OpenTox Tutorials
The OpenTox Tutorial report represents a collection of tutorial materials on several OpenTox topics including walk-throughs of the two end user prototype applications ToxPredict and ToxCreate, illustration of the use of validation and reporting services applied to predictive toxicology models, the application of OpenTox facilities in a drug discovery workflow, and detailed instructions on how to get a system set up to host an OpenTox data service.
The tutorial example on the prototype OpenTox application ToxPredict (www.toxpredict.org) accepts chemical structures and names as input from the user and generates toxicity reports based on various pre-calibrated toxicity models and existing toxicity data.
In the ToxCreate (www.toxcreate.org) tutorial, the user provides a dataset of chemical structures and target variable data. ToxCreate subsequently builds and validates a Quantitative Structure-Activity Relationship (QSAR) predictive toxicology model. The user receives a reporting on details of model results and model predictions which they may examine, and including using the model for new predictions.
In the in vitro data model building tutorial, a predictive model is built based on in vitro data using OpenTox web services. Several models can be built and inspected based on application to the US EPA ToxCast dataset.
The tutorial on web validation and reporting web services, which are also behind the end user applications ToxPredict and ToxCreate, shows how cURL calls can be used to validate a predictive model or an algorithm. A number of different validation methods are used, including K-fold split, training-test-split and bootstrapping. Furthermore, QMRF reports are generated and visualized using the QMRF Editor web start application.
The objective of the ISSMIC data analysis tutorial is to illustrate searching facilities and data visualization tools in the OpenTox framework, specifically in the context of in vivo micronucleus mutagenicity assays contained within ISSMIC, a curated database, containing critically-selected information on chemical compounds tested with the assay.
A tutorial example of a predictive toxicology application in drug discovery is provided using the data on anti-malarial compounds made available at the ChEMBL Neglected Tropical Disease (NTD) archive (www.ebi.ac.uk/chemblntd/). The anti-malarial compounds are prioritized based on a strongly conservative model for predicting oral toxicity. Experimentally-determined cytotoxicities against human cells of the compounds predicted to be safe are further examined, and their mutagenicities predicted. Sites of cytochrome P450 metabolism are predicted for selected compounds with no mutagenicity alerts, low human cytotoxicity, but high anti-malarial activity.
A tutorial is provided to guide the user through the setup of an OpenTox data service based on the download of the AMBIT software and its subsequent installation either on Windows or Linux.
The tutorial report is available for download at http://www.opentox.org/data/documents/development/opentoxreports/opentoxreportd62
All tutorials and their updates wiil made available online shortly under www.opentox.org/tutorials (to be opened 14 March 2011).
