Challenges in drug-induced liver injury (DILI) prediction
(Waring et al., Nat Rev Drug Discov, 2015, 14, 475-486)
Drug development costs more than US$2 billion and takes over 10 years. Safety (clinical safety and non-clinical toxicology) was the number one reason for the project termination according to the analysis on drug candidate attrition in the major pharmaceutical companies.
(Horner et al., Regul Toxicol Pharmacol, 2013, 65, 334-343)
According to the analysis on the preclinical reports, liver injury was the most frequently observed organ toxicity in the animal studies.
(Olson et al., Regul Toxicol Pharmacol, 2000, 32, 56-67)
DILI was one of frequent organ toxicities in human with the highest percentage of project termination among the drugs entered the clinical phases after animal testing. Particularly, idiosyncratic DILI can’t be predicted by the animal models and causes serious safety issues in the clinical phase.
(Matthews et al., Regul Toxicol Pharmacol, 2009, 54, 23-42)
In order to predict DILI in the early phase, various (Q)SAR models have been developed; however, prediction accuracy of the models is not yet sufficient to find out the DILI positives. (Specificity is prediction accuracy to detect the DILI negatives whereas sensitivity is to detect the DILI positives.) Moreover, only few prediction models were deployed as a usable software such as web program and GUI program. Thus, most of prediction models are not accessible.
ToxSTAR aims to predict DILI positives in the early phase of drug development. Drugs with DILI concerns were used to produce in vitro, in vivo, and in silico data and to collect publicly available data to predict DILI. Currently, it is under progress to develop novel in vitro, in vivo, and in silico models to predict DILI in order to support drug development projects.