Installation

Manual installation

modPhred is written in Python3 and should work in most UNIX systems.

Make sure you install all programs listed below, before runnning the pipeline.

  • first the repository

mkdir ~/src
cd src
git clone https://github.com/novoalab/modPhred

  • from conda (use miniconda3!)

conda create -y -n modPhred python=3.7
conda activate modPhred
conda install minimap2 samtools hdf5 wget

# create & activate new virtual environment
mkdir -p ~/src/venv
python3 -m venv ~/src/venv/modPhred
source ~/src/venv/modPhred/bin/activate

pip install h5py matplotlib pysam pandas seaborn sklearn mappy pebble numba

pip install pyguppyclient==0.0.6

Once you have all dependencies installed, we recommend to try running it with test dataset. It’ll be much easier to troubleshoot all potential issues this way.

Which pyguppyclient version should I install?

If you plan to perform live basecalling, you’ll need to install right version of pyguppyclient. Guppy API was changing across version and unfortunately the newer versions are not back-compatible. Therefore, you’ll need to install pyguppyclient version matching the version of guppy basecaller.

Guppy version

pyguppyclient version

>= 5.0

0.1.0

>= 4.4 & <5.0

0.0.9

>= 4.0 & <4.4

0.0.7a1

>= 3.4 & <4.0

0.0.6

< 3.4

not supported!

For example, if you intend to use guppy 4.0.15, you’ll need to install pyguppyclient v0.0.7a1 as follows:

pip install pyguppyclient==0.0.7a1

Note, only one version of pyguppyclient can be installed in your system. If you wish to use more than one version, you can install them using virtual environments as follows:

python3 -m venv ~/src/venv/pyguppyclient006
source ~/src/venv/pyguppyclient006/bin/activate
pip install pyguppyclient==0.0.6 pysam pandas seaborn

And whenever you wish to switch to this version, just execute:

source ~/src/venv/pyguppyclient006/bin/activate

Once you are finish with computation eihert close the terminal window or execute deactivate.

Docker image

We maintain docker image for below versions of guppy:

  • 3.6.1 (with pyguppyclient v0.0.6)

  • 5.0.11 (with pyguppyclient v0.1.0) - you’ll need CUDA v11.1 and most recent GPU drivers installed!

If you want to use it, make sure you have Docker, GPU drivers, CUDA and nvidia-docker installed. The easiest may be to follow nvidia-docker installation tutorial.

In order to execute test example, all you need to do is to adjust the version of guppy in the below command:

cd test
acc=PRJEB22772
docker run --gpus all -u $UID:$GID -v `pwd`:/data lpryszcz/modphred-3.6.1 \
  /opt/modPhred/run -f /data/ref/ECOLI.fa -o /data/modPhred/$acc \
  -i /data/$acc/{MARC_ZFscreens_R9.4_1D-Ecoli-run_FAF05145,MARC_ZFscreens_R9.4_2D-Ecoli-run_FAF05711} \
  -t4 --host /usr/bin/guppy_basecall_server

As you can see, the above command got a bit complicated. This is because:

  • we need to enable GPU

  • define user & group (otherwise all output files will be owned by root)

  • bind local directory within container

  • and define all input folders (because autocompletion doesn’t work inside the container)

Note, dam-dcm-cpg model has been replaced in guppy v4.5 by the new 5mc model, so additionally we’ll need to specify a model name in the recent versions of guppy.

docker run --gpus all -u $UID:$GID -v `pwd`:/data lpryszcz/modphred-5.0.11 \
  /opt/modPhred/run -f /data/ref/ECOLI.fa -o /data/modPhred5/$acc \
  -i /data/$acc/{MARC_ZFscreens_R9.4_1D-Ecoli-run_FAF05145,MARC_ZFscreens_R9.4_2D-Ecoli-run_FAF05711} \
  -t4 --host /usr/bin/guppy_basecall_server -c dna_r9.4.1_450bps_modbases_5mc_hac.cfg

If you wish to use the original dam-dcm-cpg model with the latest versions of guppy, you can find a copy in the /data folder in modPhred repository.