Halogenase

This repository provides the profile hidden Markov model (pHMM) for flavin-dependent tryptophan halogenase (Trp-FDH) and Python script for filtering result of pHMM search and leave-one-out cross-validation.

Pre-installed programs

• Python version 2.7.12 or higher with argparse library
• HMMER3.0 package version 3.1b2 or higher
• Clustal Omega version 1.2.4 or higher

Filtering result of pHMM search

• pHMM search

  • To perform the pHMM search, hmmscan in HMMER3.0 package is used with "--domtblout" option.
  • A query should be included as protein sequences in the FASTA format file.

USAGE:
$ hmmscan --domtblout output_file_name.domtblout pHMM/Trp_FDH.hmm query.fasta

• Filtering .domtblout file

  • To filter the .domtblout file, code/domtblout_filtering.py is used.
  • According to the threshold for e-value and pHMM model coverage, the search result is filtered.

USAGE:
$ python domtblout_filtering.py -i example/example_input.domtblout -o example/example_output.domtblout -e e-value -c model_coverage

Leave-one-out cross-validation (LOOCV)

• LOOCV

  • To perform the LOOCV, code/leave-one-out_hmmscan.py is used.
  • The input file of this code is in pHMM/Trp-FDH.fasta.
  • Training data and test data are generated as many as the number of sequences in the input file, respectively.
  • To build pHMMs of training data, multiple sequence alignment (MSA) is performed by using Clustal Omega (v1.2.4).
  • Using the results of MSA (.sto file), pHMM models were constructed by using hmmbuild in HMMER3.0 package.
  • The pHMM search with the pHMM or training data and test set is executed.
  • The pHMM search results are summarized in terms of evalue (avg. std., and median of evalue).

USAGE:
$ python leave-one-out_hmmscan.py --fasta pHMM/Trp-FDH.fasta