oecd qsar toolbox v.4 is a step-by-step presentation designed to take the toolbox user through the...

OECD QSAR Toolbox v.4.0

Tutorial on how to predict Skin sensitization potential taking into account alert performance

• Background

• Objectives

• Specific Aims

• Read across and analogue approach

• The exercise

• Workflow

Outlook

The OECD QSAR Toolbox for Grouping Chemicals into Categories 2April, 2017

• This is a step-by-step presentation designed to take the Toolbox user through the workflow of a data filling exercise accounting alert performance.

Background


• Background

• Objectives

• Specific Aims

• Alert performance

• The exercise

• Workflow

Outlook


• Define target endpoint;

• Relevancy of profiles and data availability;

• Calculation of alert performance (AP).

Objectives


This presentation demonstrates a number of functionalities of the Toolbox:

• Background

• Objectives

• Specific Aims


• The exercise

• Workflow

Outlook


• To introduce to the Toolbox user the workflow of the definingtarget endpoint;

• To familiarize the user with the new interface of the Toolbox;

• To familiarize the user with the coloring of the profiles anddatabases;

• To familiarize the user with the calculation of alert performance;

• To explain to the Toolbox user the rationale behind each step ofthe exercise.

Specific Aims


• Background

• Objectives

• Specific Aims


• The exercise

• Workflow

Outlook


• Alert performance (AP) is used to define how much relevant to a targetendpoint an alert is;

• AP reflects the alerts usability for category formation;

• AP is applicable for the alerts of the following profilers:

• AP can be calculate for an alert with or without accounting metabolism

Alert performanceOverview


- US-EPA New Chemical categories with respect to Carcinogenicity

- Biodeg probability (Biowin 5) with respect to Biodegradation

- Protein binding alerts for skin sensitization by OASIS with respectto Skin sensitization

- DNA alerts for AMES by OASIS with respect to Ames mutagenicity

• Background

• Objectives

• Specific Aims


• The exercise

• Workflow

Outlook


• In this exercise we will predict the skin sensitization potential foran untested compound, (pyridaphenthion) [CAS# 119-12-0],which will be the “target” chemical.

• We will preliminary define the target endpoint.

• This prediction will be accomplished by calculation of alertperformance.

• The category will be defined by the mechanism of protein bindingcommon to all the chemicals in the category.

• The prediction itself will be made by “read-across”.

The Exercise


• Allergic contact dermatitis that results from skin sensitization isa significant health concern.

• Skin sensitization is a toxicological endpoint that is complexand conceptually difficult.

• However, there is growing agreement that most organicchemicals must react covalently with skin proteins in order tobehave as skin sensitizers.

• Therefore, mechanisms by which organic chemicals bind withproteins are relevant to grouping chemicals that may be skinsensitizing agents.

The OECD QSAR Toolbox for Grouping Chemicals into Categories 12

The ExerciseSidebar On Sensitization

April, 2017

• Background

• Objectives

• Specific Aims


• The exercise

• Workflow

Outlook


• The Toolbox has six modules which are used in asequential workflow:

o Input

o Profiling

o Data

o Category Definition

o Data Gap Filling

o Report

Workflow


• Background

• Objectives

• Specific Aims


• The exercise

• Workflow

o Input

Outlook


• This module provides the user with several means of enteringthe chemical of interest or the target chemical.

• Since all subsequent functions are based on chemical structure,the goal here is to make sure the molecular structure assignedto the target chemical is the correct one.

InputOverview


InputWays of Entering a Chemical

User Alternatives for Chemical ID:

A. Single target chemical

• Chemical Name

• Chemical Abstract Services (CAS) number (#)

• SMILES (simplified molecular information line entry system)notation/InChi

• Drawing chemical structure

• Select from User List/Inventory/Databases

B. Group of chemicals

• User List/Inventory

• Specialized Databases

17The OECD QSAR Toolbox for Grouping Chemicals into Categories April, 2017

Input ScreenInput target chemical by CAS#


1

2

34

Click on CAS# (1); Insert CAS 119-12-0 in the blank field (2) and click on Search (3). When the

structure appears, click on OK (4).

InputDefine target endpoint


Calculation of alert performance (AP) is only possible if the targetendpoint is preliminary selected.Defining of the endpoint allows entering the endpoint of intereste.g. EC3, LC50, gene mutation etc., along with specific metadatainformation. Based on the metadata, different relevancy scores forprofiles could be provided for same endpoint.


When click on Define (1) and the Define target window appears (2). Select Sensitisation in

the Human health hazards category (3) and click on Next (4).


1

2

3

4


On the next step you have to select the endpoint of interest and additionalmetadata if needed.


1. Select EC3 endpoint from the drop-down menu; 2. Add consecutively the following metadata:

Assay: LLNA, Organ: Skin, Type of method: In Vivo. Selection of additional metadata happens by

click on the Add button and selection of the metadata from the drop-down menus; 3. Click on Finish

1

2

2

3



Once the endpoint is defined along with its metadata, theyappear in the endpoint tree and the corresponding row of thedata matrix is highlighted.



The whole defined target endpoint path can be saved and used for other chemicals.To do this, you have to apply right click over the row corresponding to the definedtarget endpoint and to select Target endpoint > Copy.

• Background

• Objectives

• Specific Aims


• The exercise

• Workflow

o Input

o Profiling

Outlook


• “Profiling” refers to the electronic process of retrieving relevantinformation on the target compound, other than environmentalfate, ecotoxicity and toxicity data, which are stored in theToolbox database.

• Available information includes likely mechanism(s) of action, aswell as observed or simulated metabolites.


ProfilingOverview

April, 2017

26The OECD QSAR Toolbox for Grouping Chemicals into Categories

Profiling Sidebar to profiles` relevancy

April, 2017

Once the endpoint is selected, the relevant profiles and metabolictransformations are highlighted.

o Suitable - developed using data/knowledge for the target endpoint; o Plausible – not endpoint specific; structure-based; form broader group of analogues;o Unclassified – all profilers, which are not classified in any of the categories above.


Profiling Profiling the target chemical

April, 2017

1. Select Protein binding alerts for skin sensitization by OASIS v1.4.;

2. Click on Apply

2

1


Profiling Profiling the target chemical

April, 2017

Three protein binding alerts for skin sensitization are found in the target chemical.

Alert 1

Alert 2

Alert 3


Profiling Sidebar on the hierarchical type profiles

April, 2017

Protein binding alerts for skin sensitization by OASIS is a hierarchicalprofile. The organization of the hierarchical profiles includes threelevels of information for each category – domain, mechanistic alertand structural alert.

Domain Mechanistic

alert Structural

alert

Right click over the structural alert and select on Explain opens theprofiling scheme, where the user can see more details about thecurrent alert.

• Background

• Objectives

• Specific Aims


• The exercise

• Workflow

o Input

o Profiling

o Data

Outlook


• “Data” refers to the electronic process of retrieving theenvironmental fate, ecotoxicity and toxicity data that are storedin the Toolbox.

• Data gathering can be executed in a global fashion (i.e.,collecting all data for all endpoints) or on a more narrowlydefined basis (e.g., collecting data for a single or limited numberof endpoints).

DataOverview


Once the endpoint is selected, the databases, which contain such type ofdata, are highlighted.

The more specific the endpoint is, the less databases are highlighted andthe choice is more accurate.

DataSidebar on Data availability


DataGather data


3

2

1. Go to Data; 2. Select Skin sensitization database;

3. Click on Gather.

A pop-up message informs that there is no experimental

data for the target chemical.

1

• Toxicity information on the target chemical is electronicallycollected from the selected dataset(s).

• It should be kept in mind that the search for data and analogues(and therefore calculation of AP) is performed only among thechemicals which are listed in the selected databases. In thisexample only the Skin sensitization database is selected.

• In this example, an insert window appears stating there was “nodata found” for the target chemical.


DataGather data

April, 2017

• Background

• Objectives

• Specific Aims


• The exercise

• Workflow

o Input

o Profiling

o Data

o Category definition

Outlook


• In module one, you have entered the target chemical and definethe target endpoint.

• In the second module, you have profiled the target chemicalwith profiling scheme, which is suitable for the selected targetendpoint.

• In the third module, you have seen the database correspondingto the defined target endpoint. You have found that noexperimental data is currently available in the database for thestructure.

• In other words, you have identified a data gap which you wouldlike to fill.

• Click on “Category Definition” to move to the next module.

Recap


• This module provides the user with several means of groupingchemicals into a toxicologically meaningful category thatincludes the target molecule.

• This is the critical step in the workflow.

• Several options are available in the Toolbox to assist the user inrefining the category definition.

Category DefinitionOverview


• The different grouping methods allow the user to groupchemicals into chemical categories according to differentmeasures of “similarity” so that within a category data gapscan be filled by read-across.

• For example, starting from a target chemical for which aspecific protein binding mechanism is identified, analogues canbe found which can bind by the same mechanism and for whichexperimental results are available.


Category DefinitionGrouping methods

April, 2017

• This is one of the best grouping methods in the Toolbox. It isbuilt on conventional organic chemical reactions and as such isqualitative in character.

• This method is particularly relevant for respiratory and skinsensitization and acute aquatic toxicity, but also forchromosomal aberration and acute inhalation toxicity.


Category DefinitionProtein binding by OASIS v.1.4 grouping method

• This scheme includes 99 categories organized in three level of information:

Level I: Mechanistic Domains (11 categories)

Level II: Mechanistic alerts associated to each mechanistic domain are created on the basis of a common reactive centre being activated by a number of substituents (49 categories)

Level III: A number of structural alerts specifying the substituents to a common reactive centre are made up each mechanistic alert (99 categories)


Category DefinitionSidebar to Protein binding by OASIS v.1.4

categorization

April, 2017

• Each category from level III is presented by defined 2-dimensional structural alerts that is responsible for the elicitingtoxic effects, such as skin sensitization which are a result ofprotein binding.

• The associated chemical reactions are in accordance with existingknowledge on electrophilic interaction mechanisms of variousstructural functionalities.


Category DefinitionSidebar to Protein binding by OASIS categorization

April, 2017

• There is an agreement that most organic chemicals must reactcovalently with skin proteins in order to behave as skinsensitizers.

• Therefore, chemical reactions by which organic chemicals bindwith proteins are relevant to grouping chemicals that may beskin sensitizing agents. So you have mechanistic plausibility fordefining your category based on similar protein-bindingmechanism.


Category DefinitionSidebar to Protein binding by OASIS categorization

April, 2017

• When more than one alert is found in the target structure beforeor after metabolic activation, Alert performance could be used todefine which of them is the most suitable for primarycategorization


Category Definition

April, 2017

The performance of an alert represents the number of chemicals with datarelated to the predefined scale across all chemicals from the selecteddatabases, which have the same alert. It provides and distribution of dataaccording to a given effect (e.g. positive, negative) in percentages.

Alert performanceOverview


N – Number of all chemicals from the selected databases possessing the same alert;n – Number of chemicals out of 22 having data related to the predefined scale;Strongly positive, Weakly positive, Negative – the ratio between chemicals withgiven effect and all chemicals with data related to the scale ( Effect/n*100 )

N n

Category DefinitionCalculation of Alert performance


1. Select the suitable profile for skin sensitization – Protein binding alerts forskin sensitization by OASIS v1.4; 2. Click on Define

2

1



After the click on Define, the Categorization dialog appears. It consists ofall protein binding alerts for SS found in the target structure.

The most suitable alert for category formation is determined bycomparison of their alert performance.

Additional section for calculating “Alert performance” is designed in thisdialogue, when the endpoint is preliminary defined. (see on the nextslide).

Alert performance can be calculate for only one, for combination of alertsor for all found alerts.



alert 1

alert 2

alert 3



In order to calculate the performance of an alert, first of all you have toselect a scale within the Aggregation options.

The main purpose of the scales is to unify all data available in theToolbox databases for a certain endpoint. Therefore, the mostappropriate scale is “Skin Sensitisation II (ECETOC)”. It is a dichotomousscale that converts all skin data into: Positive and Negative.

Additional option for applying different weight of the data that isavailable is also provided. Worst case scenario have been taken intoaccount, i.e. “Maximal” data is set as default.



1

3

2

1. Click on Options; 2. Select Skin sensitization II (ECETOC) scale; 3. Confirm with “OK”

Category DefinitionCalculation of Alert performance for all alerts


1

1. The button “Calculate” become active and we can calculate AP for all found alerts in the

target; 2. “Could not calculate” message appears if chemicals containing all searched alerts

are not found in the selected databases.

Category DefinitionCalculation of Alert performance for one alert


Now we will calculate AP for each of the alerts in order to see which ofthem is the most suitable for category formation.

1. Remove the second and third alerts by double click or using “Down” button; 2. Click on Calculate

1

2



Alert performance window appears. It informs us that if we decide to use thisalert we will obtain:

• 113 analogues with the searched alert;• 54 out of these 113 chemicals posses

the searched data;• 90.74% of the data for all 54

chemicals are positive;• 9.26% of the data for all 54 chemicals

are negative.

Keep in mind that the statistic is obtained from the

chemicals and data, available in the selected databases



Click on the Reset button and repeat the alert performance calculation stepsfor the second and third alert.



In summary we see that the first alert is the most suitable todefine a category.

alert 1

alert 2

alert 3

alert 2

alert 3

alert 1

Category DefinitionSidebar on Alert performance accounting

metabolism


AP can be also calculated for alert(s) identified after AU or SM activation.

No protein binding alerts are

found in the parent structure

Structural alerts are found

after applying of Skin

metabolism simulator

Example: CAS 90-05-1


The target chemical has no alert for protein binding as parent but it isactivated as a result of skin metabolism. In this respect, the primarycategory will be defined with accounting for the metabolic activation.


metabolism


The system will search for chemicals which have similar distribution of theprotein binding alerts as identified in the generated package parent andmetabolites, accounting for the skin metabolism activation for the analogues.


metabolism


The found alerts could be seen by click on the Edit button. Same sequence of actions for APcalculation could be done – remove all alerts except the alert for which AP will be calculated(three levels of mechanistic information are required – domain, mechanistic and structuralalert), select a scale in the Options and click on Calculate.


metabolism

Now let`s go back to our example (see slide 53).

• Based on the defined category (Michael Addition<<Michael addition onconjugated systems with electron withdrawing group<<α,β-Carbonylcompounds with polarized double bonds) 114 analogues have beenidentified (including the target chemical).

• The Toolbox automatically requests the user to select the endpoint thatshould be retrieved.

• The user can either select the specific endpoint or by default choose toretrieve data on all endpoints (see below).


Category DefinitionAnalogues

April, 2017

Only Skin sensitization database is selected now (slide 32) and we click on OK.

• The Toolbox automatically informs the user for the number of gathered data points across the chemicals in the category


Category DefinitionRead data for Analogues

• Click OK to confirm the appeared message

April, 2017

54 chemicals with 87 experimental results related to the defined targetendpoint are found across all 113 analogues.


Category DefinitionSummary information for Analogues

April, 2017

Chemical statistics presenting the number of chemicals and the available experimental data.

• Background

• Objectives

• Specific Aims


• The exercise

• Workflow

o Input

o Profiling

o Data


o Data Gap Filling

Outlook


• You have identified three protein binding alerts for the target chemical(pyridaphenthion).

• You have calculated and compare the alert performance for each of thealerts.

• You have now retrieved in the available experimental results on skinsensitisation (EC3) values for 54 chemicals with the same mechanism ofprotein binding as the target compound, which were found in the “SkinSensitisation” database.

• The user can now proceed to the next module; click on ”Data Gap Filling”.


Recap

April, 2017


Data Gap FillingOverview

April, 2017

• “Data Gap Filling” module give access to five different data gap filling tools:

o Read-across

o Trend analysis

o (Q)SAR models

o Standardized workflow

o Automated workflow

• Depending on the situation, the most relevant data gap mechanism should be chosen, takinginto account the following considerations:

o Read-across is the appropriate data-gap filling method for “qualitative” endpoints like skin sensitisation

or mutagenicity for which a limited number of results are possible (e.g. positive, negative, equivocal).

Furthermore read-across is recommended for “quantitative endpoints” (e.g., 96h-LC50 for fish) if only a

low number of analogues with experimental results are identified.

o Trend analysis is the appropriate data-gap filling method for “quantitative endpoints” (e.g., 96h-LC50 for

fish) if a high number of analogues with experimental results are identified.

o “(Q)SAR models” can be used to fill a data gap if no adequate analogues are found for a target chemical.

o Standardized and Automated workflows are developed to facilitate the users work. Once started, they

follow the implemented logic and finish with prediction. The general differences between the two type of

workflows are represented on the next slide.

In this example we will use the manual read-across approach.


Data Gap FillingApply Read across

April, 2017

2

1. Click on the row with the target endpoint and the cell corresponding to for thetarget chemical; 2. Select Read-across; 3. Select Skin sensitisation II (ECETOC);4. Click on OK

1

3

4



April, 2017

1. Go to Calculation options and click on Data usage; 2. Select the worst case scenario, i.e.

“Maximal” data; 3. Click on OK

1

2

3



April, 2017

1. Go to Select/filter data and click on Subcategorize; 2. Select the US-EPA New Chemical

Categories profiler; 3. Click on Remove selected

1

2

3


Data Gap FillingAccepting the predicted result

April, 2017

1. Click on Accept prediction; 2. Confirm with “Yes”

1

2

• Read-across is the appropriate data-gap filling method for“qualitative” endpoints like skin sensitisation. Since the most ofthe analogues and all five neighbouring tested chemicals in thecategory were positive, it was easy to accepting the prediction ofpositive for the target chemical.

• You are now ready to complete the final module and to create thereport.

• Click on “Report” to proceed to the last module.

Recap


• Background

• Objectives

• Specific Aims


• The exercise

• Workflow

o Chemical Input

o Profiling

o Endpoint


o Data Gap Filling

o Report

Outlook


ReportOverview


• Report module could generate report on any of predictionsperformed with the Toolbox.

• Report module contains predefined report template which userscan customize.

• Two type of report files are generated:

- Prediction report – containing information for the target

- Data matrix – containing information for the analogues usedfor the prediction.


ReportGeneration report

April, 2017

1. Go to the Report module; 2. Click on the cell with the prediction; 3. Click on the Create

button; 4. Click on Create report

1

2

3

4

ReportGeneration report

73The OECD QSAR Toolbox for Grouping Chemicals into Categories April, 2017

After the click on the Create report button, Generated report files windowappears. It contains two type of files:1) Prediction report - a PDF file containing the prediction information

related to the target.2) Data matrix - a MS Excel file containing chemicals used for

prediction along with their data for selected parameters, profiles andendpoint tree positions.

1

2


ReportGenerated report files

April, 2017

Prediction report

Data matrix report

• You have now been introduced to the defining of target endpoint;

• You have now been introduced to the coloring of profilers anddatabases.

• You have now been introduced to the consecutive steps of thecalculation of alert performance.

• Note proficiency comes with practice.


Congratulation

April, 2017

oecd qsar toolbox v.4 is a step-by-step presentation designed to take the toolbox user through the...

Documents