&EPA
www.epa.gov/research
science   in   ACTION
INNOVATIVE RESEARCH FOR  A SUSTAINABLE FUTURE
     Estimation of Toxicity Using the Toxicity Estimation Software Tool (TEST)
     Problem
     Tens of thousands of chemicals are
     currently in commerce, and
     hundreds more are introduced every
     year. Since experimental
     measurements of toxicity are
     extremely time consuming and
     expensive, it is imperative that
     alternative methods to estimate
     toxicity are developed.

     Solution
     Quantitative Structure-
     Activity/Property Relationship
     (QSAR) models can be used to
     estimate toxicity from molecular
     structure (the atoms from which a
     molecule is built and the particular
     way the atoms are linked together).

     To make use of these models, EPA
     scientists developed the Toxicity
     Estimation Software Tool (TEST).
     TEST is a freely available program
     that allows users to easily estimate
     toxicity from molecular structure.

     Endpoints
     TEST estimates toxicity values for a
     variety of toxicity endpoints
     including:
     •   Acute aquatic toxicity
     •   Acute mammalian toxicity
     •   Bioaccumulation factor
     •   Developmental toxicity
     •   Mutagenicity
     TEST can estimate physical
     properties including:
     •   Boiling point
     •   Density
     •   Water solubility
     •   Thermal conductivity
               T.E.S.T (Toxicity Estimation Software Tool)
               File  Edit
                                                                    Help
3Ti
                           -1
c
H
0
N
P 1
S
F
Cl
Br
1


Sh
As
                                               CH
                            H,C
               <
               Molecule ID: [333-41-S    | Endrjoint: [Fathead minnow LCSO (96 hr) [r]  Method: [Consensus


              Q Relax fragment constraint                                   Calct
             (Above) Structure entry with TEST

             TEST can provide toxicity values for
             the WAste Reduction Algorithm
             (WAR). WAR is used to evaluate
             the environmental friendliess of
             chemical process designs.1 TEST
             can provide toxicity and physical
             property values for the PARIS II
             (Program for Assisting the
             Replacement of Industrial Solvents)
             methodology.2 PARIS II is used to
             find more environmentally friendly
             replacements for industrial solvents.

             QSAR Methodology
             QSARs are mathematical models
             used to predict measures of toxicity
             from physical characteristics of the
             structure of chemicals (known as
             molecular descriptors). Examples of
                            molecular descriptors include
                            connectivity indices and molecular
                            fragment counts (such as the number
                            of hydroxyl groups attached to
                            aromatic rings).

                            TEST utilizes a consensus approach
                            to make toxicity predictions. In the
                            consensus approach, predictions are
                            made using the average of the
                            predictions from several different
                            QSAR methods including
                            hierarchical clustering and group
                            contribution based methods.3 The
                            QSAR models were developed using
                            a pool of more than 800 descriptors.
                            Predictions from each QSAR
                            methodology are subject to a series
                            of constraints that ensure each
                            predicted value represents an
            U.S. Environmental Protection Agency
            Office of Research and Development
                                                       EPA600-F11019
                                                             July 2011

-------
interpolation rather than an
extrapolation.

Model Validation
The QSAR models were validated
by randomly dividing the data sets
into training and prediction sets. The
results from the validation exercises
(along with the data sets) are
provided with the software. The
prediction statistics typically exceed
those from other QSAR software
packages for the same endpoints.

TEST Results
TEST provides the prediction results
using a series of web pages so the
calculation details are completely
transparent.

References
1. Comp. Chem. Eng., 1999, 23
  1477-1491.
I.Ind. Eng.  Chem. Res., 2002, 41,
  5867-5877.
3. Toxicol. Mech. Method, 2008,18,
  251-266.

Contacts
Todd Martin, Ph.D., Office of Research
& Development, 513-569-7682,
martin.toddfgiepa.gov

Douglas Young, Ph.D., Office of
Research & Development, 513-569-
7624, young.douglasfatepa. gov
(Below) Results from TEST
  Toxicity prediction results for 333-41-5 for Consensus method

                          Prediction results
Endpoint
Fathead minnow LC^o (96 hr) -Log(mol'L)
Fathead minnow LCso (96 hr) mg/L
Experimental value
CAS: 333-41-5
Semite: ECOTOX
4.81
4.70
Predicted value*
5.20
1.91
  4Note: the test chemical was present in the external test set.
          Individual Predictions
Method
Hierarchical clustering
Single model
Group contribution
FDA
Nearest neighbor
Pnedk ted value
-Log(mol/L)
5.39
4.92
4.84
4.89
5.9(5
        U.S. Environmental Protection Agency
        Office of Research and Development
                       Recycled/ recyclable
                       Printed with vegetable-based ink on
                       paper mat consists of a minimum of
                       50% post-consumer fiber content
                       processed chlorine free
EPA600/R-12/015
        July 2011

-------