Models for ADME

Novartis Institutes for BioMedical Research, 250 Massachusetts Avenue, Cambridge, MA 02139, USA 

Annual Reports in Medicinal Chemistry, Volume 42 2007 Elsevier Inc. 

Nearly 15 years after Patched s speech, we still face the same problems that he highlighted. Medicinal chemists often synthesize a potent molecule and find later that it has poor exposure in vivo, and thus poor efficacy. Poor exposure can be caused by many different factors. Most of the factors affecting exposure are commonly known by acronym ADME - absorption, distribution, metabolism, and excretion. A fifth factor, solubility, is also very important and is commonly considered to be part of ADME. 

The prevalence of ADME problems is quite high, although the pattern has changed somewhat in recent years. Kola and Landis [4] reported that for 10 large pharmaceutical companies, ADME/formulation problems were responsible for 40% of clinical failures in the year 1991 but only 12% of clinical failures in the year 2000. Clinical safety and toxicity were responsible for 22% of clinical failures in 1991 and 33% of clinical failures in 2000. For comparison, clinical failures due to poor efficacy/PD were just under 30% at both time points. 

---

Egan, W.J. (2007) Computational models for ADME. Annual Reports in Medicinal Chemistry, 42, 449—467. 

Product property prediction Limited set mainly available in daylight Limited set Very large collection, including many vendor-supplied and internally developed in silico models for ADME T end points and target SAR... 

The problems of today s in-silico models for ADME properties are mainly associated with data availability, consistency, and quality to solve particular problems. Accurate and consistent date is an indispensable prerequisite for predictive models. In particular the use of multi-mechanism data often results in failures... 

Whereas hard filters can be considered to be knowledge-driven, soft filters are the result of a data-driven approach. A quantitative structure-activity or structure-property relationship (QSAR/QSPR) is established to predict a property from a set of molecular descriptors. Examples are the above-mentioned in-silico prediction tools for frequent hitters [27] and drug-likeness [41,42] additional models for ADM E properties are described below. 

Today, researchers continue the quest to estabUsh correlations between in vitro assays and in vivo observations [96,97]. Work is also underway to use animal data collected from in vivo experiments along with data generated by human in vitro systems to predict the metabolic behavior of compounds in humans [98]. The amount and diversity of data collected will be used to develop computational models for ADME properties. In the future, computational screens may replace many of the experimental screens run today. 

Computational models for ADME properties work best when the models are based on single mechanism experimental assays. Scientists approaching an ADME compu-... 

The increased level of interest in computational models for ADME/Tox is influenced by the need that multinational pharmaceutical companies now have to lower costs and to increase the width of their discovery pipelines. The risks and expenses associated with developing drugs has increased to the point where companies have pulled forward to early development stages screens for ADME/Tox properties. These companies have started to utilize computational filters early on in their bid to lower costs and to improve the efficiency of developing molecules. 

---