Section 3: EPA's Toxicity Forecaster (ToxCast) program

EPA’s efforts to advance new approaches for better understanding and predicting chemical risk are coordinated through an umbrella research initiative called CompTox, housed at the EPA's National Center for Computational Toxicology (NCCT). CompTox houses several individual research programs focused on exposure estimation and hazard prediction -- including the development of computational models that attempt to mimic the functions of whole organs and tissues.

Arguably the most developed of the CompTox programs is the Toxicity Forecaster or ToxCast program. ToxCast uses high-throughput (HT) in vitro testing to assess potential hazards of chemicals. For more information on HT in vitro testing, visit Section 2 of the primer.

Impressively, the HT technology being employed in the ToxCast program allows for thousands of chemicals to be quickly tested—at multiple doses—for effects on hundreds of biological pathways. ToxCast assays are not conducted at EPA, but rather at EPA-contracted biotech companies (for a list of these contractors, visit this EPA CompTox webpage). The HT assays are each run in individual wells on multi-well plates requiring very small volumes of material—literally drops of liquid (see image below). The whole operation is automated and uses robots to carry out the tedious, repetitive work (see video of robots at work on right). These features—many assays run simultaneously on many chemicals at many doses, all automated—are what warrant the use of the term “high-throughput.”

HT in vitro testing systems utilize multi-well plates, such as the one pictured above, to quickly assess the toxicity of thousands of chemicals.

Photo by Nature Biotechnology

The data generated from ToxCast are being used by EPA to build predictive models of chemical toxicity related to specific adverse health outcomes, including impairments of early development, male and female reproductive function, and vascular development (Read this EPA factsheet to learn more about how ToxCast data are being used to build predictive models).

The ToxCast program divides its work into two phases. EPA refers to Phase I, now completed, as the “Proof of Concept” phase. In this phase, EPA selected nearly 300 chemicals, mostly pesticides that had been extensively tested using traditional methods. It ran each of them through approximately 600 HT in vitro assays.

Because the Phase I chemicals are data-rich, EPA is comparing the results of data generated from the ToxCast assays with data from traditional, in vivo tests. This comparison effort is a major aspect of EPA’s work to “validate” the ToxCast assays (i.e., determine how well the ToxCast assays predict outcomes seen in vivo).

According to a 2010 presentation by CompTox researchers, models developed using ToxCast Phase I data to screen for chronic, developmental and reproductive toxicity endpoints had few false positives (i.e., chemicals erroneously flagged as hazardous that are not), but had many false negatives (i.e., chemicals flagged erroneously as benign that are not). A 2011 CompTox publication that evaluated the accuracy of a rat reproductive toxicity model built using ToxCast assays indicated that the model was able to distinguish chemicals that were and were not reproductive toxicants with an accuracy of 80%. The predictive power of ToxCast toxicity models will vary from model to model, of course, depending on factors such as the quantity and quality of the underlying data. A key need will be to continuously assess and improve the predictability of such models.

Phase II of ToxCast was launched in 2010 and is ongoing. In this phase, to further explore the potential utility of ToxCast, EPA is evaluating an additional 767 chemicals drawn “from a broad range of sources including industrial and consumer products, food additives, touted “green” products, nanomaterials and drugs that never made it to the market.”

EPA is making data generated in both Phase I and II of the ToxCast program publicly available through this online database.

While the goals of the EPA efforts are laudable - filling massive data gaps and improving our ability to accurately predict adverse effects of chemicals, more quickly and at lower cost - there are important limitations and uncertainties that need to be recognized and addressed. Other sections of this primer provide an overview of both the benefits and the challenges to advancing this work.

For additional information from EPA on the ToxCast program visit:

For a description of related research initiatives, proceed to Section 4: EPA's CompTox Programs.

Screenshot from video of laboratory robot performing HT in vitro tests

Sophisticated robots conduct high-throughput (HT) in vitro tests in an automated and efficient manner. Click here to see a video of the machine in action. Image and video courtesy of the National Human Genome Research Institute