PyHMMSearch：基于 PyHMMER 的高性能 HMM 蛋白质功能注释工具

# 创建一个名为PyHMMSearch的conda环境，并安装pyhmmsearch
conda create -n PyHMMSearch -c conda-forge -c bioconda pyhmmsearch
# conda install 直接安装
conda install -c bioconda pyhmmsearch
# pip install 安装
pip install pyhmmsearch
# 激活环境
conda activate PyHMMSearch  

## Pfam数据库为例
# 离线模式
mkdir -p db/Pfam && cd db/Pfam
# 使用wget下Pfam数据库Hmm文件
wget -c -nv -a download.log -b https://ftp.ebi.ac.uk/pub/databases/Pfam/current_release/Pfam-A.hmm.gz 
# 格式化数据库
# 方式 1：输入单个合并后的 HMM 数据库文件
serialize_hmm_models -d Pfam-A.hmm.gz -b PyHMMSearch.pkl.gz
# 方式 2：输入包含多个 HMM 文件的目录
serialize_hmm_models -d hmm_directory/ -b PyHMMSearch.pkl.gz
# 方式 3：输入记录多个 HMM 文件路径的列表文件
serialize_hmm_models -l hmms.list -b PyHMMSearch.pkl.gz

pyhmmsearch -h 
usage: pyhmmsearch -i <proteins.fasta> -o <output.tsv> -d 

    Running: pyhmmsearch v2025.10.23.post1 via Python v3.12.13 | /Path/softwara/miniforge3/envs/PyHMMSearch/bin/python3.12

options:
  -h, --help            show this help message and exit
  -v, --version         show program's version number and exit

I/O arguments:
  -i PROTEINS, --proteins PROTEINS
                        path/to/proteins.fasta. stdin does not stream and loads everything into memory. [Default: stdin]
  -o OUTPUT, --output OUTPUT
                        path/to/output.tsv [Default: stdout]
  --no_header           No header
  --tblout TBLOUT       path/to/output.tblout
  --domtblout DOMTBLOUT
                        path/to/output.domtblout

Utility arguments:
  -p N_JOBS, --n_jobs N_JOBS
                        Number of threads to use [Default: 1]

HMMSearch arguments:
  -s SCORES_CUTOFF, --scores_cutoff SCORES_CUTOFF
                        path/to/scores_cutoff.tsv[.gz] [id_hmm]<tab>[score_threshold], No header.
  -f {accession,name}, --hmm_marker_field {accession,name}
                        HMM reference type (accession, name) [Default: accession]
  -t {domain,full}, --score_type {domain,full}
                        {full, domain} [Default: full]
  -m {trusted,noise,e,gathering}, --threshold_method {trusted,noise,e,gathering}
                        Cutoff threshold method [Default:  e]
  -e EVALUE, --evalue EVALUE
                        E-value threshold [Default: 10.0]

Database arguments:
  -d HMM_DATABASE, --hmm_database HMM_DATABASE
                        path/to/database.hmm cannot be used with -b/-serialized_database.  Expects a (concatenated) HMM file and not a directory. You can build a database from a directory using `serialize_hmm_models.py`
  -b SERIALIZED_DATABASE, --serialized_database SERIALIZED_DATABASE
                        path/to/database.pkl cannot be used with -d/--database_directory.  Database should be pickled dictionary {name:hmm}

https://github.com/jolespin/pyhmmsearch

# 示例蛋白序列：https://ftp.ncbi.nlm.nih.gov/genomes/all/GCF/000/005/845/GCF_000005845.2_ASM584v2/GCF_000005845.2_ASM584v2_protein.faa.gz
# 去除序列ID后的注释信息
seqkit replace -p '\s.*' -r '' GCF_000005845.2_ASM584v2_protein.faa.gz -w 0 > ASM584v2_protein.faa
# 方式 1：输出 PyHMMSearch 整理后的全部命中结果
pyhmmsearch -i ASM584v2_protein.faa -o PyHMMSearch_Pfam_all.tsv -p 16 -b ./db/Pfam/PyHMMSearch.pkl.gz
# 方式 2：同时输出整理结果、序列级结果和结构域级结果
pyhmmsearch -i ASM584v2_protein.faa -o PyHMMSearch_Pfam_all.tsv --tblout PyHMMSearch_Pfam.tblout.tsv --domtblout PyHMMSearch_Pfam.domtblout.tsv -p 16 -b ./db/Pfam/PyHMMSearch.pkl.gz

# 查看结果
tree -h 
.
├── [1.3M]  ASM584v2_protein.faa
├── [881K]  GCF_000005845.2_ASM584v2_protein.faa.gz
├── [1.4M]  PyHMMSearch_Pfam_all.tsv
├── [6.7M]  PyHMMSearch_Pfam.domtblout
└── [3.2M]  PyHMMSearch_Pfam.tblout.tsv

# 查看结果
csvtk pretty -t PyHMMSearch_Pfam_all.tsv | head