302
D. Wang
mention another concept, mode of action (MOA), which is similar in principle to
AOP but with different emphasis [14]. Most discussions in this review are applicable
to both AOP and MOA, and both terms are used in the literature cited.
Though our knowledge about mechanistic information for toxicity is far from complete, substantial efforts have been made to incorporate genomic and high-throughput
assays into risk assessment with some notable successes. In this review, we will focus
on the problem of obtaining estimates for PODs with genomic and high-throughput
assay endpoints. Compared to hazard identification, determining the POD requires
the characterization of the dose response relationship regarding the chemical under
evaluation. If the dose response experiment has been performed under settings different from the targeted POD (e.g., in vitro assays relative to in vivo PODs), then there
is the added question how to translate the concentration between different settings.
For this, using transcriptomic data to infer in vivo POD with the same animal tissue
is the most straightforward. The seminal paper by Thomas et al. [16] pioneered the
strategy of using the most sensitive pathways, which has been further studied by a
number of research groups. In the next section, we will review current methods for
inferring in vivo PODs with transcriptomic data from the same animal tissues. PODs
derived from in vitro assays will be then discussed, followed by a review of recent
work using in vitro assay endpoints to predict in vivo PODs. We will conclude with
some general discussions. Due to the size of the literature in this field, we will not
cover every aspect of in silico modeling of PODs in detail. Rather, a general outline
will be provided regarding the most often used approach in this area. Though the
references cited do not constitute a complete list of relevant literature, readers can
use them as a starting point for further reading.
15.2 Infer In Vivo PODs with Transcriptomic Data
from the Same Tissues
Inferring the in vivo PODs with transcriptomic profiles from the same target tissue
(usually from rats or mice) for apical endpoints is the most mature approach discussed
in this review. Conceptually, a toxic chemical will trigger expression changes that
underpin both direct and indirect toxicity responses. There exists significant literature
comparing transcriptionally derived PODs with PODs based on apical endpoints.
The general finding is that transcriptional PODs are usually consistent with apicalendpoint-based values. Since the work of Thomas et al. [16], a number of authors
have proposed variations with this approach. To summarize these ideas, we will first
review the approach of Thomas et al. [17] as an example to illustrate various steps
taken in this type of studies. Then, we will discuss in detail important considerations
and possible choices in each step.
D. Wang
mention another concept, mode of action (MOA), which is similar in principle to
AOP but with different emphasis [14]. Most discussions in this review are applicable
to both AOP and MOA, and both terms are used in the literature cited.
Though our knowledge about mechanistic information for toxicity is far from complete, substantial efforts have been made to incorporate genomic and high-throughput
assays into risk assessment with some notable successes. In this review, we will focus
on the problem of obtaining estimates for PODs with genomic and high-throughput
assay endpoints. Compared to hazard identification, determining the POD requires
the characterization of the dose response relationship regarding the chemical under
evaluation. If the dose response experiment has been performed under settings different from the targeted POD (e.g., in vitro assays relative to in vivo PODs), then there
is the added question how to translate the concentration between different settings.
For this, using transcriptomic data to infer in vivo POD with the same animal tissue
is the most straightforward. The seminal paper by Thomas et al. [16] pioneered the
strategy of using the most sensitive pathways, which has been further studied by a
number of research groups. In the next section, we will review current methods for
inferring in vivo PODs with transcriptomic data from the same animal tissues. PODs
derived from in vitro assays will be then discussed, followed by a review of recent
work using in vitro assay endpoints to predict in vivo PODs. We will conclude with
some general discussions. Due to the size of the literature in this field, we will not
cover every aspect of in silico modeling of PODs in detail. Rather, a general outline
will be provided regarding the most often used approach in this area. Though the
references cited do not constitute a complete list of relevant literature, readers can
use them as a starting point for further reading.
15.2 Infer In Vivo PODs with Transcriptomic Data
from the Same Tissues
Inferring the in vivo PODs with transcriptomic profiles from the same target tissue
(usually from rats or mice) for apical endpoints is the most mature approach discussed
in this review. Conceptually, a toxic chemical will trigger expression changes that
underpin both direct and indirect toxicity responses. There exists significant literature
comparing transcriptionally derived PODs with PODs based on apical endpoints.
The general finding is that transcriptional PODs are usually consistent with apicalendpoint-based values. Since the work of Thomas et al. [16], a number of authors
have proposed variations with this approach. To summarize these ideas, we will first
review the approach of Thomas et al. [17] as an example to illustrate various steps
taken in this type of studies. Then, we will discuss in detail important considerations
and possible choices in each step.