minhyukpark / hmmerge

HMMerge

---

Do you want to merge multiple profile HMMs into one?

How to Run

---

There are 2 ways to run HMMerge depending on the input data available. Regardless of which way is used, the output file will be available at <Output directory>/HMMerge.aligned.fasta.

1. Most users fall into this category. Required Inputs:

• Backbone alignment in FASTA format
• Backbone tree in NEWICK format
• Sequences to align in FASTA format

python <git root>/decompose_fasta_version.py --input-tree <Backbone tree> --sequence-file <Backbone alignment> --output-prefix <Output folder for decomposed alignments>/input_ --maximum-size <Decompose size>

pushd <Folder with decomposed alignments>
for f in *.fasta
do
    trimal -in ${f} -out ${f} -noallgaps
done
popd

python <git root>/main.py --input-dir <Folder with decomposed alignments> --backbone-alignment <Backbone alignment> --query-sequence-file <Query sequences> --output-prefix <Output directory> --num-processes <Num cpus> --model {dna|amino}

2. No backbone tree, only backbone alignment and query sequences Required Inputs:

• Backbone alignment in FASTA format
• Sequences to align in FASTA format
• Directory containing decomposed backbone alignment in FASTA format

Note: Make sure that the directory containing decomposed backbone alignment follows these two rules. a. The files need to follow the naming format input_<integer starting from 0>.fasta b. The fasta files must not have columns that are all gaps. If so, use trimal or other equivalent utilities to remove columns with all gaps.

python <git root>/main.py --input-dir <Folder with decomposed alignments> --backbone-alignment <Backbone alignment> --query-sequence-file <Query sequences> --output-prefix <Output prefix> --num-processes <num cpus> --model {dna|amino}

Requirements

---

• biopython
• click
• dendropy (if using a backbone tree)
• numpy
• pyhmmer-sepp
• pytest (for testing)
• scipy

Testing

---

git clone https://github.com/MinhyukPark/HMMerge.git
cd HMMerge
pytest test.py

What HMMerge does

---

HMMerge, given a backbone alignment, its decomposition, and a set of query sequences, builds an HMM on each of the decomposed subset backbone alignments, assigns probability weights on each HMM based on the likelihood of havingt emitted the query sequence, merges all the HMMs based on the weights, and aligns the query sequence using the merged HMM. Individual query sequences are aligned independently onto the backbone alignment and merged at the end using transitivity to the backbone alignment.