Brazil China Japan RussiaSpain

Drug Tolerability and Exposure: In silico Prediction using QSPR and PBPK Modelling

Tolerability of a compound tends to be examined through a toxicity-related point of view. However, exposure as mediated through pharmacokinetics can play an important role in determining tolerability. One clear example of this occurs when comparing LD50 via the oral route with LD50 via the intravenous route, where the oral route exhibits much higher tolerability most likely due to absorption limitations or enhanced first pass metabolism effects. Because there are now in silico tools that can accurately predict pharmacokinetic parameters directly from chemical structure (e.g., chemPKTM), we sought to investigate the in silico predictability of tolerability using outputs from chemPKTM in combination with other predictor variables such as structural descriptors.

The US FDA has published a list of marketed drugs, along with information such as maximum recommended therapeutic dose (MRTD) (a clinically-determined measure of tolerability), and chemical structure. The model developed at Cyprotex is based on 9 readily interpretable predictor variables; three of these were outputs from chemPKTM (gut wall partition co-efficient, gut wall Cmax and skin partition co-efficient) and the remaining six variables were structural descriptors calculated from 2D chemical structure (three based on bulk properties and three based on molecular surface areas). Using this model, it was possible to predict approximately 40% of the variability in MRTD. This is notable as 50% of the variability of the rat PO LD50 is predicted by rat IV LD50, meaning that the remaining 50% of variance may be explained by other factors e.g., pharmacokinetics. Further development and elaboration of this model is expected to improve of the prediction of tolerability, especially with regard to bias for less tolerable drugs.

This research was presented at the EFMC International Symposium on Medicinal Chemistry. Download the poster.

facebooktwitterlinkedin