Skip to content
/ polo Public

POLO: web interface to MARCO-scored crystallization images

License

Notifications You must be signed in to change notification settings

Merck/polo

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

POLO: web interface to MARCO-scored crystallization images

The MARCO crystallization classification model was published in 2018 and seems to be very robust for many sorts of visible crystallization images. This model has incorporated into the RockMaker software and can be implemented as an internal network service.

The POLO system is a web interface to access MARCO-classified images, which is intended to be a very rapid method to find crystals.

The motivation behind this effort is two-fold:

  • The RockMaker interface is plate-based. When examining broad screens, a sorted list may be more effective.
  • Depending on network configuration, the RockMaker interface can be slow and it may become laborious to find crystals on a set of plates in a set of experiments, to zoom in on a single image, and to evaluate a time course of images.

POLO in a nutshell

  • Images are scored periodically using an separate MARCO implementation (this is outside the scope of POLO). Note that the RockMaker MARCO scores are a weighted combination of scores, so will not match exactly the scores in POLO. These scores are stored in a MySQL database.
  • Quickly navigate or search a project/experiment tree, just as in RockMaker, to select one or more plates (useful for broad screens), even if they are from different RockMaker experiments.
  • Browse drops, sorted by their MARCO “Crystal” or “Other” scores, or other information such as well number, and see the time course. You can filter results by drop number and/or temperature.
  • “Favorite” a drop (akin to RockMaker’s “Interesting” tag), though this is not saved.
  • Export a list to a CSV file for further analysis, for example to include in a report or to bring to the lab for sample prep.
  • Keyboard shortcuts (e.g. arrows) for most of the functionality, which makes zipping through images a breeze.
  • You can manually annotate images using keyboard shortcuts (CPXO, D for dispute, and S for salt/false positive). Annotations are stored in the same MySQL database which contains the MARCO scores.
  • You can limit the tree to a single top-level folder, use https://POLOSERVER/?folder=NAME

How to run a POLO server from a Docker container

Download the POLO code from github.com.

mkdir polo-docker
cd polo-docker
git clone ...

Build the POLO Docker container. Put the following into a file called Dockerfile in the current directory.

FROM python:3.7-slim

# if you are behind a firewall, you may need to set the https proxy to enable downloading of external content
ENV https_proxy "https://webproxy.company.com:8080"

RUN apt-get update && apt-get -y install \
        build-essential \
        unixodbc \
        unixodbc-dev \
        tdsodbc \
        freetds-bin \
        freetds-dev \
        mariadb-client \
        mariadb-common \
        libmariadbclient-dev \
        host
WORKDIR /app
COPY polo/requirements.txt .
RUN pip3 install -r requirements.txt
COPY polo .
CMD bash freetds.sh && gunicorn --bind=0.0.0.0:9002 --workers=5 --preload polo:app

Build the POLO image (let's call it "polo")

docker build -t polo .

All settings are configured using a file called polo.env where each line is a key=value pair An example polo.env file looks like this:

FLASK_ENV=production

#POLO_INFO=an informational message
#POLO_WARNING=a warning message
#POLO_ERROR=an error message

# POLO database connection information
POLO_CONN=user:password@server/database

# RockMaker server configuration. The index should match the id in the POLO database `sources` table
# Up to 10 RockMaker databases can be specified
# Specify EITHER RM_x_PORT or RM_x_INSTANCE
RM_1_HOST=
RM_1_PORT=
RM_1_UID=
RM_1_PWD=
RM_1_DATABASE=

RM_2_HOST=
RM_2_INSTANCE=
RM_2_UID=
RM_2_PWD=
RM_2_DATABASE=

# OAuth2 configuration, or set BYPASS_AUTH to True
BYPASS_AUTH=False
OAUTH2_AUTHORIZATION_BASE_URL=
OAUTH2_TOKEN_URL=
OAUTH2_USERPROFILE_URL=
OAUTH2_USERNAME_FIELD=
OAUTH2_CLIENT_ID=
OAUTH2_CLIENT_SECRET=

# If running the POLO Docker container behind a reverse proxy server (e.g. nginx) or running over HTTP
OAUTHLIB_INSECURE_TRANSPORT=1

You should have a directory structure which looks like this:

Dockerfile (file)
polo.env (file)
polo (directory)

Configure the polo.env file and run the container in the background, in this case over port 9002.

docker run -d -it --rm -p 9002:9002 --env-file polo.env --name polo polo

The POLO Database

  • You need a POLO database. Currently this is configured as a MySQL database. See the file called scripts/define_tables.sql for more information. The RockMaker image server information needs to be set up in the sources table.
    • The connection string for the POLO database will look something like: USERNAME:PASSWORD@MYSQLHOST:PORT/DATABASE_NAME
    • This database can be updated regularly (e.g. using a cron entry) by running a suitable script. There is a template available called scripts/Update_POLO.py.

Image server(s)

  • You need to serve RockMaker images

    • An easy way to configure this is to mount the Windows share under linux (let's call it /mnt/RockMaker).
    • Then start a server. Using Python2, you can cd /mnt/RockMaker ; python -mSimpleHTTPServer <PORT> to serve images on port .
    • If you use Python 3, try python -m http.server <PORT>
  • Be sure that the dictionary of RockMaker instances is reflected in the MySQL sources table. Again consult the file scripts/define_tables.sql for additional information.

Credits

License

  • Distributed under the MIT License. See the LICENSE file for details
  • See the LICENSES_THIRD_PARTY file for additional license information