15 In Silico Prediction of the Point of Departure (POD) …
301
modeling approaches are expected to play critical roles for this purpose. This vision
was elaborated in several important publications, including two National Research
Council reports—Toxicity Testing in the 21st Century [1] and Using 21st Century
Science to Improve Risk-Related Evaluations [2].
The requirements regarding cost and speed provide another impetus to adopt highthroughput assays and in silico methods. There is a huge backlog of chemicals to be
evaluated by regulatory agencies around the world. Chronic animal experiments alone
are too costly and time-consuming to deal with this problem efficiently. On the other
hand, genomics and high-throughput assays can potentially provide a comprehensive
picture of perturbed pathways, which can then be used to guide targeted testing. The
significant societal interest for reducing animal testing also calls for greater use of
innovative assay methods and in silico modeling approaches [3]. The same principle
has been advocated by the EU program for registration, evaluation, authorization and
restriction of chemical substances (REACH) program and EU Cosmetic Directive
[4].
Right now, it is very common to generate transcriptomic profiles for chemicals
under consideration using either microarrays or next generation sequencing technology (RNAseq) to provide insights for toxicological mechanisms. An example of
systematic data generation efforts is the Open TG-GATEs (Toxicogenomics ProjectGenomics Assisted Toxicity Evaluation System [5]), through which a large-scale
database consisting of data for gene expression and pathology from both animaltissue- and cell-line-based experiments on 170 compounds has been generated. It
provides an excellent source for exploring transcriptomic changes caused by some
important chemicals.
In the area of high-throughput in vitro assays, several large projects are ongoing to
develop high-throughput-cell-based- or cell-free tests for toxicological evaluations
and to establish data repositories with a diverse collection of chemicals. The Tox21
program and EPA’s ToxCast are two important projects in this field. Phase I of the
Tox21 program [6, 7] focused on more than 50 assays regarding cytotoxicity, mitochondrial toxicity, cell signaling, DNA damage, nuclear-receptor activation, among
others; with testing on more than 2800 chemicals completed. The ToxCast program
[8, 9] examines high-throughput assays covering a range of cell responses and over
300 signaling pathways [10]. More than 2000 chemicals have been evaluated in ToxCast Phase I and Phase II. Both the Tox21 and ToxCast programs have generated
data in dose response format for chemical-endpoint combinations. Various models
for hazard identification and prioritization for screening have been developed using
these datasets [11–13].
Another important development is the coordinated effort to characterize adverse
outcome pathways (AOPs, [14, 15]). An AOP links a molecular initiating event (MIE)
to the adverse outcome (AO) via a series of key events (KEs), which is specified by
key event relationships (KERs). Mechanistic information can thus be connected to
apical endpoints in a formalized, quality-controlled, and transparent way. As MIEs
and KEs are often associated with certain genes, proteins, or metabolites; AOPs
provide a valuable framework to describe the biological context for mechanistic
information regarding transcriptomic and in vitro assays. Here, it is also useful to
301
modeling approaches are expected to play critical roles for this purpose. This vision
was elaborated in several important publications, including two National Research
Council reports—Toxicity Testing in the 21st Century [1] and Using 21st Century
Science to Improve Risk-Related Evaluations [2].
The requirements regarding cost and speed provide another impetus to adopt highthroughput assays and in silico methods. There is a huge backlog of chemicals to be
evaluated by regulatory agencies around the world. Chronic animal experiments alone
are too costly and time-consuming to deal with this problem efficiently. On the other
hand, genomics and high-throughput assays can potentially provide a comprehensive
picture of perturbed pathways, which can then be used to guide targeted testing. The
significant societal interest for reducing animal testing also calls for greater use of
innovative assay methods and in silico modeling approaches [3]. The same principle
has been advocated by the EU program for registration, evaluation, authorization and
restriction of chemical substances (REACH) program and EU Cosmetic Directive
[4].
Right now, it is very common to generate transcriptomic profiles for chemicals
under consideration using either microarrays or next generation sequencing technology (RNAseq) to provide insights for toxicological mechanisms. An example of
systematic data generation efforts is the Open TG-GATEs (Toxicogenomics ProjectGenomics Assisted Toxicity Evaluation System [5]), through which a large-scale
database consisting of data for gene expression and pathology from both animaltissue- and cell-line-based experiments on 170 compounds has been generated. It
provides an excellent source for exploring transcriptomic changes caused by some
important chemicals.
In the area of high-throughput in vitro assays, several large projects are ongoing to
develop high-throughput-cell-based- or cell-free tests for toxicological evaluations
and to establish data repositories with a diverse collection of chemicals. The Tox21
program and EPA’s ToxCast are two important projects in this field. Phase I of the
Tox21 program [6, 7] focused on more than 50 assays regarding cytotoxicity, mitochondrial toxicity, cell signaling, DNA damage, nuclear-receptor activation, among
others; with testing on more than 2800 chemicals completed. The ToxCast program
[8, 9] examines high-throughput assays covering a range of cell responses and over
300 signaling pathways [10]. More than 2000 chemicals have been evaluated in ToxCast Phase I and Phase II. Both the Tox21 and ToxCast programs have generated
data in dose response format for chemical-endpoint combinations. Various models
for hazard identification and prioritization for screening have been developed using
these datasets [11–13].
Another important development is the coordinated effort to characterize adverse
outcome pathways (AOPs, [14, 15]). An AOP links a molecular initiating event (MIE)
to the adverse outcome (AO) via a series of key events (KEs), which is specified by
key event relationships (KERs). Mechanistic information can thus be connected to
apical endpoints in a formalized, quality-controlled, and transparent way. As MIEs
and KEs are often associated with certain genes, proteins, or metabolites; AOPs
provide a valuable framework to describe the biological context for mechanistic
information regarding transcriptomic and in vitro assays. Here, it is also useful to