neherlab / HIVEVO_reversion

Repository for the analysis and figures of Valentin Druelle's paper on HIV-1 reversion

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

HIVEVO_reversion

Repository for the analysis and figures of the paper "Reversion to consensus are positively selected in HIV-1 and bias substitution rate estimates" by Valentin Druelle and Richard Neher. This repository contains a compressed version of the intermediate data used to create the figures along with the scripts needed for that. It also contains the scripts to generate the intermediate data yourself if needed. The raw files being too heavy for a github repository, you will have to download the full dataset yourself.

Installing conda environment

The codes in this repo needs a couple of over package to run. These dependencies can be installed using conda and the provided conda_env.yaml file (you will need to install conda first). We recommend installing the conda environment using mamba, otherwise it might take a very long time to do so. Install mamba with conda activate base; conda install mamba; mamba update --all. Then install the environment using mamba env create -f conda_env.yml. This will create a conda environment named HIV_reversion. You need to activate this environment using conda activate HIV_reversion before using the scripts in this repository.

Figure plots

All the figures are plotted using the Paper_figures.py script. You will have to uncompress the intermediate data as explained in the following section.

Working with the intermediate files

This section is intended for people who wish to work with the intermediate data used for the paper. Due to size issues, the intermediate data is saved in a compressed format in data.tar.xz/. One can use the unpack_data.sh script to uncompress it, which will generate the data/ folder and touch the files with snakemake. This can be done with the command bash unpack_data.sh on Linux distributions.

This data contains:

  • The raw and subsampled sequences for the BH analysis for the different regions and their metadata in data/BH/raw/
  • The alignments for these sequences in data/BH/alignments/to_HXB2/
  • The trees and TreeTime files in data/BH/intermediate_files/
  • The reference HXB2 sequence in data/BH/reference/
  • The bootstrap dictionnaries for the error rate estimate in data/BH/bootstraps/
  • The files necessary for visualisation with nextstrain in data/BH/visualisation. Can be used with the nextstrain view data/BH/visualisation command
  • The dictionnaries used for the WH analysis in data/WH/
  • The MSAs generated for the modeling part in data/modeling/generated_MSA/
  • The trees generated for the modeling part in data/modeling/generate_trees/

Working with the raw files

This section is intended for people that wish to regenerate the intermediate files. This is computationally intensive and will take several hours on a regular laptop. One can remove the intermediate files provided using snakemake clean --cores 1 and regenerate them as explained in the following sections.

Generate between host data

For the between host analysis, make sure the raw data is in the data/BH/raw folder, then use snakemake to execute the rule all. This can be done with the command snakemake all --cores 8 to run it on 8 cores for example. This will compute a bunch of files for the 3 HIV-1 genes studied, which can take a lot of time. Details on how to do this on the cluster in section "Sidenote: use on the cluster".

Generate within host and modeling data

The WH intermediate data is generated by using the HIVevo_access repo: https://github.com/neherlab/HIVEVO_access. It also requires the files from the between host analysis, so one has to generate them first (previous section). Generate the intermediate data by using python scripts/generate_data.py make-data. Note that to run properly, one needs to set the correct paths to the HIVevo access folder in the scripts/filenames.py file. It will generate all the intermediate data needed for the within host analysis. One can use python scripts/generate_data.py clean-data to remove this part of the intermediate data only.

Sidenote : use on the cluster

Generating the intermediate files (both for the BH and WH analysis) is computationnaly intensive. The whole thing will take several hours if run on a laptop. The generation of the BH data can be speed up using a cluster to run the snakemake rules. Command to launch the jobs on the cluster: snakemake all --jobs=16 --cluster "sbatch --time={cluster.time} --mem={cluster.mem} --cpus-per-task={cluster.n} --qos={cluster.qos}" --jobscript snake_submit.sh --cluster-config cluster.json --jobname "{rulename}_{jobid}"

About

Repository for the analysis and figures of Valentin Druelle's paper on HIV-1 reversion


Languages

Language:Python 99.6%Language:Shell 0.4%