Menu
Home
Log in / Register
 
Home arrow Computer Science arrow Computational systems pharmacology and toxicology
Source

In silico Methodologies

Overview

In silico methodologies, which includes expert alerts, Qsar models, and read-across, are being increasingly used as part of the chemical R&D process and submissions for regulatory authorities. To support their use, the oECD published five (Q)sAR model validation principles: “1. a defined endpoint; 2. an unambiguous algorithm; 3. a defined domain of applicability; 4. appropriate measures of goodness-of-fit, robustness and predictivity; 5. a mechanistic interpretation, if possible.”66 these principles support the evaluation of any model used in predicting toxicity. the following sections outline these three major in silico methodologies.

Expert Alerts

Expert alerts (also referred to as expert rule-based or structural alerts) is a methodology for generating predictions for specific toxicity endpoints. Commonly used commercial systems include the Leadscope Genetox expert Alerts67 and Derek Nexus from Lhasa limited.68 Non-commercial expert alerts include the Benigini Bossa rulebase, which is part of the oECD toolbox69 and TOxl^/ee.70 this methodology makes use of intellectually derived structural rules or alerts that are generally associated with specific toxic effects or mechanisms. these alerts are usually encoded as one or more molecular substructures that may have been reported in the public literature alongside a mechanistic justification for the structural features. information on the precise structural definition needs to be encoded alongside rules to describe any criteria where the alert would not match a test chemical.

the aromatic nitro substructure is an example of an expert rule or alert for mutagenicity, as shown in Figure 9.1. it has been cited in a number of publications.71-75 the structural definition defines both the aromatic nitro substructure along with different substituents that potentially deactivate its mutagenicity. In Figure 9.1 three deactivating fragments are shown for illustration. the alert will match any compound containing an aromatic nitro; however, if the test compound also contains one of the deactivating fragments then the alert will not match. In addition to the structural definition of the alert, information on the mechanism has also been collected from

Chapter 9

Example alert (aromatic nitro) for mutagenicity from the Leadscope expert alerts

Figure 9.1 Example alert (aromatic nitro) for mutagenicity from the Leadscope expert alerts.

different sources. The number of positive and negative known examples is reported along with data behind these tested chemicals.

Any alert reported in the literature should be critically evaluated before it is incorporated into the knowledge base of the alert’s system and used to make prediction. A database of known positive and negative results (i.e. a reference set) can help to qualify the alerts by performance. Each prospective alert should have a significant positive association with the reference set data. This positive association results from when there is a significantly higher than expected number of positive compounds containing the alert than expected from a random sample. The number of positive and negative compounds matching the aromatic nitro alert is shown in Figure 9.1. In this example, 911 were positive and 100 were negative (or 90% positive). In comparison to the overall number of positive and negative examples in the reference set (48% positives), the association was determined to be significant and the aromatic nitro was included as an alert for mutagenicity.

If an expert alert system adheres to the OECD (Q)SAR validation principle 4, before any prediction is made, it should be established whether the test chemical is within the system’s applicability domain. A positive prediction occurs when an alert is present in the test compound (based on the structural definition of the alert). If no alert occurs and the chemical is within the

Four examples that were predicted using an expert alert system

Figure 9.2 Four examples that were predicted using an expert alert system.

applicability domain, then a negative prediction is made. It is possible that there is some conflicting evidence for the alert which may result in the alert being classified as indeterminate.

Figure 9.2 shows four examples where predictions for bacterial mutagenicity were generated using the Leadscope Genetox Expert Alerts system.67 Compound a was determined to be within the domain of applicability for the expert alert system, no alerts were identified and it was predicted to be negative. Compound B was predicted to be positive based on the presence of an aromatic nitro group highlighted in red. Based on the alert definition, no other group appears to deactivate the mutagenicity and hence it was predicted to be positive. An alkyl hydrazine alert was identified in compound C. There was conflicting data in the reference set matching this alert and hence it was assigned as indeterminate. like compound B, compound D also contains an aromatic nitro group (shown in gray) but was predicted to be negative based on the mitigated fragment highlighted in light blue/green. Further information on any of the matched alerts, as illustrated in Figure 9.1, would provide additional supporting information to assess the result.

 
Source
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
 
Subjects
Accounting
Business & Finance
Communication
Computer Science
Economics
Education
Engineering
Environment
Geography
Health
History
Language & Literature
Law
Management
Marketing
Mathematics
Political science
Philosophy
Psychology
Religion
Sociology
Travel