

​
Add Consensus Sequence

Include consensus as master sequence: 
hhmake -i alignment.a3m -o alignment.hhm -add_cons

​
Key Parameters

  • -i <file> - Query alignment (A2M, A3M, FASTA) or HMM file
  • -o <file> - HMM output file (default: <infile>.hhm)
  • -a <file> - Append to existing HMM file instead of overwriting
  • -name <name> - Use this name for HMM (default: first sequence name)
  • -v <int> - Verbose mode (0=no output, 1=warnings, 2=verbose)
  • -id [0,100] - Maximum pairwise sequence identity % (default: 90)
  • -diff [0,inf] - Filter for diversity, keeping Ndiff sequences per 50-column block (default: 100)
  • -cov [0,100] - Minimum coverage with query % (default: 0)
  • -qid [0,100] - Minimum sequence identity with query % (default: 0)
  • -qsc [-inf,100] - Minimum score per column with query (default: -20.0)
  • -neff [1,inf] - Target diversity (effective number of sequences)
  • -M a2m - Use A2M/A3M format (default)
  • -M first - Use FASTA format, columns with residue in 1st sequence are match states
  • -M [0,100] - Use FASTA format, columns with <X% gaps are match states
  • -add_cons - Generate consensus sequence as master sequence
  • -seq <int> - Maximum number of sequences to display in HMM (default: 10)
  • -maxseq <int> - Maximum number of input sequences (default: 65535)
  • -maxres <int> - Maximum number of HMM columns (default: 20000)

​
Output Format

The HHM file is a text-based format containing: 
HHsearch 1.5
NAME  protein_name
FAM   protein_family  
FILE  alignment.a3m
LENG  250 match states, 500 columns in alignment
FILT  150 out of 200 sequences passed filter
NEFF  8.5
The HHM format stores:
  • Emission probabilities for each match state
  • Transition probabilities between states
  • Effective number of sequences (Neff) per column
  • Secondary structure predictions (if available)

​
Tips and Best Practices

Filtering redundancy: Use -id 90 to remove very similar sequences. This speeds up searches while maintaining sensitivity.
Overly aggressive filtering (e.g., -id 50 -cov 90) can remove too many sequences and reduce the profile quality. Balance diversity with information content.
Match state assignment:
  • Use -M a2m (default) when your alignment already has match/insert states defined
  • Use -M first to make the first sequence define match states
  • Use -M 50 to make columns with <50% gaps be match states

​
Advanced Options

​
Context-Specific Pseudocounts

Enable context-specific pseudocounts for better HMMs: 
hhmake -i alignment.a3m -o alignment.hhm \
  -pc_hhm_contxt_mode 2 \  # Diversity-dependent mode
  -pc_hhm_contxt_a 0.9 \   # Overall admixture (0-1)
  -pc_hhm_contxt_b 4.0 \   # Neff threshold
  -pc_hhm_contxt_c 1.0     # Extinction exponent

​
No Pseudocounts (for raw counts)

Build HMM without any pseudocounts: 
hhmake -i alignment.a3m -o raw.hhm \
  -pc_hhm_contxt_mode 0 \  # No pseudocounts
  -nocontxt                # Disable context-specific pseudocounts

​
Custom Sequence Weighting

Control how sequences are weighted in the profile: 
hhmake -i alignment.a3m -o alignment.hhm -wg
This uses global sequence weighting instead of local weighting.

​
Build Database from Directory

Process all alignments in a directory: 
#!/bin/bash
for ali in alignments/*.a3m; do
  base=$(basename "$ali" .a3m)
  hhmake -i "$ali" -o "hmms/${base}.hhm" -id 90 -cov 50
done

# Create database index
ffindex_build -s hmms.ffdata hmms.ffindex hmms/

​
Format Conversion

​
HMMER to HH-suite Format

Convert HMMER3 HMM to HH-suite format: 
hhmake -i hmmer_profile.hmm -o hhsuite_profile.hhm

​
A3M to A2M Format

While HHmake primarily creates HMMs, you can use it in a pipeline: 
# Generate HMM then use with other tools
hhmake -i alignment.a3m -o temp.hhm
hhsearch -i temp.hhm -d database -o results.hhr

​
Understanding Match State Assignment

The -M parameter controls how columns are designated as match vs. insert states: 
Input:  MSTPQRLLAGAIDSFSLTESDKPTYRlvgpsgcsGKTTLLNAIAG
        Upper = Match, lower = Insert
Result: Match states at uppercase positions

​
Troubleshooting

If filtering removes too many sequences:
  • Relax -id threshold (increase value)
  • Lower -cov and -qid requirements
  • Check input alignment quality
If you hit the maximum HMM size:
  • Use -maxres to increase the limit
  • Check if your alignment has excessive columns
  • Consider trimming the alignment first
If the resulting HMM performs poorly:
  • Ensure input alignment is high quality
  • Try context-specific pseudocounts
  • Adjust filtering to keep more diverse sequences

​
References

Steinegger M, Meier M, Mirdita M, Vöhringer H, Haunsberger SJ, Söding J (2019). HH-suite3 for fast remote homology detection and deep protein annotation. BMC Bioinformatics, 473. doi: 10.1186/s12859-019-3019-7

Build docs developers (and LLMs) love

Get started for freeTalk to us