Prediction of human population responses to toxic compounds by a collaborative competition

Abstract

The ability to computationally predict the effects of toxic compounds on humans could help address the deficiencies of current chemical safety testing. Here, we report the results from a community-based DREAM challenge to predict toxicities of environmental compounds with potential adverse health effects for human populations. We measured the cytotoxicity of 156 compounds in 884 lymphoblastoid cell lines for which genotype and transcriptional data are available as part of the Tox21 1000-Genomes Project. The challenge participants developed algorithms to predict inter-individual variability of toxic response from genomic profiles and population-level cytotoxicity data from structural attributes of the compounds. 179 submitted predictions were evaluated against a blinded experimental dataset. Individual cytotoxicity predictions were better than random, with modest correlations (Pearson’s r<0.28), consistent with complex trait genomic prediction. In contrast, predictions of population-level response to different compounds were higher (r<0.66). The results highlight the possibility of predicting health risks associated with unknown compounds, although risk estimation accuracy remains suboptimal.

Crowdsourcing biomedical research: leveraging communities as innovation engines

“…The NIEHS–NCATS–UNC DREAM Toxicogenetics Challenge 60 was a collaboration between the US National Institute of Environmental Health Sciences (NIEHS), the US National Center for Advancing Translational Sciences (NCATS) and the University of North Carolina (UNC). It was designed to assess the capabilities of current methodologies to address two crucial issues in the context of chemical safety testing: first, the use of genetic information to predict cellular toxicity in response to environmental compounds across cell lines with different genetic backgrounds; and second, the use of compound structure information to predict population-level cellular toxicity in response to new environmental compounds.…”

DeepTox: Toxicity Prediction using Deep Learning

“…In the case of the Tox21 challenge dataset, 12,707 compounds were reduced to 8694 distinct fragments. To counteract the reduction in the training set size, an optional augmentation step was introduced to DeepTox: kernel-based structural and pharmacological analoging (KSPA), which has been very successful in toxicogenetics (Eduati et al, 2015). The central idea of KSPA is that public databases already contain toxicity assays that are similar to the assay under investigation.…”

A tiered, Bayesian approach to estimating of population variability for regulatory decision-making_suppl

“…In addition, further refinement of the prior distribution may be possible through chemo-informatic approaches -using chemical structure information to give greater weight to chemicals in the database that are more similar to the chemical of interest, such as incorporating the models reported in Eduati et al (2015). Indeed, we found that the distance between chemicals in chemical property space (e.g., as in Fig.…”

“…One solution is to develop computational models based on the already collected data, either to predict susceptibility to chemicals based on the constitutional genetic makeup of an individual or to forecast which chemicals may be most prone to eliciting widely divergent responses in a human population. Indeed, the large-scale population based in vitro toxicity data of Abdo et al (2015b) enabled development of an in silico approach to predicting individual-and population-level toxicity associated with unknown compounds (Eduati et al, 2015). This exercise showed that in silico models that produced predictions which were statistically significantly better than random could be developed, but the correlations were modest for individual cytotoxicity response and only somewhat better for population-level responses, consistent with predictive performances for complex genetic traits.…”