Research Article: Predicting Chemical Toxicity Effects Based on Chemical-Chemical Interactions

Date Published: February 15, 2013

Publisher: Public Library of Science

Author(s): Lei Chen, Jing Lu, Jian Zhang, Kai-Rui Feng, Ming-Yue Zheng, Yu-Dong Cai, Gajendra P. S. Raghava.


Toxicity is a major contributor to high attrition rates of new chemical entities in drug discoveries. In this study, an order-classifier was built to predict a series of toxic effects based on data concerning chemical-chemical interactions under the assumption that interactive compounds are more likely to share similar toxicity profiles. According to their interaction confidence scores, the order from the most likely toxicity to the least was obtained for each compound. Ten test groups, each of them containing one training dataset and one test dataset, were constructed from a benchmark dataset consisting of 17,233 compounds. By a Jackknife test on each of these test groups, the 1st order prediction accuracies of the training dataset and the test dataset were all approximately 79.50%, substantially higher than the rate of 25.43% achieved by random guesses. Encouraged by the promising results, we expect that our method will become a useful tool in screening out drugs with high toxicity.

Partial Text

Toxicity is a key cause of late-stage failures in drug discovery. Even some approved drugs such as Phenacetin [1] and Troglitazone [2] have been withdrawn from the market because of unexpected toxicities that were not detected during Phase III clinical trials. Thus, early toxicology data on compounds are needed to reduce R&D costs. Evaluating toxicity and assessing risks of diverse chemicals require comprehensive experimental testing against a broad spectrum of toxicity end points. These tests can cost millions of dollars, involving several thousand animals, and take many years to complete. As a result, very few chemicals have undergone the degree of testing needed to support accurate health risk assessments or meet regulatory requirements for drug approval. In recent years, the number of synthetic compounds has surged with the advance of combinatorial chemistry, and accordingly large quantities of toxicity data are urgently demanded.

As described in the Section “Benchmark dataset”, 10 test groups were constructed to evaluate the method described in Section “Prediction method”. In each test group, there were one training dataset consisting of 15,510 compounds and one test dataset containing 1,723 compounds. The predicted results for each test group obtained by the proposed method are as follows.