- Highlighted Item:
- Webinar:
- Publisher:Lhasa Limited
- Publication Date:SEP 2020
- Reference:
- DOI:
- PMID:
- Publication Type:Presentation
- Scientific Area:
- Endpoint:Carcinogenicity
- Industry Type:
Integrating Knowledge Of Carcinogenicity Adverse Outcome Pathways (AOPs) With Experimental Data_EMGS Data Challenge 2020
Integrating Knowledge Of Carcinogenicity Adverse Outcome Pathways (AOPs) With Experimental Data
Steven Kane, Alex N. Cayley, Robert Foster, Susanne A. Stalford and Richard V. Williams
The assessment of carcinogenicity and related toxicity endpoints is a principal area of research in the development of alternative methods. However, it remains unclear how results from these alternative methods can be combined with each other, and with more traditional assay results, to give the best overall prediction of genotoxicity and carcinogenicity.
Previously, we have described how knowledge relating to the carcinogenicity endpoint contained in the expert rule-based prediction system Derek Nexus (DX) was rearranged to generate a network of adverse outcome pathways (AOPs) that share common key events (KEs) and which can be interrogated at different levels (Figure 1). The network developed from this work currently captures knowledge on 39 AOPs with over 350 pathways and 361 KEs relating to the endpoint of carcinogenicity.
Figure 1: An AOP network created from the tubulin binding MIE.
Subsequently, evidence relating to assays, measurements and in silico models (in the form of DX alerts) have been associated with the appropriate place on the network. To date we have associated 291 alerts, 73 assays and 69 measurements with the pathways. Assays and measurements are a mixture of traditional tests, binding assay data and relevant biomarker assays.
In this work we outline how this network can be used to provide the basis for carcinogenicity predictions for individual compounds and that presenting knowledge in this way will allow the user to more intelligently combine in vitro and in vivo data with hypotheses for a predicted mode of action (MOA). The methodology was tested using a published dataset of EPA pesticides with a human carcinogenicity category (Table 1).
|
Balanced accuracy |
Sensitivity |
Specificity |
No. Compounds |
DX carc. prediction |
61 |
52 |
70 |
310 |
AOP Prediction |
62 |
67 |
57 |
310 |
Tiered Prediction |
66 |
60 |
72 |
284 |
Table 1: Results from a validation study. Chemicals Evaluated for Carcinogenic Potential. Office of Pesticide Programs U.S. Environmental Protection Agency. Annual Cancer Report 2018.
It is hoped that this approach will allow the user to build a weight of evidence (WOE) to predict the carcinogenicity of a query compound and determine the most appropriate next steps in the testing of a hypothesis.