Capturing a Comprehensive Picture of Biological Events From Adverse Outcome Pathways in the Drug Exposome

Abstract

Background: The chemical part of the exposome, including drugs, may explain the increase of health effects with outcomes such as infertility, allergies, metabolic disorders, which cannot be only explained by the genetic changes. To better understand how drug exposure can impact human health, the concepts of adverse outcome pathways (AOPs) and AOP networks (AONs), which are representations of causally linked events at different biological levels leading to adverse health, could be used for drug safety assessment. Methods: To explore the action of drugs across multiple scales of the biological organization, we investigated the use of a network-based approach in the known AOP space. Considering the drugs and their associations to biological events, such as molecular initiating event and key event, a bipartite network was developed. This bipartite network was projected into a monopartite network capturing the event-event linkages. Nevertheless, such transformation of a bipartite network to a monopartite network had a huge risk of information loss. A way to solve this problem is to quantify the network reduction. We calculated two scoring systems, one measuring the uncertainty and a second one describing the loss of coverage on the developed event-event network to better investigate events from AOPs linked to drugs. Results: This AON analysis allowed us to identify biological events that are highly connected to drugs, such as events involving nuclear receptors (ER, AR, and PXR/SXR). Furthermore, we observed that the number of events involved in a linkage pattern with drugs is a key factor that influences information loss during monopartite network projection. Such scores have the potential to quantify the uncertainty of an event involved in an AON, and could be valuable for the weight of evidence assessment of AOPs. A case study related to infertility, more specifically to "decrease, male agenital distance" is presented. Conclusion: This study highlights that computational approaches based on network science may help to understand the complexity of drug health effects, with the aim to support drug safety assessment.

Keywords: AON; AOP; NAM; bipartite network; drug-AOP; network science.

Figure 1

Workflow of the procedure to assess AOPs in the drug exposome. A multistep procedure was developed: (1) Data compilation: drug–event associations were extracted from various data sources (solid lines). Several event types were considered, which are molecular initiating events (MIEs) (purple circles), key events (KEs) (light green circles), and adverse outcomes (AOs) (red circles); (2) Bipartite network development: a bipartite drug-event network was created based on the compiled data. To enrich the created network, KE relationships (KERs) (dash lines) extracted from the AOP-wiki database were identified, allowing to add other KEs (green circles); (3) Generation of the monopartite network: if two events shared at least one drug in the bipartite network, a link was created between these two events in the monopartite network. The transformation of the bipartite network to a monopartite network was done by using the “guilt by association” principle; (4) Network analysis: the loss of information from the bipartite network to the monopartite network was quantified by calculating the “increase of uncertainty” and the “loss of coverage.”

Figure 2

Illustration of bipartite network projection to a monopartite network. In a bipartite network G^b, containing two types of nodes (events represented as circles and drugs as diamonds), the associations between drugs and events are defined as “linkage patterns”. For example, in G^b1, the A, B, C events are connected to a drug 1 via 3 edges, which define a linkage pattern ABC with a linkage profile of 1 (occurrence of 1). In the meantime, the A and B events are both connected to drug 2 one time, therefore defining another linkage pattern AB with a linkage profile of 1. Similarity, in G^b2, the linkage pattern AB has a linkage profile of 2 as it appears two times (drugs 1 and 2). The linkage profile of linkage pattern ABCDE is 1 (drug 3). Then, from the bipartite network, a projection to a monopartite network G^m is performed for the events. If two events share at least one common drug in the bipartite network, an association between these two events is generated. Finally, G^b1 and G^b2 were transformed into G^m1, G^m2, respectively. As observed, G^m1 and G^m2 are present as “cliques” meaning that all the events in G^m1 and G^m2 are fully interconnected. It is showed that the linkage pattern AB appears twice in G^b2 while oit is overlapping as an edge AB in G^m2, thus resulting in the information loss.

Figure 3

Representation of the bipartite network. The network illustrates the drug–event associations data compiled from the CompTox and the AOP-wiki databases. Each circle node represents one event, colored by the event type to which it belongs (MIE, KE or AO). Each yellow diamond node represents one drug. The size of the event node corresponds to the number of drugs known to be linked to event(s) (from 1 to 134). The size of the drug node corresponds to the number of events associated with this drug (from 1 to 12). The added KEs, using the KERs information from the AOP-wiki database, are in directed green dash lines. Directed solid lines indicate direct drug–event associations extracted during te compilation phase. These edges are colored by the event type to which drugs are connected i.e., drug-MIE in purple, drug-KE in green, and drug-AO in red.

Figure 4

Uncertainty increase (A) and loss of coverage (B) of the monopartite network. (A) Illustrates the increase in uncertainty, and (B) represents the loss of coverage, both for the monopartite network of events. Each circle node represents one event, colored by the event type to which it belongs (MIE, KE, AO). Solid edges indicate event–event associations. In (A), the width of nodes increase according to the value of the increase in uncertainty. In (B), the width of nodes is raised with the value of loss of coverage. The dark color of the edges shows the highest value of loss of coverage.