/hgt_detection

Primary LanguagePython

HGT detection pipeline v.1

Usage

Add your own (sub)-taxonomy

Insert tutorial

Installation

Dependencies

  • snakemake - conda
  • hgtector2 - pip
  • taxonkit - conda
  • blast suite - conda
  • mafft - conda
  • proteinortho - conda
  • famsa - conda
  • seqkit - conda
  • trimal - conda
  • fasttree - conda
  • FastRoot - pip
  • tidyverse R package - CRAN
  • phytools -CRAN
  • ggtree - Bioconductor
  • ggtreeExtra - Bioconductor
  • Biostrings - Bioconductor
  • ComplexHeatmap - Bioconductor

You can use conda to create this environment, ideally use mamba:

conda create -c conda-forge -c bioconda -n hgt snakemake
conda activate hgt

# install hgtector
conda install -c conda-forge pyyaml pandas matplotlib scikit-learn bioconda::diamond
pip install git+https://github.com/qiyunlab/HGTector.git

# install other dependencies
conda install -c bioconda mafft trimal fasttree taxonkit proteinortho famsa seqkit

# install R if you need
conda install -c conda-forge r-base
conda install -c conda-forge -c bioconda r-tidyverse r-optparse r-magic r-phytools bioconductor-ggtree bioconductor-ggtreeextra bioconductor-complexheatmap 
bioconductor-biostrings

# install fastroot
python3 -m pip install FastRoot
fr=$(which FastRoot.py)
dos2unix $fr

TODOs

  • mask proteins and then do dataframe with protein length, %masked proportion to eventually filter downstream results

  • add link to deps

  • what to do if less than 2 seqs in plot_Tree?

  • more elegant way to get sequences to do the tree

  • add QC plots

  • what to do with foldseek?

  • matreex plots?

  • mark HGT tree by topological rule

  • get df of tree with all different properties so user can filter

  • when deleting hits from the same species you may remove paralogs that are not identified by proteinortho but still may be close in the tree

  • annotate domains and protein functions

  • maybe one strategy would be to compute big trees in a fast way then reduce taxonomic redundancy and compute a more accurate tree

  • remove fastroot dependency

  • option to define "close" outgroup and compute other indexes

  • identify potential donors from blast results (recurring weird taxons)

  • one day 3 modalities

    • --interdomain: easy mode for proka2euka lgt
    • --donors: user-defined putative donors
    • --reconcile: with small curated database and species tree run reconciliation