On August 10, 2015 the European Bioinformatics Institute (EMBL-EBI) reported about a new study “Prediction of human population responses to toxic compounds by a collaborative competition” published in the peer-reviewed scientific journal Nature Biotechnology. The aim of the study was to predict chemicals’ toxic effects in humans using computational methods. The cytotoxicity of 156 chemicals was measured in 884 different human lymphoblastoid cell lines from bone marrow. In a next step, over 200 participating scientists developed predictions for toxic responses of 106 compounds in 620 cell lines, based on the chemicals’ structural attributes. Next, they tested their predictive models on 50 substances for which only structural data were provided. The chemicals were substances found in consumer products and the environment, but also included inactive compounds, as well as pharmaceuticals. Finally, the predictions were compared to the experimental data set. The computational method worked better for predicting population-level effects, rather than for individual responses. The authors write that their findings “highlight the possibility of predicting health risks associated with unknown compounds”. They further acknowledge that the accuracy of risk estimates generated with the new method needs to be improved, but in general, the “wisdom of crowds” can provide a “more robust estimate than any individual prediction”.

Currently the Threshold of Toxicological Concern is commonly used in the area of food contact materials and flavorings to estimate risk for compounds where no toxicological data are available.

Read more

EMBL-EBI (2015). “Predicting the effect of toxic compounds on individuals: Crowdsourcing initiatve for systems biomedicine.” Science Daily.


Eduati, F. et al. (2015). “Prediction of human population responses to toxic compounds by a collaborative competition.” Nature Biotechnology (online August 10, 2015).