2

Volume

IDEACONSULT

ToxtreeUserManual
 I D E ACONS UL T

Toxtree User Manual

 Ideaconsult Ltd.
4 Angel Kanchev St.
1000 Sofia, Bulgaria
Phone +359 886802011 • Email nina@acad.bg

Version of 15 July 2009

Table of Contents
Introduction _____________________________________________________________________ 1
Background _____________________________________________________________________ 3
Main features at a glance __________________________________________________________ 5
Development tools ________________________________________________________________ 6
Launching Toxtree________________________________________________________________ 7
Main screen layout ________________________________________________________________ 8
Opening/saving a file _____________________________________________________________ 10
Classifying/reading the result ______________________________________________________ 11
Decision tree visualisation _________________________________________________________ 12
Decision tree options visualisation __________________________________________________ 13
Decision tree selection ____________________________________________________________ 16
Decision tree editing______________________________________________________________ 20
Splitting the file into groups _______________________________________________________ 23
Structure diagram editor __________________________________________________________ 26
Batch processing ________________________________________________________________ 27
A typical usage scenario __________________________________________________________ 28
Command line options ____________________________________________________________ 29
Cramer rules specifics ____________________________________________________________ 30
Verhaar scheme specifics__________________________________________________________ 33
Skin irritation/corrosion rules specifics ______________________________________________ 34
Eye irritation/corrosion rules specifics _______________________________________________ 37
Benigni / Bossa rulebase (for mutagenicity and carcinogenicity) specifics __________________ 43
Decision tree editing - typical tasks __________________________________________________ 48
Creating a simple decision tree __________________________________________________________ 48
Using a decision tree ___________________________________________________________________ 59
T O X T R E E U S E R M A N U A L

Introduction
Toxtree is a full-featured and flexible user-friendly open source application, which is
able to estimate toxic hazard by applying a decision tree approach. Currently it includes
the following plug-ins:

• Cramer rules (Cramer G. M., R. A. Ford, R. L. Hall, Estimation of Toxic

Hazard - A Decision Tree Approach, J. Cosmet. Toxicol., Vol.16, pp. 255-
276, Pergamon Press, 1978);

• Verhaar scheme for predicting toxicity mode of actions (Verhaar HJM, van
Leeuwen CJ and Hermens JLM (1992) Classifying environmental pollutants. 1.
Structure-activity relationships for prediction of aquatic toxicity. Chemosphere
25, 471-491);

• A decision tree for estimating skin irritation and corrosion potential, based on
rules published in “The Skin Irritation Corrosion Rules Estimation Tool
(SICRET), John D. Walker, Ingrid Gerner, Etje Hulzebos, Kerstin Schlegel,
QSAR Comb. Sci. 2005, 24, pp. 378-384”;

• A decision tree for estimating eye irritation and corrosion potential, based on
rules published in “Assessment of the eye irritating properties of chemicals by
applying alternatives to the Draize rabbit eye test: the use of QSARs and in
vitro tests for the classification of eye irritation, Ingrid Gerner, Manfred
Liebsch & Horst Spielmann, Alternatives to Laboratory Animals, 2005, 33, pp.
215-237”;

• A decision tree for estimating carcinogenicity and mutagenicity, based on the

rules published in the document: “The Benigni / Bossa rulebase for
mutagenicity and carcinogenicity – a module of Toxtree”, by R. Benigni, C.
Bossa, N. Jeliazkova, T. Netzeva, and A. Worth. European Commission report
EUR 23241 EN1;

• START (Structural Alerts for Reactivity in Toxtree) biodegradation and

persistence plug-in is based on a compilation of structural alerts for
environmental persistence and biodegradability. These structural alerts are
molecular functional groups or substructures that are known to be linked to
the environmental persistence or biodegradability of chemicals. The rulebase
utilizes the structural alerts in logical decision trees. If one or more the

1 http://ecb.jrc.it/documents/QSAR/EUR_23241_EN.pdf

1
T O X T R E E U S E R M A N U A L

structural alerts embedded in the molecular structure of the chemical are

recognized, the system flags the potential persistence or biodegradability of the
chemical. Installation and user manuals available2;

• Structure Alerts for the in vivo micronucleus assay in rodents, based on the
rules, published in the document “Development of structural alerts for the in
vivo micronucleus assay in rodents”, by Romualdo Benigni, Cecilia Bossa,
Olga Tcheremenskaia and Andrew Worth3, European Commission report
EUR 23844 EN;

• Cramer rules with extensions: This plug-in is a copy of the original plug-in,
plus minor extensions. Like the Cramer plug-in, this plug-in works by assigning
compounds to Class I, II, or III, according to the rules from Cramer, and
some extra ones. Several compounds were classified by Munro in 19964 as
Class I or Class II compounds according to the Cramer rules, even though
Munro reported low NOEL values upon oral administration (indicating
relatively high toxicity). To overcome such misclassifications, five rules have
been introduced to capture the possible toxicity of these compounds;

• Structure Alerts for identification of Michael Acceptors: This plug-in contains

structural alerts, able to identify Michael Acceptors, as defined in T. Wayne
Schultz, Jason W. Yarbrough, Robert S. Hunter, Aynur O. Aptula (2007)
Verification of the Structural Alerts for Michael Acceptors. Chem. Res.
Toxicol. 20, 1359–1363.

Toxtree could be applied to datasets from various compatible file types. User-defined
molecular structures are also supported - they could be entered by SMILES, or by
using the built-in 2D structure diagram editor.

The Toxtree application is suitable for a standalone PC. It has been designed with
flexible capabilities for future extensions in mind (e.g. other classification schemes that
could be developed at a future date). New decision trees with arbitrary rules can be
built with the help of graphical user interface or by developing new plug-ins.

2 http://ecb.jrc.ec.europa.eu/DOCUMENTS/QSAR/QSAR_TOOLS/Toxtree_start_manual.pdf

3 http://ecb.jrc.ec.europa.eu/DOCUMENTS/QSAR/EUR_23844_EN.pdf

4 I.C. Munro, R.A. Ford, E. Kennepohl, and J.G. Sprenger, Correlation of structural class with No-Observed-

Effect Levels: A proposal for establishing a threshold of concern, Food Chem. Toxicol.34 (1996), pp. 829–
867.

2
T O X T R E E U S E R M A N U A L

Background
Cramer rules

The threshold of toxicological concern (TTC) is an open research topic with significant
practical implications. Two principal approaches exist in the thresholds developed to
date; a general threshold or a threshold developed in relation to toxicological data or
structural information. Thresholds based on structural information have typically been
developed by the principles established by Cramer. Chemicals are divided into three
structural classes based on a decision tree. This comprises some 33 structural rules and
places evaluated compounds into one of three classes:

• Class I substances are simple chemical structures with efficient modes of

metabolism suggesting a low order of oral toxicity;

• Class III substances are those that permit no strong initial presumption of
safety, or may even suggest significant toxicity or have reactive functional
groups;

• and finally, Class II are intermediate.

The Cramer classification scheme was the first decision tree scheme, implemented in
Toxtree. If it is not possible to reduce the rule to a specific substructure search, an
extendable list of compounds is used as the input, for making a decision.

This plug-in is developed by IdeaConsult Ltd (Sofia, Bulgaria) on behalf of JRC.

Cramer rules with extensions

This plug-in is a copy of the original plug-in, plus minor extensions. Like the Cramer
plug-in, this plug-in works by assigning compounds to Class I, II, or III, according to
the rules from Cramer, and some extra ones. Several compounds were classified by
Munro in 1996 as Class I or Class II compounds according to the Cramer rules, even
though Munro reported low NOEL values upon oral administration (indicating
relatively high toxicity). To overcome such misclassifications, five rules have been
introduced to capture the possible toxicity of these compounds.

This plugin is developed by Curious-IT, The Netherlands, on behalf of JRC.

Verhaar scheme

This plugin implements the Verhaar scheme for predicting toxicity mode of action,
according to:

3
T O X T R E E U S E R M A N U A L

Verhaar H.J.M., Van Leeuven C., Hermens J.L.M., Classifying Environmental

Pollutants. 1: Structure-Activity Relationships for Prediction of Aquatic Toxicity,
Chemosphere, Vol.25, No.4, pp.471-491, 1992.

This plug-in is developed by Ideaconsult Ltd (Sofia, Bulgaria) within the AMBIT
project, funded by CEFIC LRI.

Skin irritation and corrosion prediction

Estimates skin irritation and corrosion potential by physicochemical property limits

and structural rules, according to:

1. Ingrid Gerner, Kerstin Schlegel, John D. Walker, and Etje Hulzebosc, Use of
Physicochemical Property Limits to Develop Rules for Identifying Chemical
Substances with no Skin Irritation or Corrosion Potential, QSAR Comb. Sci.
2004, 23, pp.726-733

2. John D. Walker, Ingrid Gerner, Etje Hulzebos, Kerstin Schlegel, The Skin
Irritation Corrosion Rules Estimation Tool (SICRET), QSAR Comb. Sci.
2005, 24, pp.378-384

3. Etje Hulzebos, John D. Walker, Ingrid Gerner, and Kerstin Schlegel, Use of
structural alerts to develop rules for identifying chemical substances with skin
irritation or skin corrosion potential, QSAR Comb. Sci. 2005, 24, pp.332-342

This plug-in is developed by IdeaConsult Ltd (Sofia, Bulgaria) on behalf of JRC.

Eye irritation and corrosion prediction

Estimates eye irritation and corrosion potential by physicochemical property limits and
structural rules, according to:

Ingrid Gerner, Manfred Liebsch & Horst Spielmann, Assessment of the eye irritating
properties of chemicals by applying alternatives to the Draize rabbit eye test: the use of
QSARs and in vitro tests for the classification of eye irritation, Alternatives to
Laboratory Animals, 2005, 33, pp. 215-237.

This plug-in is developed by Ivanka Tsakovska and Nina Jeliazkova on behalf of JRC.

Benigni / Bossa rulebase (for mutagenicity and carcinogenicity)

Estimates potential carcinogenicity and mutagenicity, by using: 1) a series of Structural

Alerts (SA); and 2) a number of Quantitative Structure-Activity Relationship (QSAR)
models. Details on the alerts and QSARs are provided in the document: “The Benigni
/ Bossa rulebase for mutagenicity and carcinogenicity – a module of Toxtree”, by R.

4
T O X T R E E U S E R M A N U A L

Benigni, C. Bossa, N. Jeliazkova, T. Netzeva, and A. Worth. European Commission

report EUR 23241 EN.

START (Structural Alerts for Reactivity in Toxtree) biodegradation and

persistence plug-in

A compilation of structural alerts for environmental persistence and biodegradability.

These structural alerts are molecular functional groups or substructures that are known
to be linked to the environmental persistence or biodegradability of chemicals. The
rulebase utilizes the structural alerts in logical decision trees. If one or more the
structural alerts embedded in the molecular structure of the chemical are recognized,
the system flags the potential persistence or biodegradability of the chemical.
Installation and user manual.

This plug-in is developed by Molecular Networks GmbH (Erlangen, Germany).

Structure Alerts for the in vivo micronucleus assay in rodents

Provides a list of structural alerts for a preliminary screening of potentially in vivo

mutagens. These structural alerts are molecular functional groups or substructures that
are known to be linked to the positive in vivo micronucleus assay. Details on the alerts
are provided in the document “Development of structural alerts for the in vivo
micronucleus assay in rodents”, by Romualdo Benigni, Cecilia Bossa, Olga
Tcheremenskaia and Andrew Worth European Commission report EUR 23844 EN
and Installation and user manual.

This plug-in is developed by Instituto Superiore di Sanita (Rome, Italy).

Structure Alerts for identification of Michael Acceptors

Identifies Michael Acceptors by structural alerts, defined according to:

1. T. Wayne Schultz, Jason W. Yarbrough, Robert S. Hunter, Aynur O. Aptula

(2007) Verification of the Structural Alerts for Michael Acceptors.Chem.
Res. Toxicol. 20, 1359–1363

This plug-in is developed by Ideaconsult Ltd (Sofia, Bulgaria) within the AMBIT XT
project, funded by CEFIC LRI.

Main features at a glance

 Toxtree is a standalone software application, implementing the Cramer
decision tree, the Verhaar scheme, a decision tree for skin corrosion and
irritation prediction, and the Benigni / Bossa rulebase (for mutagenicity and
carcinogenicity). Toxtree runs on Microsoft Windows ™ operating system as

5
T O X T R E E U S E R M A N U A L

well as on different platforms with Java ™ 2 Runtime Environment, Standard

Edition 1.4 or newer installed;

 the chemical structures for analysis may be submitted by using an interactive

2D graphical editor, or in a batch mode by using CSV, TXT or SDF file
formats;

 the classification results are displayed in a graphical user interface and can be
optionally saved as a file (CSV, SDF or TXT) file, together with the
classification path explanation (list of applied rules). This provides a
transparent audit of the decision path followed;

 users may modify the tree by adding their own structural rules and save a
modified tree for future re-use;

 the software provides a flexible and documented (in the doc/src application
subfolder) plug-in interface, allowing the integration of different classification
schemes developed at a future date;

 the software is packaged in a self-installing file for Windows ™, as well as in a

ZIP archive;

 the installation procedure is easy and is supplemented with an installation

manual (located in the doc application subfolder);

 an user manual is provided in the doc application subfolder;

 documentation of the source code can be found in the doc/src application

subfolder;

 the source code of the application is located in the src application subfolder;

 a comprehensive README file is located in the main application folder.

Development tools
The Toxtree application is implemented in Java ™ . The basic cheminformatics
functionality relies on the open source LGPL licensed Java ™ library Chemistry
Development Kit (CDK). The Integrated Development Environment (IDE) Eclipse,
in conjunction with Apache Ant is the main development tool. Some of the Toxtree
capabilities are provided through the following open source libraries:

 JChemPaint – a structure diagram editor, which recently became part of CDK;

6
T O X T R E E U S E R M A N U A L

 Joelib – SMARTS support (to be replaced by Ambit SMARTS in subsequent

releases)

 Ambit – SMARTS support

 org.xmlcml – CML support;

 gnujaxp – XML support;

 jgrapht – graph algorithms library;

 apache log4j – application logging;

 javax.vecmath – vector and matrix classes;

 OpenBabel5 - molecule file conversion and pattern matching;

 smi23d6 - 3D coordinate generation;

 junit – test suite framework;

 L2fprod7 – GUI components;

 prefuse8 – decision tree GUI;

Launching Toxtree
In Windows ™ platforms, Toxtree can be launched either by using the “Start” menu
(Figure 1), or by double clicking on the Toxtree-X.YZ.jar file (the full path name is
“C:\Ideaconsult\Toxtree-vX.YZ\Toxtree\Toxtree-X.YZ.jar”).

5 http://openbabel.sourceforge.net/wiki/Main_Page

6 http://www.chembiogrid.org/cheminfo/smi23d/

7 http://www.l2fprod.com/

8 http://prefuse.org/

7
T O X T R E E U S E R M A N U A L

Figure 1: Launching Toxtree on Windows™ platforms

In all platforms (having Java ™ 2 Runtime, Standard Edition 1.5 or newer installed),
Toxtree can be launched by executing the following command (after decompressing
the ZIP archive distribution of Toxtree):

java –jar Toxtree-X.YZ.jar

Please, note that in the above mentioned command “java” and Toxtree-
X.YZ.jar” should be eventually prefixed with the full path to java and Toxtree
on the destination platform.

Main screen layout

The main Toxtree application window comprises a title bar, menu bar, data areas,
button bar and status bar. The data areas are highlighted in Figure 2 in several different
colours and are labelled in blue.

The SMILES INPUT AREA can be used for entering a SMILES string.

Pressing the Go! button draws the structure diagram of the corresponding compound
in the COMPOUND STRUCTURE DIAGRAM AREA.

8
T O X T R E E U S E R M A N U A L

The << and >> buttons on the left hand side of the SMILES INPUT AREA are used for
navigating the history of entered SMILES’. This is also accessible also by means of a
drop-down list.

The COMPOUNDS PROPERTIES AREA is used to summarise the available information

about the current compound.

The FILE BROWSING AREA provides an easy way to navigate through the list of
compounds in the current open file.

The CLASSIFICATION AREA provides access to the Cramer classification results for
the current compound.

Pressing the Estimate button starts the classification routines for the current
compound (shown both in the COMPOUNDS PROPERTIES AREA and the
COMPOUND STRUCTURE DIAGRAM AREA).

The classification result is shown in graphical form (green highlight for class I, yellow
highlight for class II and red highlight for class III), as well as in text form.

The Verbose explanation checkbox determines the level of detail of the text-based
classification results.

9
T O X T R E E U S E R M A N U A L

SMILES INPUT AREA

COMPOUNDS PROPERTIES AREA

COMPOUND STRUCTURE DIAGRAM

AREA

CLASSIFICATION AREA

FILE BROWSING AREA

Figure 2: Toxtree main application window

Opening/saving a file
Toxtree can open CML, CSV, HIN, ICHI, INCHI, MDL MOL, MDL SDF, MOL2,
PDB, SMI, TXT and XYZ file types.

Please, note that CSV files can be read/written by MS Excel™. Attention

should be paid to cells type (should be ‘text’, otherwise MS Excel™ could
interpret/show incorrectly their values). Also, note that input of CSV and TXT
requires a column with "SMILES" heading in order the structure to be read.
All other fields are optional, will be read as molecule properties and
displayed as such.

You can open a supported type of file by using the “File ► Open” menu, as shown on
Figure 3. Molecules can be written to SDF, CSV or TXT files, together with their
classification data (class & path), by using the “File ► Save” menu.

10
T O X T R E E U S E R M A N U A L

Figure 3: Opening a file with Toxtree

Classifying/reading the result

In order to apply the active decision tree on the current compound (displayed on the
left hand side of the main application window), you should press the Estimate button,
located in the upper part of the CLASSIFICATION AREA.

An example of classification result visualisation is show on Figure 4. Hexane is assigned

Class I (green highlight), and a verbose text explanation is printed, after pressing the
Estimate button.

11
T O X T R E E U S E R M A N U A L

Figure 4: Classification result visualisation

Decision tree visualisation

The current decision tree is accessible through the “Method ► View decision tree”
menu. It is shown in a child window, as illustrated on Figure 5. Rule details are printed
after clicking on the respective tree nodes. Leaves are coloured according to classes (I -
green, II – yellow, III - red). There are example molecules for each rule outcome
(“Yes” or “No”), selectable by radio buttons.

12
T O X T R E E U S E R M A N U A L

Figure 5: Decision tree visualisation

Decision tree options visualisation

A new menu item, “Method► Decision Tree Options” ha s been introduced since
Toxtree v1.40. The options dialog contains several sections, of which “General” is
common for all decision trees, and “Rules” is specific to the currently loaded decision
tree, and may be missing, if no rule specific options are available (as for Cramer rules).

13
T O X T R E E U S E R M A N U A L

Figure 6: Decision tree options

• Default directory: Remembers the directory of the last opened/saved file

o Default: Empty

• Show circle in an aromatic ring: Toggles displaying aromatic rings

o Default: Checked

• Generate 2d coordinates if none exist: Generate 2D coordinates of the

structures, entered as SMILES

o Default: Checked

• Use Openbabel SMILES parser: Toggles usage of Openbabel9 vs. CDK

SMILES parser.

o Default: Checked

9 http://openbabel.org/

14
T O X T R E E U S E R M A N U A L

• Path to OpenBabel (Windows): Specifies the path to Openbabel on

Windows platform.

o Default: helper/openbabel/win/babel.exe

• Path to OpenBabel (Linux): Specifies the path to OpenBabel on Linux

platform.

o Default: helper/openbabel/linux/babel

• MOPAC directory: Directory where MOPAC resides

o Default: helper

• MOPAC executable: Name of the MOPAC executable. Used to calculate

electronic descriptors as eHOMO/eLUMO, required by some plug-ins.

o Default: MOPAC_7.1.exe10

• mengine (3D builder - Windows): MMFF94 force field by mengine11.

Structures without 3D coordinates are submitted to mengine before running
MOPAC. Specifies the path to mengine on Windows platform.

o Default: helper/smi23d/win/mengine.exe

• mengine (3D builder - Linux): MMFF94 force field by mengine. Structures

without 3D coordinates are submitted to mengine before running MOPAC.
Specifies the path to mengine on Linux platform.

o Default: helper/smi23d/linux/mengine

• smi2sdf (Windows) (used by mengine): Generates rough 3D structure12.

Preparatory step before running mengine. Specifies the path to smi2sdf on
Windows platform.

o Default: helper/smi23d/win/smi2sdf.exe

10 OpenMopac, http://openmopac.net/

11 mengine, http://www.chembiogrid.org/cheminfo/smi23d/

12 smi2sdf, http://www.chembiogrid.org/cheminfo/smi23d/

15
T O X T R E E U S E R M A N U A L

• smi2sdf (Linux) (used by mengine): Generates rough 3D structure.

Preparatory step before running mengine. Specifies the path to smi2sdf on
Windows platform.

o Default: helper/linux/win/smi2sdf.exe

These general options are automatically saved in the toxtree.pref file, located in the
same directory as Toxtree-X.YZ.jar.

Decision tree selection

The “Method ► Select decision tree” menu can be used in order to change the active
decision tree, as shown on Figure 7. The Load from file button enables users to select a
different decision tree, which was written from scratch or by editing an existing
decision tree.

16
T O X T R E E U S E R M A N U A L

Figure 7: Decision tree selection

When the “Verhaar scheme” option is selected, the Verhaar scheme for predicting
toxicity mode of actions is loaded. The tree consists of 5 classes, as shown on Figure 8.

17
T O X T R E E U S E R M A N U A L

Figure 8: Verhaar scheme

All of the functionality explained for the “Cramer rules” decision tree is valid for any
other tree selected.

The “Skin irritation/corrosion” decision tree for estimating skin irritation and
corrosion potential by physicochemical property limits and structural rules can be
loaded by selecting the corresponding line from the decision tree selection dialog
(Figure 7). The result of this operation is displayed in Figure 9.

18
T O X T R E E U S E R M A N U A L

Figure 9: Skin irritation prediction

For more details about skin irritation prediction, please consult the Skin
irritation/corrosion rules specifics section of the manual.

The “Eye irritation and corrosion” decision tree for estimating eye irritation and
corrosion potential by physicochemical property limits and structural rules can be
loaded by selecting the corresponding line from the decision tree selection dialog
(Figure 7). For more details about eye irritation prediction, please consult the Skin
irritation/corrosion rules specifics section of the manual.

Selecting the “Benigni / Bossa rulebase (for mutagenicity and carcinogenicity)” option
loads the corresponding decision tree as displayed in Figure 10.

19
T O X T R E E U S E R M A N U A L

Figure 10: Mutagenicity and carcinogenicity prediction

For more details about mutagenicity and carcinogenicity prediction, please consult the
Benigni / Bossa rulebase (for mutagenicity and carcinogenicity) specifics
section of the manual.

Decision tree editing

The “Method► Edit decision tree” menu can be used i n order to edit an existing
decision tree or construct a new one from scratch (Figure 11). It provides access to the
following 3 submenus:

 the “New decision tree” submenu can be used for constructing a new decision
tree from scratch;

20
T O X T R E E U S E R M A N U A L

 the “Select from list” submenu provides a list of known decision trees, which
can be loaded and subsequently edited in the “Decision tree editor”;

 the “Load from file” submenu can be used for loading a user-supplied decision
tree in the “Decision tree editor”.

A copy of the chosen decision tree is loaded in memory for editing and can be
subsequently saved for future use.

You should always try to give a suitable (self-explaining) name, when saving
newly constructed from scratch or modified decision trees.

Figure 11: Edit decision tree submenus

The “Decision tree editor” is loaded in a separate (child) window (Figure 12) and
provides tools for decision tree editing, as suggested by its title.

21
T O X T R E E U S E R M A N U A L

The Decision tree editor comprises several areas:

 Decision tree area – placing the mouse on a tree node and right-clicking on it
provides access to a context menu. This menu can be used for rule editing, as
well as for modifying the “Yes” and/or “No” branches coming out of the
current node (highlighted in orange);

 Rules area – provides means to add new (supported by Toxtree itself or by an

added plug-in) rules, as well as to edit existing rules;

 Categories area – add/remove supported categories (classes);

 Decision node area – displays details about the currently selected node from
the Decision tree area. Left-clicking on a node in the Decision tree area
changes the current node.

Before exiting from the Decision tree editor, users should save the edited decision tree
by using the “File► Save” menu on the upper left corner of the main Decision tree
editor window. A reminder is displayed if the user tries to exit the Decision tree editor
without having saved his work.

22
T O X T R E E U S E R M A N U A L

Figure 12: Decision tree editor on a copy of Cramer rules

Detailed instructions how to create and edit a decision tree are provided in section
“Decision tree editing - typical tasks”.

Splitting the file into groups

Toxtree allows to split the file into subsets, defined by decision tree categories. For this
purpose:

1. Load a file with chemical compounds (the following examples use the
DSSTox EPA Fathead Minnow file
EPAFHM_v3b_617_10Apr2006.sdf, downloaded from
http://www.epa.gov/ncct/dsstox/sdf_epafhm.html).

2. Run “Toxic Hazard► Estimate All” to apply Cramer rules for all
compounds.

3. Click on “Chemical Compounds ► Subsets” to display the subsets

selection dialog.

23
T O X T R E E U S E R M A N U A L

Figure 13: Subsets

The subsets selection dialog (Figure 13) displays three subsets, corresponding to
Cramer toxicity classes, and the number of compounds in each subset. Select the
second row (Intermediate (Class II)) and click OK. The subset of 22 compounds will
be loaded into Toxtree main window (Figure 14).

24
T O X T R E E U S E R M A N U A L

Figure 14: 22 compounds classified as Cramer class II

From this point on, all actions (e.g. File Save, Estimate All) are applied on the subset
loaded in the main window. Use the same menu to select another subset or to return to
the entire file. For the latest, select “All” options and then click OK (Figure 15).

25
T O X T R E E U S E R M A N U A L

Figure 15: Select “All” option to load the entire file into the main Toxtree window

Structure diagram editor

A 2D structure diagram editor is integrated in Toxtree. It can be accessed through the
“Chemical Compounds ► Edit compound” menu (Figure 16).

The structure diagram editor provides a convenient way to edit the current compound
in Toxtree.

If needed, users could consult the structure diagram editor help and/or tutorial,
accessible through the “Help” menu of the editor.

26
T O X T R E E U S E R M A N U A L

Figure 16: 2D structure diagram editor

Batch processing
When dealing with large datasets (more than 1000 molecules), Toxtree should be used
in batch processing mode. It can be accessed through the “File► Batch processing”
menu. Users are invited to select the input and output files (supported types are CSV,
TXT and SDF), before starting the batch processing. The batch can be paused or
aborted upon user request. Batch configuration/state can be saved and loaded at a later
time (Figure 17). In case of unexpected interruption (e.g. power failure, hardware
failure, operating system failure, etc…), the batch job can be continued from the last
previously fully processed record. In order to achieve this, the interrupted batch
configuration should be loaded (either from a user-specified file, or from the system
TEMP folder, where it is automatically stored if the user has not specified a file).

27
T O X T R E E U S E R M A N U A L

Figure 17: Batch processing

A typical usage scenario

Typical use of Toxtree could follow the following scenario:

1. launch the application;

2. enter a molecular structure in the SMILES field (or open a file, containing
molecular structures - compatible file types are: CML, CSV, HIN, ICHI,
INCHI, MDL MOL, MDL SDF, MOL2, PDB, SMI, TXT and XYZ). Please,
note that CSV files can be read/written by MS Excel ™ ;

3. press "Estimate" (on the right upper part of the application window);

4. read the toxic hazard classification of the structure by application of Cramer

rules (either low - class I, intermediate - class II or high - class III);

5. read the classification explanation (which Cramer rules have been applied);

6. go through the list of structures, loaded from a compatible file type (CML,
CSV, HIN, ICHI, INCHI, MDL MOL, MDL SDF, MOL2, PDB, SMI, TXT
and XYZ) by using the navigation links at the bottom of the left side of the
application window (note: CSV files can be read/written by MS Excel ™);

28
T O X T R E E U S E R M A N U A L

7. repeat steps from (3) to (5) for any chosen (in step (6)) molecular structure;

8. consult the decision tree (accessible through the Toxtree "Method ► View
decision tree" menu);

9. classify all the molecules, loaded from a file, by using the batch processing
facility;

10. save the processed molecules, together with classification data (class & path) in
a file (compatible types are CSV, SDF & TXT). Please, note that CSV files can
be read/written by MS Excel ™;

11. create (or edit existing) decision trees through the Toxtree "Method ► Edit
decision tree" menu;

12. change the decision tree used in the estimation through the Toxtree
"Method ► Select decision tree" menu;

13. modify the current molecule by using the integrated structure diagram editor
through the “Chemical Compounds ► Edit compound” menu.

Command line options

Toxtree supports the following command line options:

 -b <file name> specifies that <file name>should be

opened for batch processing;

 -DtoxTree.debugging=true turns on verbose console output (useful

for debugging purposes, as well as for detailed study of decision tree results),
default value - off;

 -f <file name> specifies that <file name>should be

opened for browsing.

The verbose console output could be redirected to a log file, by using the following
command:

java -DtoxTree.debugging=true -jar Toxtree-X.YZ.jar > Toxtree.log

29
T O X T R E E U S E R M A N U A L

Cramer rules specifics

Lists of compounds

Cramer rules # 1 and # 22 depend explicitly on user-defined lists of compounds, which

are normal constituents of the body or common components of food. We provide
example lists of such compounds in the files bodymol.sdf and foodmol.sdf
respectively. If the files are removed from the application directory, than a message
stating that Q1 and/or Q22 are not implemented is displayed. This is normal
behaviour and a reminder, that Q1 and Q22 cannot be implemented by structural
rules.

You can check if the files bodymol.sdf and foodmol.sdf are found by Toxtree by using
the “Help► Files info” menu ( Figure 18).

Figure 18: Files information (bodymol.sdf & foodmol.sdf found and used by Toxtree)

If bodymol.sdf and/or foodmol.sdf are not found by Toxtree, the “Files information”
box reflects this, as shown on Figure 19.

Figure 19: Files information (bodymol.sdf & foodmol.sdf not found by Toxtree)

Please, note that the bodymol.sdf and foodmol.sdf files are provided mainly
as an example. They contain currently only a very limited number of
“Normal constituents of the body” and “Common components of food”
respectively, following an expert advice. Users should consider expanding
these files with appropriate molecules.

30
T O X T R E E U S E R M A N U A L

Hydrolysis/metabolic reactions

A limited number of hydrolysis (Cramer rules # 15, # 17, # 29, # 30 and # 31) and
metabolic (Cramer rule # 33) reactions are implemented, based on an expert advice.
The reactions are stored and read as CML files.

The implemented six hydrolysis reactions are shown on the following figures.

Figure 20: Reaction-hydrolize-S(=O)(=O)O

Figure 21: Reaction-hydrolize-ester

Figure 22: Reaction-hydrolize-C(=S)O

31
T O X T R E E U S E R M A N U A L

Figure 23: Reaction-hydrolize-C(=O)S

Figure 24: Reaction-hydrolize-C(=S)S

Figure 25: Reaction-hydrolize-P(=O)(OR)(OR)OR

The implemented four metabolic reactions are show on the following figures.

Figure 26: Reaction-metabolize-R-N

32
T O X T R E E U S E R M A N U A L

Figure 27: Reaction-metabolize-C=C

Figure 28: Reaction-metabolize-N+=

Figure 29: Reaction-metabolize-N=N

Verhaar scheme specifics

The Verhaar scheme classifies compounds in the following categories:

Class 1 (narcosis or baseline toxicity)

Class 2 (less inert compounds)

Class 3 (unspecific reactivity)

33
T O X T R E E U S E R M A N U A L

Class 4 (compounds and groups of compounds acting by a specific

mechanism)

Class 5 (Not possible to classify according to these rules)

The second rule verifies whether the LogP of a compound is within the [0,6] range.
LogP is calculated on the fly through the XlogP procedure13,14, implemented by the
CDK library. The implementation was validated by comparison with other LogP
implementations15.

The Verhaar scheme does not define structural rules for Class 4 compounds, but only
examples. The implementation includes only a limited set of examples.

Skin irritation/corrosion rules

specifics
The skin irritation/corrosion rules classify compounds into following categories:

• Not Corrosive

• Not Irritating Or Corrosive

• Not Irritating

• Irritating

• Corrosive

• Irritating Or Corrosive

• Unknown

The classification is done by physicochemical property limits and structural rules.

13R. Wang, Y. Fu, and L. Lai. A New Atom-Additive Method for Calculating Partition Coefficients. J. Chem.
Inf. Comput. Sci., 37:615–621, 1997.

R. Wang, Y. Gao, and L. Lai. Calculating partition coefficient by atom-additive method. Perspectives in
14

Drug Discovery and Design, 19:47–66, 2000.

15 Uli Fechner, Kristina Grabowski, QA of the XlogP Descriptor, CDK News - The Newsletter of CDK

project, Volume 3/1, March 2004, ISBN 1614-7553, pp. 12-14, http://wiki.cubic.uni-
koeln.de/cdknews/index.php/CDKNews

34
T O X T R E E U S E R M A N U A L

Physicochemical properties

The decision rule depends on the following physicochemical properties: molecular

weight, LogP, melting point, water solubility, lipid solubility and surface tension.

The Toxtree software calculates on the fly only LogP and Molecular weight, for the
rest it expects the values to be read from the file or manually entered by the user.

The file should contain properties with exactly the following names (column names in
CSV file of SDF properties in SDF file):

“Vapour Pressure”

“Water Solubility”

“Lipid Solubility”

“Melting Point”

“Surface Tension”

If fields with exactly these names are missing from the file, the software shows a dialog,
asking properties to be entered manually for each compound (Figure 30).

35
T O X T R E E U S E R M A N U A L

Figure 30: Skin irritation prediction - options

If the values are unknown, you might select the checkbox “Silent”. In this case, the
result of a silent rule asking for missing properties will always be “No” and therefore
the left branch of the tree will be followed. This effectively means that physicochemical
rules will be skipped and only structural rules will be applied. Please note that this may
result in a low quality prediction.

36
T O X T R E E U S E R M A N U A L

Eye irritation/corrosion rules

specifics
The eye irritation/corrosion caused by a chemical is characterized using the following
EU risk phrases:

R36: Irritating to eyes (moderate eye irritation reversible within ≤ 21 days)

R41: Risk of serious damage to eyes (moderate but persistent eye lesions, eye
corrosion)

According to international risk assessment guidelines, skin corrosion potential excludes

further considerations on a similar hazardous potential to eyes, since the chemical has
already proved to have corrosive properties. Thus the following effects induced by
local contact to skin are also assumed to be predictive of eye damage:

R34: Causes burns (skin corrosion caused by a 4-hour skin contact)

R35: Causes severe burns (skin corrosion caused by a 3-minute skin contact)

Based on this, the eye irritation/corrosion rules implemented in Toxtree classify

compounds into the following categories:

• NOT skin corrosion R34 or R35 category 1

• NOT lesions R34, R35, R36 or R41 category 2

• NOT eye irritation R41 category 3

• NOT eye irritation R36 category 4

• NOT corrosion R34, R35 or R41 category 5

• NOT lesions R34, R35 or R36 category 6

• NOT eye irritation R36 or R41 category 7

• Serious lesions to the eye R41 category 8

• Moderate reversible irritation to the eye R36 category 9

• Skin corrosion R34 or R35 category 10

37
T O X T R E E U S E R M A N U A L

• Unknown category 11

Similarly to the skin irritation/corrosion rules, the classification here is done by

physicochemical property exclusion rules and structural inclusion rules.
Physicochemical exclusion rules are used to identify chemicals with no skin
irritation/corrosion or eye irritation/corrosion potential and structural inclusion rules
are used to identify chemicals with skin irritation/corrosion or eye irritation/corrosion
potential.

Physicochemical exclusion rules

The decision rules depend on the following physicochemical properties: molecular

weight, octanol-water partition coefficient LogP, melting point, aqueous solubility, lipid
solubility (Table 2). The rules are valid for all groups of chemicals or are specific for the
chemicals from a given chemical class (Table 1).
Table 1: Chemical classes for which specific rules are defined

Class (designation) Empirical Formula

C CxHyOz

CN CxHyOzNa

CNHal CxHyOzNaHalb

CNS CxHyOzNaSb

CHal CxHyOzHalb

Table 2: Physicochemical exclusion rules for eye irritation/corrosion as implemented in Toxtree

RuleID Group IF parameter Qualifier Value Unit Category

a
1 All m.p. > 200 °C 1
2 All logP > 9 2
3 All logP < -3.1 1
b
4 All l.s. < 0.01 g/kg 1
c
5 All a.s. < 0.000005 g/l 4
6 All a.s. < 0.00002 g/l 3
d
7 All m.w. > 650 g/mol 4
8.1 C m.p. > 55 °C 1
8.2 C m.w. > 380 g/mol 2
8.3 C a.s. < 0.0005 g/l 7
8.4 C a.s. < 0.0001 g/l 2
9.1 CN l.s. < 0.4 g/kg 1
9.2 CN m.w. > 290 g/mol 1
9.3 CN a.s. < 0.1 g/l 1
9.4 CN logP > 4.5 1
10.1 CNHal logP > 3.8 5

38
T O X T R E E U S E R M A N U A L

RuleID Group IF parameter Qualifier Value Unit Category

10.2 CNHal a.s. < 0.1 g/l 1
10.3 CNHal m.w. > 370 g/mol 1
10.4 CNHal l.s. < 400 g/kg 1
10.5 CNHal a.s. < 0.004 g/l 3
11.1 CNS m.w. > 620 g/mol 6
11.2 CNS m.p. > 200 °C 4
11.3 CNS m.p. > 50 °C 1
11.4 CNS log P < -2 1
11.5 CNS log P > 1.5 4
11.6 CNS log P > 3.6 3
11.7 CNS a.s. < 0.006 g/l 7
12.1 CHal m.w. > 370 g/mol 6
12.2 CHal m.w. > 280 g/mol 1
12.3 CHal m.p. > 65 °C 1
12.4 CHal logP > 4.5 7

a
m.p. – melting point; b l.s. – lipid solubility; c a.s. – aqueous solubility; d m.w. – molecular weight

The Toxtree software calculates on the fly only LogP and molecular weight, for the
rest it expects to be read from the file or manually entered by the user similarly to skin
irritation/corrosion rules.

If the values are unknown, you might select the checkbox “Silent”. In this case, the
physicochemical rules will be skipped and only the structural rules will be applied.
Please note that this may result in a low quality prediction.

Structural inclusion rules

The decision rules depend on the structural inclusion rules, given in Table 3, Table 4
and Table 5.

Table 3: Structural inclusion rules for predicting serious local lesions to the eye as implemented in Toxtree
(category 8)

RuleID Structural alert Chemical class Limits

R = aliphatic
chain
R1 R1,2 = H or
aliphatic
R OH Aliphatic
13 chain
monoalcohols
C3-C11 - eye
R2 damage
C12-C14 -
eye irritation
OH
Aliphatic glycerol R = aliphatic
14 R O CH2 HC
CH2OH monoethers chain

39
T O X T R E E U S E R M A N U A L

RuleID Structural alert Chemical class Limits

Hal O
R1-4 = H,
R1
Derivatives of 2- aliphatic
OH
halogen benzoic acids chain or
15
and corresponding halogen
R2 R4 alkali salts Hal = F, Cl or
Br
R3
O R1 = H or
Hal O CH Halogen benzenes halogen
with substituents R2 = aliphatic
16 OH
R2 containing carboxylic chain
acid groups Hal = F, Cl or
R1 Br
O
Aliphatic esters of R = aliphatic
17 Cl
chloro formic acid chain
O R
Cl R4 R1 = aliphatic
chain
Chlorinated aliphatic
18 R1 CH2OH R2-4 = H or
alcohols
R2 R3
aliphatic
chain
+
I
R1 = aliphatic
Diphenyl iodonium
19 chain
salts
R2 = any
R1 R2
R1 R1 = H or
CH N Derivatives of alpha aliphatic
20
amino benzene chain
R3 R2 R2,3 = any

R = H or
21 R N O Pyrrolidones aliphatic
chain
R
R
R=H or OH
R1=H or
R2
22 Substituted indoles aliphatic
N ketone
R
R2= any
R1
R R R R

N R N R
N R = H, NH2
N or aliphatic
23 H Substituted pyrazoles
chain
R1 R1 R1 = any

R1
a) b)

40
T O X T R E E U S E R M A N U A L

RuleID Structural alert Chemical class Limits

R1
+ +
NH3 N R2 R,R3=any; if
containing
R3 halogen, thio-
a) b) or sulpho-
Aromatic ammonium
24 groups then
salts
R R1 R R1 R41
R1,2 = H or
C N+ R2 C N+ R2 aliphatic
chain
R R3 R R3

c) d)
R = any
R SO 3H Organic sulphonic
25 (aliphatic or
salts
aryl group)
R R R2
N N

R S R1 R R1 R = any
S
R1 = H or -C-
a) b)
Thiazoles and any
26 R2
thiazolidines R2 = H or
N N aliphatic
chain
R1 R1
S S
c) d)
R
S S
N R
N R = aliphatic
27 R Thiazolones
chain

O O
a) b)

+ Triphenylphosphonium
28 P CH2 R R = any
salts

O R = aryl or
Organic phosphinic aliphatic
29 R P R1 acids and their chain
derivatives R1 = H or -
OH CH2-any

41
T O X T R E E U S E R M A N U A L

Table 4: Structural inclusion rules for predicting moderate reversible irritation to the eye as implemented in Toxtree
(category 9)

RuleID Alert Chemical class Limits

R=aliphatic
chain
R1 R1,2=H or
aliphatic chain
R C OH Aliphatic
30 C12-C14 - eye
monoalcohols
irritation
R2 > C15 - only
slight eye
irritation
OH O R,R2 = aliphatic
Aliphatic alpha chain;
31 R C C
hydroxyesters R1 = H or
O CH2 R2
R1 aliphatic chain
R1-3=aliphatic
chains with or
without further
R2 O oxygen
Aliphatic containing
32 R1 C C carboxylic substituents;
OH acids C6-C14 - strong
R3 eye irritation
> C20 - only
slight eye
irritation

R1
+ +
NH3 N R2

R3 R,R3=any; if
a) b) containing
R R1
halogen, thio-
Aromatic
or sulpho
33 ammonium
C N +
R2 groups then
salts
R41
R R3 R1,2=H or
R R R1 aliphatic chain

C C N+ R2

R R R3
c) d)

42
T O X T R E E U S E R M A N U A L

Table 5: Structural inclusion rules for predicting skin corrosion as implemented in Toxtree (category 10)

RuleID Alert Chemical class Limits

Hal R1

C R2
O
Substituted benzoic acid Hal = Cl or F
34
halides R1-4 = any
R3

R4
O C N CH2 R
a)
35 Aliphatic iso(thio)cyanates R=aliphatic chain
S C N CH2 R
b)
R2
R1
R1-3 = any (e.g., further halogen)
36 Si Chlorosilanes
R3
Cl
H3C
O
37 H3C Si O C Mixed oxy-carboxysilanes R1,2 = any
R1
O R2
R1
R = aliphatic chain which may
38 R N Aliphatic amines contain ether functions
R2
R1,2 = H or aliphatic chain
R1
- + Alkali salts of aliphatic R1 = H or aliphatic chain
39 CH O Li
alcohols R2 = aliphatic chain
R2

Benigni / Bossa rulebase (for

mutagenicity and carcinogenicity)
specifics
The processing of a query chemical can give rise to a limited number of different
outcomes, namely: a) no presence of SAs for carcinogenicity; b) one or more SAs are
recognized; c) SAs relative to aromatic amines or αβ -unsaturated aldehydes are
recognized, and the chemical goes through QSAR analysis, which may result in a
negative or positive outcome. The system flags either outcome through one, or a
combination of several labels, as follows:

43
T O X T R E E U S E R M A N U A L

• No alerts for carcinogenic activity – no SAs have been recognized by the

system;

• Structural Alert for genotoxic carcinogenicity – the system recognizes the

presence of one or more SAs, and specifies a genotoxic mechanism;

• Structural Alert for nongenotoxic carcinogenicity – the system recognizes

the presence of one or more SAs, and specifies a nongenotoxic mechanism;

• Potential S. typhimurium TA100 mutagen based on QSAR – assigned

according to the output of QSAR6 or QSAR13;

• Unlikely to be a S. typhimurium TA100 mutagen based on QSAR –

assigned according to the output of QSAR6 or QSAR13;

• Potential carcinogen based on QSAR – assigned according to the output of

QSAR8 (aromatic amines);

• Unlikely to be a carcinogen based on QSAR – assigned according to the

output of QSAR8 (aromatic amines);

• For a better assessment a QSAR calculation could be applied – assigned

when one of QSAR6, QSAR8 or QSAR13 is applicable, but the user chooses
not to apply a QSAR.

If the query chemical belongs to the classes of aromatic amines or αβ -unsaturated

aldehydes, the appropriate QSAR is applied. A QSAR provides a more refined
assessment than SAs, and should be given higher importance in a weight-of-evidence
scheme. Thus, a QSAR analysis might point to an estimated lack of toxic effects, in
spite of the presence of SAs.

All molecular descriptors for the QSARs are calculated on the fly by the software.
LogP is calculated through the XlogP procedure16,17, implemented by the CDK library.
The implementation has been validated by comparison with other LogP
implementations18. EHOMO and ELUMO are calculated by launching OpenMopac

16 R. Wang, Y. Fu, and L. Lai. A New Atom-Additive Method for Calculating Partition Coefficients. J. Chem.

Inf. Comput. Sci., 37:615–621, 1997.

R. Wang, Y. Gao, and L. Lai. Calculating partition coefficient by atom-additive method. Perspectives in
17

Drug Discovery and Design, 19:47–66, 2000.

18 Uli Fechner, Kristina Grabowski, QA of the XlogP Descriptor, CDK News - The Newsletter of CDK

project, Volume 3/1, March 2004, ISBN 1614-7553, pp. 12-14, http://wiki.cubic.uni-
koeln.de/cdknews/index.php/CDKNews

44
T O X T R E E U S E R M A N U A L

7.119. An additional validation step was performed by comparing the results with the
original values presented by the authors of the QSARs.

In order to reproduce the values of the descriptors present in the original papers, the
following correction factors are applied to the calculated values (the corrected values
are used for predictions):

QSAR13

MR QSAR13 = 0.8718 * MR – 2.3452

LogP QSAR13 = 0.99738 * LogP – 0.10589

ELUMO QSAR13 = 1.07907 * ELUMO – 0.01463

QSAR6

EHOMOQSAR6 = 1.03383 * EHOMO + 0.30348

ELUMO QSAR6= 0.98963 * ELUMO – 0.04037

QSAR8

EHOMO QSAR8 =0.88239 * EHOMO – 1.0381

ELUMO QSAR8 =0.96239 * ELUMO – 0.01521

QSAR calculations can be time consuming due to the requirement to calculate

electronic descriptors. The software provides options to skip the QSAR calculation,
and in this case will assign the category “For a better assessment a QSAR
calculation could be applied”. Toxtree will show the dialogs at Figure 31 or Figure
32, if a QSAR is to be calculated.

19 Available at http://openmopac.net/Downloads/MOPAC_7.1executable.zip

45
T O X T R E E U S E R M A N U A L

Figure 31: Options for QSAR6 and QSAR8 calculations

Figure 32: Options for QSAR13 calculations

The available options have the following semantics:

• Yes: The QSAR will be calculated for the current chemical, and the question
dialog will appear for each subsequent chemical

46
T O X T R E E U S E R M A N U A L

• No: The QSAR will not be calculated for the current chemical, and the
question dialog will appear for each subsequent chemical

• Yes to all: The QSAR will be calculated for the current chemical and for all
subsequently processed chemicals, for which it is applicable. The question
dialog will not appear anymore, unless the option is changed via the
“Method ► Decision Tree Options” menu.

• No to all: The QSAR will be not calculated for the current chemical and for
all subsequently processed chemicals, for which it is applicable. The question
dialog will not appear anymore, unless the option is changed via the
“Method ► Decision Tree Options” menu.

The rule options can also be accessed through the Rules section of the
“Method► Decision Tree Options ” menu (Figure 33).

Figure 33: Decision tree specific options

47
T O X T R E E U S E R M A N U A L

Decision tree editing - typical tasks

Creating a simple decision tree

1. Use the “Method ► Edit Decision tree ► New Decision Tree” menu option to
create a new (empty) decision tree. This will launch the decision tree editor, as
shown on Figure 34.

Figure 34: Decision tree editor on empty decision tree

2. Click the “Decision rules ► New Rule” menu option to create a new rule. This
will launch the decision rule wizard, as shown on Figure 35.

48
T O X T R E E U S E R M A N U A L

Figure 35: Decision tree editor – new rule wizard

3. Click on the “Aromatic” option to create a rule that will verify if the substance
is aromatic. Then click OK. The next page of the wizard will appear, as shown
on Figure 36.

49
T O X T R E E U S E R M A N U A L

Figure 36: Decision tree editor – aromatic rule editor

4. Click on the “Rule ID” field, and type in “Q1”. This will be used as a rule
identifier. You could also modify the rule’s title, explanation and examples.
When finished, click OK. The rule will be added as a top node of the tree as
shown on Figure 37. Clicking on the rule will highlight it with orange colour
and will display decision node details on the left and decision rule details at the
bottom.

50
T O X T R E E U S E R M A N U A L

Figure 37: Decision tree editor – new rule added as a top node

5. Click on the “Categories” tab. This will show a list of categories used so far in
the tree. Click on “Default class 2” (second row), as shown on Figure 38.

51
T O X T R E E U S E R M A N U A L

Figure 38: Decision tree editor – modifying a category

6. Then click on “Categories ►E dit category” menu option on the left. This will
launch the category editor, as shown on Figure 39. Type in “Not aromatic” as
category title and similar text in the explanation field. Click OK when ready.

52
T O X T R E E U S E R M A N U A L

Figure 39: Decision tree editor – category editor

7. Right click on the decision node in order to invoke a popup menu, as shown
on Figure 40. These menu options allow editing the rule and modifying the left
and right branches. The left branch is followed when the answer of the rule is
“No” and the right branch is followed if the answer is “Yes”.

Figure 40: Decision tree editor – popup menu

8. Proceed with modifying the right branch, which in the current tree will be
followed when the structure is aromatic. Click on the “Modify <YES>

53
T O X T R E E U S E R M A N U A L

branch” menu option. This will launch a wizard, as shown on Figure 41. Select
the “Rule” option and click “Next”.

Figure 41: Decision tree editor – branch wizard

9. The next wizard page shows the same options as in Figure 35. Click on the
“Heteroaromatic” option and follow the wizard’s instructions. At the end a
new node will be added to the tree, as shown on Figure 42.

54
T O X T R E E U S E R M A N U A L

Figure 42: Decision tree editor – decision tree with two nodes

10. Proceed with modifying the left branch, which in the current tree will be
followed when the structure is aromatic but not heteroaromatic. Right click on
the new node and invoke the “Modify <NO>branch” menu option. The
same wizard as in Figure 41 will appear. Select the “Category” option and click
“Next”.

Figure 43: Decision tree editor – category options

55
T O X T R E E U S E R M A N U A L

11. Select the “Create new category” from the three options, shown on Figure 43,
and click “Next”. A category editor will appear, as in Figure 39. Type in
“Aromatic, but not heteroaromatic” as a title and “The structure is not
heteroaromatic” in the explanation field. Then click OK. This will change the
left branch of the second node.

12. Finally, we will modify the result that will be assigned if the structure is
heteroaromatic. Right click on the second (Heteroaromatic) node and invoke
the “Modify <YES>branch” menu option. The same wizard as in Figure 41
will appear. Select the “Category” option and click “Next”.

13. From the category wizard (Figure 43) select the first option – “Select from
categories used in this tree” and click “Next”. The next page displays the
categories used in the tree. Select “Default class 1”, as shown on Figure 44, and
follow the wizard’s instructions.

Figure 44: Decision tree editor – list of categories used in the tree

The next page will be a category editor as in Figure 39. Type in “Heteroaromatic” in the
title field. Click OK when ready. The resulting tree should look like as shown on Figure
45 (2 nodes, 3 categories).

56
T O X T R E E U S E R M A N U A L

Figure 45: Decision tree editor – the resulting tree (2 nodes, 3 categories)

14. Use the “Decision Tree ► Modify tree caption” menu option on the left. This
will ask you to enter the tree’s title. Type in “Aromatic substances” and click
OK.

15. Save the file as *.tree or *.tml file and close the decision editor. You might then
load the tree by “Method ► Select Decision Tree” menu option Figure 46.

The *.tree file is a binary file and normally .tree files are incompatible
between releases. A new XML based format *.tml was introduced, in order to
avoid incompatibility of .tree files. The *.tml file is the preferred format,
while *.tree format is likely to be abandoned in subsequent releases.

57
T O X T R E E U S E R M A N U A L

Figure 46: Loading the new decision tree from a file

16. The decision tree will be loaded as a default decision method in Toxtree. The
classification area will reflect the categories, defined in the tree (Figure 47).

58
T O X T R E E U S E R M A N U A L

Figure 47: The new “Aromatic substances” decision tree displayed on the main Toxtree screen

Using a decision tree

1. Apply the decision tree, created in the previous section, to the default structure
(hexane) by clicking on the “Estimate” button. This applies the decision tree to
the current structure. The result says that the compound is not aromatic, as
shown on Figure 48. The result is also assigned as a property of the compound
and will be saved along with other properties when using the “File► Save”
menu option.

59
T O X T R E E U S E R M A N U A L

Figure 48: The result of applying the new tree to Hexane

2. Click on the “Chemical compounds ► Edit compound” menu option. This

will launch the JChemPaint20 structure diagram editor and allows editing the
current structure. Add a benzene ring to the chain, as shown on Figure 49.
Click OK when ready. This will update the structure, displayed on the main
screen.

20 More information about JChemPaint can be found at http://almost.cubic.uni–koeln.de/cdk/jcp

60
T O X T R E E U S E R M A N U A L

Figure 49: JChempaint structure diagram editor

3. Click again on the “Estimate” button. This will apply the decision tree to the
new structure. The result is shown on Figure 50.

61
T O X T R E E U S E R M A N U A L

Figure 50: The result of applying the new tree to the new structure

62