February 1, 2005 (Vol. 25, No. 3)

Implementing ADME-Tox Profiling Earlier in Discovery

Fail fast, fail cheap is a common clich in drug development. Studies project that anywhere from 4060% of potential drugs candidates fail in development due to absorption, distribution, metabolism, excretion, and toxicology (ADME-Tox) problems, and that per drug, companies allocate about one-fifth of their R&D budgets to evaluate ADME-Tox compounds that do not reach the clinic.

In discovery, potential lead compounds are not advanced through the pipeline based upon ADME-Tox profiles, but upon their affinity to a biologically relevant target.

But now many drug companies are integrating ADME-Tox programs earlier in discovery to assess ADME-Tox properties of compound libraries and high throughput screening in parallel. In addition to re-engineering the pipeline, companies are developing new, physiologically integrated cell-based screening methods and in silico predictive models, and creating new high throughput technologies.

Predicting Bioavailability

“Predicting human bioavailability is a major challenge in drug discovery,” states Kuo-Chi Cheng, Ph.D., a research fellow who, with colleagues at Schering-Plough (Kenilworth, NJ) developed an in vitro Caco-2/hepatocyte hybrid model system to predict oral bioavailability (F; an F equal to 1.00, for example, is equivalent to 100% bioavailability).

The system uses a two-compartment transwell device. The donor compartment contains Caco-2 cells, and the receiver compartment contains hepatocytes. The system’s goal is to mimic the in vivo system where a compound gets absorbed through the GI tract and is metabolized by the liver.

The liver can be damaged by potential pharmaceutical and environmental insults, thus biliary excretion and hepatic uptake of potential drug candidates in liver assay systems must be evaluated.

Liver injury can occur directly by a toxin, indirectly by a xenobiotic metabolizing into a toxin, or by a drug or metabolite acting as a hapten complexing with an antigenic carrier and converting a cell protein, for example, into an immunogen.

The Caco-2/hepatocyte hybrid system was used to test 24 randomly chosen marketed drugs by comparing the area under the curve (AUC), a measure of the exposure of the body to the drug with in vivo oral bioavailability of that sample set reported in the literature.

Linear regression analysis showed reasonable correlation (R2=0.86) with in vitro AUC and oral bioavailability reported in the literature. Data suggest that this hybrid system is useful for predicting oral absorption and first-pass human bioavailability.

Early Discovery Screens

“Accessibility, cost, and throughput of various methods for testing ADME-Tox properties differ between the properties one needs to evaluate,” says Ian J. Mehr, Ph.D., head of business development at Qualyst (Research Triangle Park, NC).

“For example, high throughput in vitro analysis is currently available for metabolic stability (e.g., Cyp450 liability), but excretion studies are typically performed in vivo, after deciding to promote a compound for development.

“Ideally, excretion would be assessed in vitro proactively, and be used to decide candidate promotion, rather than coming into play after a candidate is promoted, forcing a reactive response to drug transport challenges when they arise later in development and are much more costly.”

Qualyst developed B-Clear, an in vitro hepatobiliary disposition model, to screen candidate compounds for susceptibility to biliary excretion. B-Clear uses cultured hepatocytes sandwiched between two collagen layers to create a bile canaliculi network, which are channels formed by grooves between plasma membranes of facing hepatocytes.

Intracellular actin and myosin microfilaments surround each canaliculus to help propel secreted biliary fluid along these channels. The system provides physiological data on hepatic drug transport, including uptake, excretion, and biliary clearance, to help eliminate compounds with an undesirably high susceptibility for biliary excretion from further evaluation as therapeutic agents early in drug development.

B-Clear is deployed as an early discovery screen for promotion of compounds, as a source of lead optimization guidance, and as a tool to address development-stage drug transport questions.

“Drug companies are performing ADME/Tox studies earlier in drug discovery, and using computational approaches to do it,” states Sean Eakins, Ph.D., vp at GeneGo (St. Joseph, MI).

“Key questions include which of many possible lead series does a company pursue that will provide, not only the potency, but also a good balance of ADME-Tox properties that will have a high probability of success further down the pipeline. Molecules need to be screened by considering many different properties, and those properties can be predicted and/or measured.”

Genego’s MetaDrug platform of computational technologies aids in predicting human drug metabolism and in visualizing high throughput data in the context of a complete biological system.

MetaDrug consists of rule-based prediction of metabolites, quantitative structure activity relationships (QSAR) to predict affinity to various enzymes, transporters, and other ADME-Tox properties, a large database of human, curated drug metabolism data, and algorithms to generate networks of genes and molecules using omics data input on top of the empirical data.

MetaDrug’s human specific database of molecules with drug metabolism data, used to derive rules to predict metabolism, can be searched to find molecules similar to those for which predictions are being made. The platform allows users to make in silico predictions and upload microarray data to assist in understanding implications of particular genes being up or down regulated as they can generate interaction networks.

Quality of Analysis and Selection

“Drug discovery and development involve many iterations of analysis and testing, and analysis means much more than getting statistical correlations,” states Alanas Petrauskas, Ph.D., president, Pharma Algorithm (Toronto).

“One has to understand the causal reasons of positive and negative results, suggest viable compound modifications, predict many different ADME-Tox effects for new compounds, extrapolate predictions to physiological levels, identify possible new limiting factors, and select proper experimental assays.”

The company’s Advanced Algorithm Builder consists of two parts: data analysis tools (algorithm builder) and ADME-Tox predicting tools (ionization, solubility, absorption, oral %F, inhibition, dose dependences, acute and chronic toxicity, mutagenicity, and organ-specific effects, among others).

Analytical tools include automated conversion of experimental data into predictive algorithms. This enables high laboratory automation and allows analyzing experimental data using a knowledge-based approach to understanding the limiting factors that can direct new experiments.

The tools speed up these repetitions by helping “automate” complex decision-making processes. “They differ from traditional informatics tools in that they help to generate new knowledge rather than plain statistics,” says Dr. Petrauskas.

“The problem of conventional drug design is that automation compromises quality of analysis and compound selection. Discovery driven by plain statistics is really driven by chance. Without good direction of how to interpret new experimental results, you cannot make good drugs. Virtual screening will become a lot like cherry-picking,’ based on the ability to predict most of ADME-Tox effects in silico.”

Cyprotex’ (Manchester, U.K.) Cyprotex Lead Optimization Engine (Cloe) Screen is an ADME and physicochemical property in vitro screening facility that delivers detailed information on individual compound properties via high throughput robotics and information technology.

Cloe PK prediction software integrates in vitro ADME and physicochemical data to deliver plasma and tissue concentration/ time profiles and PK parameter summary statistics. It enables ranking of compounds based on a compound’s likely pharmacokinetics in humans or rats, and provides guidance for scientists to improve pharmacokinetics of individual compounds.

Cloe PK and Screen technologies are integrated with Cyprotex’ in-house predictive ADME capability to develop QSAR models from relevant in vitro data within hours of generating that data. These technologies allow users to understand the level and duration of compounds in the body following a particular dose to provide an important basis for interpreting other toxicity and pharmacology data.

Combining Cloe PK results with potency data (such as an IC50 inhibition value), allows determination of the potential efficacy of a compound from a particular dose. Or, if the concentration of a compound that delivers a toxic effect is known, one can see whether or not this level from a particular dose will ever be reached in vivo.

“A key challenge is interpreting in vitro data and extracting the greatest value from it, states Karen Jones, Ph.D., product marketing manager at Cyprotex. “If you have lipophilicity, permeability, metabolism, and solubility data in front of you, which one do you choose to work on first? You need to understand how these properties interplay and consider pharmacokinetics as a whole.

“As in silico technologies become more accepted and better integrated, they will provide a way to consider ADME/PK in drug design stages, so compounds that are synthesized have greater potential for success because they already have good PK properties. This should decrease the need to synthesize vast libraries when this capability is combined with design programs that consider biological properties.”

Lead Optimization Strategy

“The trend to bring ADME-Tox in earlier in discovery is moving to prime time,” believes Robert S. De Witte, business relationship manager, ADME-Tox, life and laboratory science, at Thermo Electron (Burlington, Canada).

Thermo Electron’s ADME-Tox platform integrates software and laboratory automation (Please see “Thermo Introduces New ADME-Tox Platform,” on p. 38). The software component, the Orchestrator, marshals the flow of samples and data through the system. Samples are selected to be assayed, and plates are formatted There are configurable pass/fail criteria for conditional assays. For example, a stability study may be done if solubility is greater than a threshold value.

Ideally, De Witte would like scientists to be able to characterize all of their hits. “Whether or not ADME-Tox is a bottleneck depends on your strategy,” he says. “If the strategy is to characterize every hit, existing capacity is insufficient. If your strategy is to winnow hits down and devote more resources to the ones you’ve chosen to advance, then ADME-Tox becomes less of a bottleneck.”

Ultimately De Witte would like to see lead optimization made more efficient to ensure that compounds are in better shape in the preclinical stage. “I’d like to see improvement in metrics, like the number of medicinal chemists needed to advance a drug to the clinic, and I’d also like to see a drop in clinical attrition rates,” he states.

De Witte feels that a major challenge is not so much a data deluge, but correctly interpreting data. “Once reliable data are delivered, the challenge becomes how to understand and interpret that data. Another challenge is to develop reliable in vitro toxicity testing; we still rely on in vivo models. In terms of cost and throughput, the industry is not yet where it needs to be.”

Previous articleBuilding Momentum in Scandinavian Cluster
Next articleMerops: The Peptidase Database