Over the past decade plus, a variety of in vitro assays have been developed for the prediction of ADME/Tox properties. These include, permeability, metabolic stability, solubility, cytochrome P450 inhibition, plasma protein binding, log D, metabolite identification, drug transport and several mechanisms of toxicity. Each of these assays have been developed using a variety of protocols and biological components. Most of these assays offer a considerable increase in assay throughput and reduction in cost relative to in vivo assays. However, most assays also have limitations and caveats which make them more suitable for some chemical series and not others. Moreover, the mechanism of action for the ADME endpoint can switch within a chemical series which can confound data interpretation. The state of the science will be reviewed, caveats and limits discussed and several new technologies profiled.

A number of molecular properties of interest for ADMET profiling are obtained from in-vitro assays of various complexity and throughput. In the past decade, a number of traditional, low throughput assays have been converted to higher throughput format and allow screening for a larger number of compounds for lead selection and optimization. While high-throughput in profiling does not have the same meaning that in HTS, it allows to characterize hundreds of compounds a week instead of a handful using standard technology. How useful for decision making are these assays at lead selection and during lead optimization profiling? In the last few years, we have gained some answers to these questions through experience with real life drug discovery programs. While not all assays are relevant in one particular program, profiling assays showed useful to guide chemistry decisions at the lead selection or during lead optimization. For logistics, sample management and economic reasons, we found useful to pool assays by packages We will illustrate how the various assay packages and the associated decision trees have been used to guide chemistry decision(s) using concrete examples.

At GSK efforts are underway to determine how best to leverage high throughput assay technologies to generate large, off target data sets to assist with rational compound design. With late stage attrition of otherwise promising compound candidates a common problem, there is a perception that access to early information relating to key ‘developability’ targets, even at the hit to lead stage could be used to flag, priorities or discount active series and ultimately circumvent problems later in the drug development pipeline.

To this end, we have embarked on collaborative efforts between preclinical development and early stage discovery functions to better understand structure activity relationships around the interactions of small molecule ligands with various transporter proteins. Transporter targets with key roles in drug disposition, drug-drug interactions and essential physiological processes are chosen based on the expertise and experience of DMPK and Safety Assessment scientists and high throughput compatible assays are then developed by discovery scientists. In certain instances novel assay methodologies have been required to enable the production of this data.

Once such data sets are available, it is possible to design computational models to predict liabilities, use tool compound data sets to assess the physiological relevance and significance of the in vitro data and specifically screen out liabilities eg therapeutic area-specific drug-drug interactions, toxicity, undesirable PK etc. A vital part of this overall process is the communication of findings from high throughput experiments with the necessary caveats. A highly integrated, multidisciplinary effort is therefore essential for success.

Establishing quantitative relationships between molecular structure and broad biological effects has been a long standing challenge in science. Up until a short time ag