Skip to main content
HHconsensus calculates the consensus sequence for a multiple sequence alignment. It’s useful for generating representative sequences from MSAs, creating database entries, or analyzing alignment conservation patterns.

​
Overview

HHconsensus reads an alignment in A2M, A3M, or FASTA format and generates a consensus sequence based on the most frequent amino acid at each position, weighted by sequence similarity. It can output the consensus alone or as part of the full alignment.

​
Key Features

  • Consensus generation: Creates representative sequence from MSA
  • Multiple output formats: FASTA, A2M, A3M formats
  • Filtering support: Apply filters before consensus calculation
  • Flexible output: Consensus only or full alignment with consensus

​
When to Use HHconsensus

Use HHconsensus when you need to:
  • Create representative sequences: Generate consensus for database entries
  • Analyze conservation: Identify conserved positions in alignments
  • Reduce MSA to single sequence: Simplify analysis or visualization
  • Build consensus databases: Create searchable consensus sequence sets
  • Quality control: Check alignment quality via consensus inspection
The consensus sequence is automatically calculated during HMM building in hhmake, but HHconsensus gives you direct access to it.

​
Basic Usage

1

Generate consensus sequence

Extract consensus in FASTA format:
hhconsensus -i alignment.a3m -s consensus.seq
2

Output alignment with consensus

Include consensus in the output alignment:
hhconsensus -i alignment.a3m -o output.a3m
3

Pipeline usage

Use stdin/stdout for pipeline integration:
hhconsensus -i stdin -s stdout < alignment.a3m > consensus.seq

​
Common Use Cases

​
Extract Consensus Sequence

Generate just the consensus sequence: 
hhconsensus -i alignment.a3m -s consensus.fas
Output format: 
>alignment consensus
MSTVKGYRILLAGAIDSFSLTESDKPTYRLVGPSGCSGKTTLLNAIAGESPTSGKVTLSGG

​
Generate Consensus with Filtering

Filter alignment before calculating consensus: 
hhconsensus -i alignment.a3m -s consensus.fas \
  -id 90 \   # Remove sequences >90% identical  
  -cov 50    # Require 50% coverage

​
Create Consensus Database

Build a database of consensus sequences: 
#!/bin/bash
for ali in alignments/*.a3m; do
  base=$(basename "$ali" .a3m)
  hhconsensus -i "$ali" -s "consensus/${base}_consensus.fas"
done

# Concatenate all consensus sequences
cat consensus/*.fas > all_consensus.fas

​
Output Alignment in Different Formats

Generate alignment with consensus in various formats: 
# A3M format (with inserts as lowercase)
hhconsensus -i alignment.a3m -oa3m output.a3m

# A2M format (with inserts as lowercase)
hhconsensus -i alignment.a3m -oa2m output.a2m

# FASTA format (inserts removed)
hhconsensus -i alignment.a3m -ofas output.fas

​
Batch Processing

Process multiple alignments: 
for file in *.a3m; do
  hhconsensus -i "$file" -s "${file%.a3m}_consensus.seq"
done

​
Key Parameters

  • -i <file> - Input alignment (A2M, A3M, FASTA) or HMM
  • -s <file> - Output consensus sequence in FASTA (default: <infile>.seq)
  • -o <file> - Output alignment with consensus in A3M format
  • -oa3m <file> - Same as -o (explicit A3M format)
  • -oa2m <file> - Output alignment with consensus in A2M format
  • -ofas <file> - Output alignment with consensus in FASTA format
  • -v <int> - Verbose mode (0=silent, 1=warnings, 2=verbose)
  • -id [0,100] - Maximum pairwise sequence identity % (default: 100)
  • -diff [0,inf] - Filter for diversity, keeping at least this many sequences per 50-column block (default: 0)
  • -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)
  • -M a2m - A2M/A3M format (default): upper=match, lower=insert
  • -M first - FASTA: first sequence defines match states
  • -M [0,100] - FASTA: columns with <X% gaps are match states
  • -maxseq <int> - Maximum number of input sequences (default: 65535)
  • -maxres <int> - Maximum number of HMM columns (default: 20000)

​
How Consensus is Calculated

The consensus sequence is determined by:
  1. Sequence weighting: Similar sequences are down-weighted
  2. Position-specific frequencies: Count amino acids at each position
  3. Most frequent residue: Select the most common amino acid
  4. Tie breaking: Use a fixed priority order if frequencies are equal
Position:     1234567890
Sequence 1:   MSTIKGYRIL
Sequence 2:   MSTIKGYRIL  (identical, down-weighted)
Sequence 3:   MATIVGYRVL
Sequence 4:   MSAIKGFRIL
Sequence 5:   LAATVKGYKIL (different, higher weight)

​
Output Formats

>alignment_name consensus
MSTVKGYRILLAGAIDSFSLTESDKPTYRLVGPSGCSGKTTLLNAIAGESPTSGKVTLSGG

​
Tips and Best Practices

Filter before consensus: For better consensus quality, filter the alignment first:
hhconsensus -i alignment.a3m -s consensus.fas -id 90 -cov 50
This removes redundant and low-quality sequences.
The consensus is only as good as the input alignment. Poor alignments or highly divergent sequences will produce poor consensus sequences.
Check consensus quality: View the consensus alongside the alignment to verify it represents conserved positions:
hhconsensus -i alignment.a3m -ofas alignment_with_cons.fas
# View in alignment viewer

​
Advanced Use Cases

​
Named Consensus Sequences

Specify a custom name for the consensus: 
hhconsensus -i alignment.a3m -s consensus.fas -name "MyProtein_consensus"

​
Consensus with Strict Quality Control

Generate consensus from only high-quality sequences: 
hhconsensus -i alignment.a3m -s high_quality_consensus.fas \
  -id 70 \      # Remove highly similar sequences
  -cov 80 \     # Require 80% coverage
  -qid 30 \     # At least 30% identity with master
  -qsc 5.0      # Minimum score per column

​
Build NR-style Consensus Database

Create a non-redundant database from alignments: 
#!/bin/bash
# Extract consensus from each alignment
for ali in protein_families/*.a3m; do
  family=$(basename "$ali" .a3m)
  hhconsensus -i "$ali" -s temp.fas -name "${family}_consensus" -id 90
  cat temp.fas >> nr_consensus.fas
done
rm temp.fas

# The result is a FASTA database of consensus sequences

​
Integration with Structure Prediction

Generate consensus for AlphaFold input: 
# Build MSA with HHblits
hhblits -i query.fas -d uniclust30 -oa3m msa.a3m

# Extract consensus
hhconsensus -i msa.a3m -s consensus.fas

# Use consensus for validation or as alternative input

​
Comparison with Master Sequence

The consensus differs from the master sequence:
AspectMaster SequenceConsensus Sequence
DefinitionFirst sequence in alignmentMost frequent residue per position
CoverageMay have gapsAlways complete (no gaps)
RepresentsOne real sequenceAverage of all sequences
Use caseReference sequenceRepresentative sequence

​
Pipeline Integration

​
Workflow Example

#!/bin/bash
# Complete pipeline: search → filter → consensus

# 1. Build MSA
hhblits -i query.fas -d uniclust30 -oa3m raw_msa.a3m -n 3

# 2. Filter MSA  
hhfilter -i raw_msa.a3m -o filtered_msa.a3m -id 90 -cov 50

# 3. Generate consensus
hhconsensus -i filtered_msa.a3m -s consensus.fas

# 4. Generate HMM
hhmake -i filtered_msa.a3m -o profile.hhm -add_cons

​
Batch Consensus Generation

# Process all MSAs in parallel
find alignments/ -name "*.a3m" | \
  parallel -j 8 \
  'hhconsensus -i {} -s {.}_consensus.fas -id 90 -cov 50'

​
Troubleshooting

If the consensus sequence has gaps (dashes):
  • This shouldn’t happen - consensus always has a residue at each match state
  • Check if input alignment is properly formatted
  • Ensure you’re using A3M/A2M format correctly
If the consensus seems wrong:
  • View alignment to check sequence quality
  • Apply filters to remove outliers
  • Check if similar sequences are dominating (use -id filter)
  • Remember: consensus is frequency-based, not conservation-based
If no consensus is generated:
  • Check input file format
  • Verify alignment has at least one sequence
  • Look for error messages with -v 2
  • Ensure output path is writable
  • hhmake - Build HMMs (includes consensus in HMM file)
  • hhfilter - Filter alignments before consensus
  • hhblits - Build MSAs for consensus generation

​
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