Flow ships with tasks that do not have all of their dependencies included.  On startup Flow will attempt to install the dependencies, but not every system is equipped to install them.

In the case of any difficulties, it is highly recommended to instead use a docker deployment  (cluster installations may require singularity instead, which is somewhat still a work-in-progress)

CNVkit

Requires Python 2.7 or later.

On startup Flow will attempt to install additional python packages using the command pip install --user cnvkit==0.89.5

If pip cannot be installed on the target machines then the packages can be downloaded from: 

customer.partek.com/cnvkit_packages-0.8.5.zip for ubuntu

customer.partek.com/cnvkit_packages-centos7.zip  for centos7

Extract the file in the home directory (Make .local a symlink if the home directory doesn't have enough free space)

Requires R 3.2.3 or later.

...

If these packages can't be built locally, it may be possible for the user to download them from us (see below).

RcppArmadillo may also have dependencies on multi-threading shared objects that may not be on the LD_LIBRARY_PATH

...

Additional dynamic libraries (such as libxml2.so) may be missing and we can provide a copy appropriate for the target OS.

HTSeq

Requires Python 2.7 or 3.4 or above

On startup Flow attempts to install using pip

Python

If there are any conflicts with preinstalled python packages, Flow should be configured to run with its own virtual environment:

pip install virtualenv

virtualenv ~/.partekflow/.local

...

(In the next build this file, if it exists will be auto sourced by start_flow.sh)

pip install HTSeq==0.11.0

pip install cnvkit==0.89.5

or 

wget customer.partek.com/python-dependencies.zip

unizp -d ~/.partekflow/ python-dependencies.zip

Variant Effect Predictor

This is a compiled Perl script (so it has no direct dependency on Perl itself) we have had one report (istem.fr) of it failing to run. 

...

For older operating systems R is not available and will need to be installed from source 

Currently, we offer a set of R packages compatible with some versions of R

...

These packages include the dependencies for both CNVkit and DESeq2

When running R diagnostic commands outside flow, it can simplify things if the environment includes a reference to the ~/.R folder:

...

.libPaths("~/.R")

in ~/.Rprofile

list loaded packages: 

(.packages())

get the version:

packageVersion("packageName")

R_HOME=/path/to/R

see also:  Minimum System Requirements

...

...