LAZAR-WEBSERVICE

This is an example OpenTox webservice for the lazar prediction system and public data from the DSSTOX project.

REST API

Description	Method	URI	Parameters
List all models	GET	/models
Show a single model	GET	/models/{id}
Predict smiles	GET	/models/{id}	smiles
Create a model	POST	/models	structure_file, activity_file, username, password
Delete a model	PUT	/models/{id}	username, password

Return values

Success: XML file with requested information
Failure: Text with error description

Input file format

To create a model you have to POST files with training data. Both input files should have 2 columns, delimited by a tab or whitespaces. No headers are allowed.

structure_file:

First column: (Unique) compound ID
Second column: Smiles string (structure_file)

activity_file:

First column: Compound ID
Second column: Toxic activity (1 (active) or 0 (inactive) for classification)
Note: numerical values for regression problems will be supported in a future version

You can provide more than one activity per compound ID (for replicated experiments). Activites without structures and structures without activities will be ignored.

Examples

Display all available models

curl -X GET http://localhost:3000/models

Display information for a single model

curl -X GET http://localhost:3000/models/1

Note: Detailed model annotations (data sources, endpoint definitions, validation results, …) will be provided by a separate component.

Predict rodent carcinogenicity (CPDB multi_cell_call) for Aniline

curl -X GET http://localhost:3000/models/1?smiles=c1cccc1N

Important: Smiles strings have to be URL encoded!

Create a new prediction model for hamster carcinogenicity

curl -X POST -F structure_file=hamster_carcinogenicity.smi -F activity_file=hamster_carcinogenicity.class -F username=secret -F password=secret http://localhost:3000/models

Delete a model

curl -X DELETE -d username=secret -d password=secret http://localhost:3000/models/1

Error codes

200 OK
404 Not Found
500 Internal Server Error
503 Service Unavailable

Installation

• Get the source code for the webservice

git clone git://github.com/helma/lazar-webservice.git

• Install lazar-core and prepare the lazar webservice

cd lazar-webservice

rake lazar:setup

	

If lazar-core does not compile

* Make sure, that you have the development libraries Swig, GSL, Boost, OpenBabel, R installed
* Configure R with ./configure --without-blas --enable-BLAS-shlib --enable-R-shlib
* Install the R kernlab package:

wget http://cran.r-project.org/src/contrib/kernlab_0.9-8.tar.gz

R CMD INSTALL kernlab_0.9-8.tar.gz

			

* Check the paths in the Makefile of lib/lazar-core and adjust them to your system settings

• Start the service

script/server

Now you should be able to contact localhost:3000 and use the REST API.