1. QSAR Application Toolbox: First Steps - Data Gap Filling (Read-Across by Analogue Approach)

2. Background This is a step-by-step presentation designed to take the first time user of the Toolbox through the workflow of the least complex of the data filling exercises. This presentation reflects the “Getting Started” document available at [www.oecd.org/env/existingchemicals/qsar]www.oecd.org/env/existingchemicals/qsar

3. Objectives This presentation demonstrates a number of functionalities of the Toolbox : –Identify analogues for a target chemical, –Retrieve experimental results available for those analogues, and –Fill data gaps by read-across.

4. Specific Aims To introduce the first time user to the work flow of the Toolbox. To familiarize the first time user with the six modules of the Toolbox. To familiarize the first time user with the basic functionalities within each module. To explain to the first time user the rationale behind each step of the exercise.

5. The Exercise In this exercise we will predict the skin sensitization potential for an untested compound, (4-nitrobenzoyl chloride) [CAS # 122-04-3], which will be the “target” chemical. This prediction will be accomplished by collecting a small set of test data for chemicals considered to be in the same category as the target molecule. The category will be defined by the mechanism of protein binding common to all the chemicals in the category. The prediction itself will be made by “read- across”.

6. Read-across & the Analogue Approach Read-across can be used to estimate missing data from a single or limited number of chemicals using an analogue approach. In the analogue approach, endpoint information for a single or small number of tested chemicals is used to predict the same endpoint for an untested chemical that is considered to be “similar.” Analogous sets of chemicals are often selected based on the hypothesis that the toxicological effects of each member of the set will show a common behavior.

7. Side-Bar On Sensitization Allergic contact dermatitis that results from skin sensitization is a significant health concern. Skin sensitization is a toxicological endpoint that is complex and conceptually difficult. However, there is growing agreement that all organic chemicals must react covalently with skin protein in order to behave as skin sensitizers. Therefore, mechanism by which organic chemicals bind with proteins are relevant to grouping chemicals that may be skin sensitizing agents.

8. Tracks After opening the Toolbox, the user has to choose between three use tracks (or workflows): –(Q)SAR Track –Category Track –Flexible Track For a first time users of the Toolbox, it is recommended to select the Category Track. As the user becomes more familiar with the functionalities of the Toolbox, it is recommended to use the Flexible Track.

9. Workflow Each track follows the same general six- step workflow: –Chemical Input –Profiling –Endpoints –Category Definition –Filling Data Gaps –Reporting

10. Tracks and Workflow

11. Chemical Input Click on the “Category Track”. This takes you to the first module, which is “Chemical input”. This module provides the user with several means of entering the 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 assigned is the correct one.

12. Chemical Input Screen

13. Ways of Entering a Chemical There are several ways to enter a target chemical. Most often used are: –Chemical Abstract Services (CAS) number (#), –SMILES (simplified molecular information line entry system) notation, and –Drawing the structure using the structural editor included in the Toolbox. Click on CAS#.

14. Enter CAS # 122043; Click on Search

15. The Search for the Structure The Toolbox now searches the databases to find out if the CAS# you entered is linked to a molecular structure stored in the Toolbox. If the structure is identified, it is displayed as a 2D image.

16. Click on OK to Accept Target

17. Target Chemical You have now selected your target chemical and have a structure for it. This is important as from here on the workflow will be based on the structure coded in SMILES. Click on the box next to “Substance Information”; this displays the chemical identification information. The workflow on the first module is now complete; Click “Next” to move to the module “Profiling”

18. Chemical Identification Information

19. Profiling “Profiling” refers to the electronic process of retrieving relevant information on the target compound, other than environmental fate, ecotoxicity, and toxicity data, which are stored in the Toolbox. Available information includes likely mechanism(s) of action.

20. Profiling Target Chemical Select the “Profiling methods” you wish to use. When first using the Toolbox, all the profiling methods are pre-selected (red checked boxes). You can therefore deselect (red check disappears) those profilers you DO NOT wish to use by clicking on the red checked box before the name of the profiler. For this example, deselect all the profilers except for the “mechanistic” methods. Click on “Apply”.

21. Profilers for 4-nitrobenzoyl chloride

22. Side-Bar to Profiling Detailed explanations of the different profilers are provided in the Manual accompanying the Toolbox. In addition, for most of the profilers, background information can be retrieved by highlighting one of the profilers (for example ‘’Protein Binding’’) and clicking on the button ‘’Show Category Boundaries’’. This is demonstrated in the next two slides.

23. Highlight ‘’Protein Binding’’ and Click on ‘’Show Category Boundaries’’

24. Insert Window Appears with the Explanation

25. Profiling The actual profiling will take several seconds to minutes depending on the number and type of profiler(s) selected. The results of profiling automatically appear as a dropdown box under the target chemical (see next slide).

26. Profiles of 4-nitrobenzoyl chloride (1)

27. Profiles of 4-nitrobenzoyl chloride (2) Very specific profiling results are obtained for the target compound. Please note the specific protein- binding profile (see side-bar on sensitisation above). These results will be used to search for suitable analogues in the next steps of the exercise. Click on “Next” to move to the module “Endpoints”.

28. Endpoints “Endpoints” refer to the electronic process of retrieving the environmental fate, ecotoxicity and toxicity data that are stored in the Toolbox. Data gathering can be executed in a global fashion (i.e., collecting all data of all endpoints) or on a more narrowly defined basis (e.g., collecting data for a single or limited number of endpoints).

29. This Example In this example, we limit our data gathering to a single toxicity endpoint (skin sensitization). The default setting is to gather data from all databases except for the “Danish EPA” database (which contains estimated properties). Clicking on the box with the black check mark deselected the database. Deselect all databases except for “ OASIS Skin Sensitization” (a black check only appears in the box next to this database). Click on “Gather data”.

30. Oasis Skin Sensitization Data Gathering

31. Next Step in Data Gathering Toxicity information on the target chemical is electronically collected from the selected dataset(s). In this example, an insert window appears stating there was “no data found” for the target chemical (see next slide). Close the insert window.

32. No data for Target Chemical

33. Side-Bar to Data Collection Since data is endpoint specific the data selection is presented in a drop-down menu. By double clicking on an endpoint, the data tree is expanded.

34. Recap You have entered the target chemical being sure of the correct structure. You have profiled the target chemical. You have found that no experimental data is currently available for this structure. In other words, you have identified a data gap, which you would like to fill. Click on “Next” to move to the category definition module.

35. Category Definition This module provides the user with several means of grouping chemicals into a toxicologically meaningful category that includes the target molecule. This is the critical step in the workflow. Several options are available in the Toolbox to assist the user in refining the category definition.

36. Grouping Methods Allow the user to group chemicals into chemical categories according to different measures of “similarity” so that within a category data gaps can be filled by read- across. For example, starting from a target chemical for which a specific protein binding mechanism is identified, analogues can be found which can bind by the same mechanism and for which experimental results are available.

37. Protein Binding This is one of the best grouping methods in the Toolbox. It is built on conventional organic chemical mechanisms and as such is qualitative in character. This method is particularly relevant for respiratory and skin sensitization and acute aquatic toxicity, but also for chromosomal aberration and acute inhalation toxicity.

38. Protein Binding Categorization This scheme includes 38 categories such as haloalkanes, isocyanates, isothiocyanates, diketones, aldehydes, acyl halides, alkyl sulfates, sulfonates, etc. Each category is presented by defined 2D structural alerts that are responsible for the eliciting of effects as a result of protein binding, such as skin sensitization. The associated mechanisms are in accordance with the existing knowledge on electrophilic interaction mechanisms of various structural functionalities.

39. Remember There is agreement that all organic chemicals must react covalently with skin protein in order to behave as skin sensitizers. Therefore, mechanism by which organic chemicals bind with proteins are relevant to grouping chemicals that may be skin sensitizing agents. So you have mechanistic reasons for defining your category based on similar protein- binding mechanism. Highlight “Protein Binding”. Click on “Defining Category”.

40. Defining the Category

41. Confirmation of Mechanism An insert window confirming the protein binding mechanism of the target chemical appears. Click on “OK”.

42. Naming Category Another insert window listing the default category name appears. Click “OK”.

43. Analogues The data is automatically collated. Based on the defined category (nucleophilic substitution of acyl halides) 6 analogues have been identified. These 6 compounds along with the target chemical form a category, which can be used for data filling (see next slide).

44. Mechanistic Analogues

45. Recap You have identified a mechanistic category (nucleophilic substitution of acyl halides) for the target chemical (4- nitrobenzoyl chloride). Available experimental results on skin sensitisation for six chemicals with the same mechanism of protein binding were found in the “Oasis Skin Sensitization” database. The available data for these 6 chemicals can now be collected.

46. Next Step in Gathering Data Highlight the “[7] Nucleophilic Substitution …..” under “Single Chemical” in the “Defined Categories” box. The inserted window entitled “Read Data?” appears (see next slide).

47. What data to collect?

48. Data Selection To select the data to be read you click on the box(s) before the name of the data type. This selects (a red check mark appears) or deselects (red check disappears) the data type. Click on the box next to “Toxicological Information”. This places a red check mark in the box next to this data type (the one we want to read). Click on “OK” (see next slide).

49. Reading the Selected Data

50. Summary of Skin Sensitization Information for Analogues

51. Side-Bar on Data Note the structure of the 6 compounds with experimental results are shown. Double clicking on a structure enlarges the view of the structure. Details on the experimental results can be retrieved by double-clicking on any cell in the data matrix line.

52. Navigating Through the Data Tree The user can navigate through the data tree by closing or opening the nodes of the tree. Double-click on the node next to “Toxicological Information” and then “Sensitisation”. In this example, results from skin sensitization testing are available (see next slide).

53. Data Tree

54. Recap You have now read in the available experimental results on skin sensitisation for six chemicals with the same mechanism of protein binding as the target compound, which were found in the “Oasis Skin Sensitization” database. You are ready to fill the data gap. Click on “Next” to access the module filling data gaps.

55. Filling Data Gaps This step in the work flow provides the user with three options for making an endpoint-specific prediction for the target molecule. As noted earlier, these options, in increasing order of complexity, are – by read-across, – by trend analysis, and – through the use of QSAR models. In this example we only use read-across.

56. The Filling Data Gap Window Take a moment to examine the filling data matrix on the next slide. Note it contains –information on the chemicals, which form the category, –the 3 options for data filling, and –a means of selecting data points used to fill the data gap (see next slide).

57. Selecting the Data Point Before applying read-across, the Toolbox allows the user to decide which type of results should be used in case more than one result is available for any analogue, (i.e., all values, average values, minimum or maximum results) (see next slide). It should be noted that averaging results is only useful for quantitative endpoints. Click “ All values”.

58. Data Point Selection

59. Applying Read-across Highlight the data endpoint box (Toxicological Information_Sensitisation_skin) under the target chemical. It will be empty as it is the data gap. Next with the “read-across” box highlighted, click “Apply” (see next slide).

60. Apply Read-across

61. Results of Read-across

62. Interpreting the Read-across Figure The resulting plot is experimental results of all analogues (Y axis) according to a descriptor (X axis) with the default descriptor of log Kow (see next slide). The RED dot represent s the target chemical, while the PURPLE dots the experimental results available for the analogues that are used for the read- across; the BLUE dot represent the experimental results available for the analogues but not used for read-across.

63. Interpretation of Read-across In this example, all results of the analogues are consistent; they all present a high sensitizing potential. The same high sensitising potential is therefore predicted for the target chemical. Accept the prediction by clicking “Accept” (see next slide).

64. Accepting Prediction

65. Filled Data Gap By accepting the prediction the data gap is filled (see next slide). You are now ready to complete the final module and to download the report. Click on “Next” to access the module “Report”.

66. Filled Data Gap

67. Report The final step in the workflow, report, provides the user with a downloadable written audit trail of what the Toolbox did to arrive at the prediction. This study history can be printed or copied to be inserted in a more detailed report (see next slide). Click on “Finish”.

68. Finished