OST Docker
==========

.. note::

  For many docker installations it is required to run docker commands as root. As
  this depends on set up, we skip the ``sudo`` in all commands.

Obtain Docker image from the OST registry
-----------------------------------------

OST has its own container registry inside GitLab. There we try to keep one
image for the latest stable version of OST. You can import it by

.. code-block:: bash

  docker pull registry.scicore.unibas.ch/schwede/openstructure:<TAG>

and just start using it without the overhead to build it yourself. A list of
available tags can be found on our
`GitLab registry page <https://git.scicore.unibas.ch/schwede/openstructure/container_registry/>`_.
A tag named ``latest`` is always available, pointing to the current stable OST release.


Build Docker image
------------------

In order to build OST image:

.. code-block:: bash

  cd <PATH TO OST>/docker
  docker build --tag <IMAGE NAME> -f Dockerfile .

or if you downloaded the Dockerfile directly:

.. code-block:: bash

  docker build --tag <IMAGE NAME> -f <DOCKERFILE NAME> <PATH TO DOCKERFILE DIR>

You can chose any image name (tag) eg. ost.
Here we only keep the recipe for the most recent version of OpenStructure. To
build an image for a different version, you can either adapt the
``OPENSTRUCTURE_VERSION`` variable in the recipe or look in the git history for
an older recipe.

Testing the image
-----------------

One can find a exemplary script (``test_docker.py``) in the downloaded directory.
To run it do:

.. code-block::

  cd <PATH TO OST>/docker
  docker run --rm -v $(pwd):/home <IMAGE NAME> test_docker.py

As the last line you should see ``OST is working!``.

Run script and action with OST
------------------------------

.. note::

  If script or action requires some external files eg. PDBs, they have to be located in the
  path accessible via mounted volume and should be accessed via docker (NOT LOCAL)
  path. Eg. assuming that we have a struc.pdb file in /home/user/pdbs directory and
  a script.py in /home/user we could mount the /home/user to /home in docker as
  above by specifying -v /home/user:/home. To run the script we thus need to
  provide the (relative) path to the script and (relative) path to the file eg:

  .. code-block:: bash

    docker run --rm -v /home/user:/home <IMAGE NAME> script.py pdbs/struct.pdb

  or with absolute paths:

  .. code-block:: bash

    docker run --rm -v /home/user:/home <IMAGE NAME> /home/script.py /home/pdbs/struct.pdb
  
  An easy solution to mount a CWD is to use $(pwd) command in the -v option
  of the Docker. For an example see the action exemplary run.
  The same reasoning is valid for the output files.

Actions
#######

To see the list of available actions do:

  .. code-block::

    docker run --rm <IMAGE NAME> -h

To run chosen action do:

  .. code-block::

    docker run --rm <IMAGE NAME> <ACTION NAME>

 
Here is an example run of the compare-structures action:

.. code-block::

  docker run --rm -v $(pwd):/home <IMAGE NAME> compare-structures \
      --model model.pdb \
      --reference reference.cif \
      --output scores.json \
      --lddt \
      --local-lddt

In order to see all available options for this action run:

.. code-block::

  docker run --rm <IMAGE NAME> compare-structures -h

CASP15 used lDDT for RNA scoring. lDDT runs stereochemistry checks by default,
removing sidechains if they have problematic stereochemistry. This gives lower
lDDT scores. The full residue is removed if the backbone has problematic
stereochemistry resulting in an lDDT score of 0.0 for that particular residue.
Stereochemistry checks for RNA were not yet available in CASP15. To reproduce
these results, use the ``--lddt-no-stereochecks`` flag. This disables
stereochemistry checks for lDDT computation but stereochemical irregularities
are still reported in the output.

Scripts
#######

In order to run OST script do:

.. code-block:: bash

  docker run [DOCKER OPTIONS] --rm -v <PATH TO SCRIPT DIR>:/home <IMAGE NAME> /home/<SCRIPT NAME> [SCRIPT OPTIONS]

Run ost with utility command
###############################

One can also use provided utility bash script ``run_docker_ost`` to run basic
scripts and actions:

.. code-block:: bash

  <PATH TO OST>/docker/run_docker_ost <IMAGE_NAME> [<SCRIPT_PATH>] [SCRIPT OPTIONS]

One just needs to provide image name and optionally a script/action and its
options. It is useful to link the command to the binary directory eg. in linux:

.. code-block:: bash

  ln -s <PATH TO OST>/docker/run_docker_ost /usr/bin/run_docker_ost

In order to run an exemplary script (``test_docker.py``) do:

.. code-block::

  cd <PATH TO OST>/docker
  ./run_docker_ost <IMAGE NAME> test_docker.py

To see the help for compare-structures action run:

.. code-block::

  cd <PATH TO OST>/docker
  ./run_docker_ost <IMAGE NAME> compare-structures


Running other commands
----------------------

The default entrypoint of the Docker image is "ost" thus in order to run other
available commands (or other commands in general) one need to override
the entrypoint:

.. code-block::

  docker run --rm -ti --entrypoint <COMMAND> <IMAGE NAME> [COMMAND OPTIONS]

Eg. to run molck type:

.. code-block::

  docker run --rm -ti --entrypoint molck <IMAGE NAME> --help

.. note::

  Note how the options to the command are specified after the image name.

.. _docker_compound_lib:

The Compound Library
--------------------

At build time of the container, a :class:`~ost.conop.CompoundLib` is generated.
Compound libraries contain information on chemical compounds, such as their
connectivity, chemical class and one-letter-code. The compound library has
several uses, but the most important one is to provide the connectivy
information for the rule-based processor.

The compound library is generated with the components.cif dictionary provided by
the PDB. As the PDB updates regularly, the compound library shipped with the
container is quickly outdated. For most use cases, this is not problematic.
However, if you rely on correct connectivity information of the latest and
greatest compounds, you have to keep the compound library up to date manually.

If you work with ligands or non standard residues, or simply if you download
files from the PDB, it is recommended to generate your own compound library and
mount it into the container.

The simplest way to create a compound library is to use the
:program:`chemdict_tool` available in the container. The program allows you
to import the chemical description of the compounds from a mmCIF dictionary,
e.g. the components.cif dictionary provided by the PDB.
The latest dictionary can be downloaded from the
`wwPDB site <http://www.wwpdb.org/ccd.html>`_.
The files are rather large, it is therefore recommended to download the
gzipped version.

After downloading the file use :program:`chemdict_tool` in the container to
convert the mmCIF  dictionary into our internal format:

.. code-block:: bash

  docker run --rm -v $(pwd):/home --entrypoint chemdict_tool <IMAGE_NAME> \
  create components.cif.gz compounds.chemlib

To run a script with the updated compound library, use the -v option for
mounting/overriding, and the --env option to set the ``OST_COMPOUNDS_CHEMLIB``
environment variable inside the container, pointing to the path of the
mounted file:

.. code-block:: bash

  docker run --rm -v /home/<USER>:/home \
  -v <COMPLIB_DIR_LOCALHOST>/compounds.chemlib:/compounds.chemlib \
  --env OST_COMPOUNDS_CHEMLIB=/compounds.chemlib \
  <IMAGE_NAME> script.py pdbs/struct.pdb

with COMPLIB_DIR_LOCALHOST being the directory that contains the newly generated
compound library with name compounds.chemlib.

You can check whether the default lib is successfully overriden by looking at the
output when running a Python script with following code in the container:

.. code-block:: python

  from ost import conop
  lib = conop.GetDefaultLib()
  print(lib.GetCreationDate())