xiongzhp/OCEAN

OCEAN - Optimized Cross rEActivity estimatioN

So what is OCEAN?

OCEAN is a web-tool for target-prediction of chemical structures which uses ChEMBL as datasource. For more details please feel free to read the paper.

Software dependencies

I recommend to use myChembl21-VM which has almost everything you need to run OCEAN.

Bare metal - if you have a clean Ubuntu or CentOS box with root access and want to install myChEMBL software directly, then you may run:

wget https://raw.githubusercontent.com/chembl/mychembl/master/bootstrap.sh && chmod +x bootstrap.sh && bash bootstrap.sh

Nevertheless here are the core-depencies:

RDKit (2015.03.1) http://www.rdkit.org/
Django (1.7.1)
numpy (1.9.0)
sciPy (0.15.1)
your preferred database-server (Oracle or PostgreSQL, we didn't test MySQL or sqlite but they are supported by Django as well)
python module for your database-server (cx_oracle for Oracle, psycopg2 for PostgreSQL)

Optional dependencies

matplotlib (1.3.0)
pillow (2.6.1)
images2gif (1.0.1)

Easy setup using myCHEMBL-VM

Get a clone of OCEAN from the Repo:

git clone https://www.github.com/rdkit/OCEAN.git
cd OCEAN

Create a PostgreSQL-User, the database for OCEAN and fill the OCEAN-DB:

# create PostgreSQL-User for OCEAN
# password 'ocean_pw' should be used when asked,
# according to database-entry in settings.py
createuser -P -s -d -r -e ocean_user

# create PostgreSQL-Database for OCEAN
createdb --owner=ocean_user ocean

If you want to use ChEMBL17-data for OCEAN you should:

change line CHEMBL_VERSION = "chembl_%d" % (20) in file ocean/settings.py to CHEMBL_VERSION = "chembl_%d" % (17)

createdb --owner=ocean_user chembl_17

# create and populate OCEAN-DB (duration <2 min)
psql chembl_17 < postgres_createOceanDB.sql
./importOcean17DB.sh

# Chembl17 is not a part of myChembl20-VM, so we need to create and populate
# a few tables (duration <10 sec)
psql chembl_17 < postgres_createChembl17Tables.sql
./importChembl17Tables.sh

For ChEMBL21-data you should:

psql chembl_21 < postgres_createOceanDB.sql
./importOcean21DB.sh

It's time to start OCEAN:

# On first start, you have to sync the django-server to create all missing tables.
# When asked, you should create a superuser and set a password.
# This is necessary to have access to the django-admin portal of OCEAN.
python manage.py syncdb

# Start OCEAN
python manage.py runserver 0.0.0.0:8080

The next steps are necessary to create the fingerprints for all structures:

Login to ocean-admin portal http://0.0.0.0:8080/admin
Select CHEMBL and 6 (section a), press createFPs to create Morgan-Fingerprints for all entries. This may take a few minutes. You should see a message at the bottom of the site when the process is done.

The last step is calculating the statistical parameters for OCEAN's scoring-function:

In the Django-Admin portal select CHEMBL, 6 and activate Checkbox rebuild Pairs (section b), press recalc Statistics for Fingerprint and Datasource to calculate statistic-parameters for the scoring-function. This will also take a few minutes. Wait until you see a new message at the bottom of the site.

If you have a MarvinJS licence, you can copy the MarvinJS-folder to ocean/media/MarvinJS and change the line url(r'^$', main_smiles, name='home') in ocean/urls.py to url(r'^$', main_marvin, name='home') to use MarvinJS as sketcher.

Now lets try it:

Open ocean-website http://0.0.0.0:8080
Write a smiles-string into the smiles-field or sketch a molecule (if you have MarvinJS) and press Button Start OCEAN-Search. You should see the results after a few seconds.

Questions or Problems?

If you have any questions or problems with OCEAN please do not hesitate to contact Paul Czodrowski or Wolf-Guido Bolick.

xiongzhp / OCEAN