HHM Format

​

Overview

​

Format Specification

​

Header Section

HHsearch 1.6
NAME  sp|Q5VUD6|FA69B_HUMAN Protein FAM69B OS=Homo sapiens GN=FAM69B PE=2 SV=3
FAM   
FILE  query
COM   hhmake -i /path/to/query.a3m 
DATE  Wed Jan  4 15:14:55 2012
LENG  431 match states, 431 columns in multiple alignment
FILT  149 out of 270 sequences passed filter (-id 90 -cov 0 -qid 0 -qsc -20.00 -diff 100)
NEFF  5.2

​

Header Fields

• HHsearch: Version identifier
• NAME: Protein name and description
• FAM: Family information (optional)
• FILE: Base filename
• COM: Command used to generate the HMM
• DATE: Creation timestamp
• LENG: Number of match states and alignment columns
• FILT: Filter statistics
• NEFF: Effective number of sequences (diversity measure)

​

Sequence Section

SEQ
>Consensus
xxxxxxxxxxxxxxxxxxxxxxrxxxxxxxxxxxxwxxxxxxsxxxyxxyssxselcrxxxcxxxiCxxYxxGxisGxlCxxLCxxxxlxxxxClxxxxx
>sp|Q5VUD6|FA69B_HUMAN Protein FAM69B
MRRLRRLAHLVLFCPFSKRLQGRLPGLRVRCIFLAWLGVFAGSWLVYVHYSSYSERCRGHVCQVVICDQYRKGIISGSVCQDLCELHMVEWRTCLSVAPG

• Consensus sequence (derived from alignment)
• Representative sequences from the MSA

​

NULL Model

NULL   3706	5728	4211	4064	4839	3729	4763	4308	4069	3323	5509	4640	4464	4937	4285	4423	3815	3783	6325	4665

​

HMM Header Line

HMM    A	C	D	E	F	G	H	I	K	L	M	N	P	Q	R	S	T	V	W	Y
       M->M	M->I	M->D	I->M	I->I	D->M	D->D	Neff	Neff_I	Neff_D

​

Match State Entries

M 1    2443	*	*	*	*	*	3455	*	*	*	1095	*	1962	*	*	*	*	*	*	*	1
       0	*	*	*	*	*	*	1695	0	0

• State type (M for match)
• Position number
• Emission probabilities for 20 amino acids (negative log probabilities, * = -infinity)
• Neff value

• Transition probabilities (M->M, M->I, M->D, I->M, I->I, D->M, D->D)
• Neff values for insert and delete states

​

Emission Probabilities

• Negative log-scale probabilities (base unclear, often -log2)
• * represents probability zero (negative infinity in log space)
• Smaller numbers indicate higher probability

​

Transition Probabilities

• M->M: Match to match
• M->I: Match to insert
• M->D: Match to delete
• I->M: Insert to match
• I->I: Insert to insert
• D->M: Delete to match
• D->D: Delete to delete

​

Example Match State

R 2    *	*	*	*	*	*	*	*	*	*	*	*	*	2443	293	*	*	*	*	*	2
       0	*	*	*	*	*	*	1695	0	0

• Position 2
• Amino acid R (Arginine)
• High probability for R (value 293) and Q (value 2443)
• Neff = 2 (low diversity at this position)

​

Creating HHM Files

​

From Alignment

hhmake -i alignment.a3m -o model.hhm

​

With Custom Name

hhmake -i alignment.a3m -o model.hhm -name MyProtein

​

With Pseudocounts

hhmake -i alignment.a3m -o model.hhm -pc_hhm_contxt_a 0.9

​

Using HHM Files

​

As Query

hhsearch -i query.hhm -d database

​

As Template

hhalign -i query.a3m -t template.hhm

​

Binary Format

• More compact storage
• Faster parsing
• Generated automatically by some tools

​

Best Practices

​

Building Quality HMMs

1. Diverse alignments: Use sequences with varied identity (filter with hhfilter)
2. Sufficient sequences: Aim for Neff > 4 for good coverage
3. Quality filtering: Remove low-coverage sequences
4. Pseudocounts: Use context-specific pseudocounts for better profiles

​

Neff Values

• Neff < 2: Low diversity, may need more sequences
• Neff 4-8: Good diversity for most purposes
• Neff > 10: High diversity, excellent for sensitive searches

​

File Size Considerations

• Larger than input alignments (due to probability matrices)
• Smaller than storing full alignments
• Text format: ~5-10 KB per 100 residues
• Binary format: ~50% smaller

​

See Also

• A3M Format - Input alignment format
• HHR Format - Results output format
• hhmake - Build HMMs from alignments
• hhsearch - Search with HMMs