Prediction of AUCpo in humans - A modern in-silico-based system that outperforms traditional in-vitro- and animal-methods

U. Fagerholm. Prosilico, Huddinge, Sweden

Introduction The aims were to investigate limitations with lab methodologies for prediction of area-under-the-curve following oral dosing (AUCpo) in man, and to benchmark the performances of these methods versus our proprietary, validated in-silico-based prediction system.

Method Lab-based predictions of AUCpo (n=107) were taken from an investigation of 66 different in-vitro- and animal-methods by 11 big pharmas and one university (1). The literature was searched for data demonstrating limitations with lab methods. Results were compared to in-silico-predictions for several hundred compounds (molecular weight ∼150-750).

Results A significant portion of commonly used in-vitro and animal-based prediction methods are associated with >1000-fold maximum prediction errors and systemic errors (1). Million-fold errors have been shown for some methods, allometry inclusively (1-3). A large portion of non-quantifiable compounds (>50%) and large maximum errors (at least ∼2,000-fold) are among limitations with human hepatocyte-based predictions (2,4). Binding to material, low solubility, methodological differences (up to ∼200-fold for fu (2)) and contribution by efflux, conjugation and excretion are other limitations/challenges with in-vitro methods. Consequences include jeopardized safety in early clinical trials (a 5,800-fold underprediction of AUC following iv dosing with allometry is reported (3)), poor understanding of drug candidate characteristics, additional costs and delays. In benchmarking studies the in-silico system succeeded in predicting hundreds of drugs inaccessible to in-vitro methods, and outperformed in-vitro methods in 56 out of 57 difficult cases. The Prosilico in-silico system was also comparable, or better than, the best of animal-based prediction methods.

Conclusions Limitations with lab-methods for prediction of AUCpo in man were explored and highlighted. With our new proprietary in-silico system predictions can be improved, with the potential to enhance human safety, reduce animal experiments, costs and time, frontload decision-making (predictions before compound synthesis) and find better drugs.

References

1. Ring BJ et al. (2011). J Pharm Sci 100: 4090-4110.

2. Prosilico. Data on file.

3. Fuse E et al. (1998). Cancer Res 58: 3248-3253.

4. Stringer R et al. (2008). Xenobiot. 38: 1313-1329.