Use ToxPredict to obtain REACH-relevant information on compounds
Facilitated by David Gallagher and Barry Hardy
Objectives
To demonstrate and evaluate a prototype OpenTox application (ToxPredict) that accepts chemical structures and names as input and automatically generates a toxicity report based on various pre-calibrated toxicity models. We will examine the relevance and usefulness of the information provided for potential REACH reporting purposes.
Description of ToxPredict
ToxPredict is a web-based interface for predicting toxicity of individual or sets of chemicals. Users can either search for a compound in the OpenTox prototype database, which currently includes quality labeled data for 243,648 chemicals (including all chemicals currently registered with REACH) grouped in 25,170 datasets, or upload their own chemical structure in the SDF format. It runs the selected calculations automatically using a collection of distributed computational services. ToxPredict currently includes 24 validated models addressing 11 different endpoints, and is extensible.
Prerequisites: A computer with browser (Internet Explorer or Firefox) and fast internet connection.
Note: any compounds not found in the OpenTox database, must be uploaded as SDF or MOL files.
Summary of Activities
In this workshop activity we will
a) Review the Information Gathering Requirements of REACH
b) Work through examples using the OpenTox-based application ToxPredict to discover and discuss what REACH-relevant information OpenTox currently provides
c) Provide time for users to work on their own examples (You may for example bring an sdf file with you for a compound or a library of ca. 5-100 molecules or you can draw your own structure and submit it for predictions)
d) Discuss requirements and features you would like to see included in the application for your future work.
e) Complete an evaluation providing your feedback.
Exercise Sequence
1) Login to Toxpredict (www.toxpredict.org) using userid "guest" and password "guest".
2) Example 1: Type the name Benzene into the query text box, select the "Auto Detect" option, and click "Search".
3) Check that benzene and its associated information page has been returned to you. Click the "Run All" button on this new page.
4) Take a few minutes to review the toxicity-related information presented to you.
5) Focus in on the IST Kazius-Bursi Salmonella mutagenicity classification model. Detailed model validation information can be found linked into the Browse Models area of ToxPredict. (Or is also available at http://lazar-services.in-silico.ch/validation/report/crossvalidation/6) What does the prediction mean? How reliable are the results?
6) Example 2: Try searching for nitrosamines (e.g., enter the SMILES CCN(CC)N=O in query box, select substructure match) and click "Search".
7) View structures presented to you, select one of interest, and Click the "Run All" button.
8) Take a few minutes to review the toxicity-related information presented to you.
9) Focus in on the IST EPA v4b Fathead Minnow Acute (Fish) Toxicity regression model. Detailed model validation information can be found linked into the Browse Models area of ToxPredict. (Or is also available at http://lazar-services.in-silico.ch/validation/report/crossvalidation/28) What does the prediction mean? How reliable are the results?
10) Example 3: Repeat the procedure for the dioxin 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), [SMILES = Clc2cc1Oc3c(Oc1cc2Cl)cc(Cl)c(Cl)c3]
11) Example 4: Try out your own examples and discuss results.
12) Complete the evaluation form following this link: Evaluation Form
