Citation:

Oki, N., S. Bell, M. Angrish, C. Grant, R. Wang, AND S. Edwards. Computationally predicted Adverse Outcome Pathway networks for liver-related diseases using publicly available data sources: Case studies and lessons learned. SOT Annual Meeting, Baltimore, MD, March 12 - 16, 2017.

Impact/Purpose:

Adverse Outcome Pathways can expand and enhance the use of ToxCast and other in vitro toxicity information by providing a mechanistic link to adverse outcomes of regulatory concern, but the expert-driven development of AOPs is labor-intensive and time-consuming. This work is intended to complement that process by creating a broad array of computationally-predicted AOPs (cpAOPs) by data mining of publicly available data. This increases the coverage of AOPs and provides minimal information in many cases where nothing is available otherwise. It also provides a starting point for expert-driven development of AOPs and thereby potentially accelerating that process.

Description:

The Adverse Outcome Pathway (AOP) framework summarizes key information about mechanistic events leading to an adverse health or ecological outcome. In recent years computationally predicted AOPs (cpAOP) making use of publicly available data have been proposed as a means of accelerating the AOP development process. Previously, we developed a method for building cpAOP networks using market basket analysis. Here we merged networks derived from distinct input datasets to create a more comprehensive global network and automated the extraction of subnetworks focusing on liver-related outcomes and diseases. We generated and extracted cpAOP subnetworks for 46 liver diseases as classified in CTD. Network similarity was assessed by comparing nodes and edges between network pairs.Using the cpAOP networks for 3 fatty liver related diseases from CTD, we compared our results with an expert derived AOP network (Angrish et al, 2016). The findings showed good concordance between our network and the expert derived AOP. Our case studies also showed that communities (phenotypes and pathway clusters) obtained when focusing on one subgroup of liver disease may be intermediary stages for other later stage outcomes, illustrating the progression that occurs in nature. Similarity comparisons between the disease networks showed that closely related diseases, e.g. alcoholic vs non-alcoholic fatty liver disease, had relatively low dissimilarity (S) scores 0 < S < 0.3 (where maximum dissimilarity is 1). However these results also revealed potential bias due to differences in sample size and specificity of phenotypic measurements. These biases should be reduced by the incorporation of additional data, which is easily accomplished via our workflow. Our findings demonstrate that methods and procedures presented here are able to build networks across different levels of biological organization and construct communities that are more meaningful for key event identification and AOP development. The views expressed in this abstract are those of the authors and do not necessarily reflect the views or policies of the U.S. EPA

Record Details: