Model behavior

Model behavior

Philip Burton looks at how ADMETRx is using data more effectively in drug discovery

It is well recognized that productivity has been disappointing in recent years in the drug discovery market. Despite significant increases in R&D spending, the number of new drug approvals has been flat. This has contributed to increases in new drug development costs and concerns about stagnation in the industry.

Given the interest in the productivity predicament, it is informative to carefully explore the nature of the problem and its historical context. Recent data shows that the number of INDs approved per year has kept pace with the increased R&D spending. This supports the contention that discovery has been effective in introducing potential new drug candidates which also tend to survive through phase I clinical trials.

This success is probably influenced by strategies implemented in the 1990’s to address an earlier productivity problem. In 1991, when major causes of clinical failure of drug candidates were evaluated, 40 percent were due to unacceptable pharmacokinetic characteristics unidentified in preclinical development. This led to the development and implementation of the in vitro ADME models to help identify these liabilities early and eliminate those molecules from further consideration. This constituted part of the ‘fail fast – fail cheap’ model resulting from high throughput profiling of early discovery compounds.

An analysis in 2000 found that ADME/PK failure had fallen to about 10 percent. These in vitro models have significantly reduced ADME/PK failure of clinical candidates. Clearly, this strategy is effective. So where, and why, are candidates failing now? Failure is occurring in later phases, primarily because of lack of efficacy and toxicity. Given that mechanism-based toxicity is expected early, the high failure may be influenced by idiosyncratic toxicities, which are harder to identify in preclinical studies. The lack of efficacy may suggest that the best candidates are not being advanced into clinical development.

It is tempting to speculate that the high attrition of later Phase candidates is due to the fail fast –fail cheap model. While the experimental models have been effective in reducing specific causes of clinical failure, they may have contributed to the elimination of viable clinical candidates, effectively throwing the baby out with the bath water.

How can this be improved?

Additional tools for predicting efficacy and toxicity are clearly needed. More effective models for utilizing all the data to improve decisions about which candidates to advance are also needed. It is the composite profile of a molecule, rather than any specific attribute, which dictates its success as a drug. Models that accommodate deficiencies in one characteristic but are compensated for in another will improve significantly the advancement of better development candidates.

In the case of ADME, a typical profile will consist of solubility, permeability, metabolism, protein binding and CYP inhibition data. Increasingly, affinity for drug transporters such as P-glycoprotein (P-gp), multi-drug resistance associated protein (MRP), breast cancer resistance protein (BCRP) and others, is also evaluated. The tendency has been to move these measurements earlier in the discovery process and obtain data on all compounds. Frequently, the quality of the data has been compromised in order to accommodate the quantity of compounds to be profiled.

After the decision of what and how to measure has been made, the challenge of using the data remains. The concept of minimally acceptable values for the individual properties has often been used to advance or reject candidates. A problem with this approach, particularly in HTS assays, is propagation of uncertainty error in the measurement. If each assay is 90 percent accurate in differentiating between acceptable and unacceptable values, by the time several properties have been selected for, the ‘yield’ of compounds meeting the desired profile is significantly reduced. The impact of accuracy on compound selection can be minimized by improving the accuracy of the assay, generally at the cost of throughput.

If one accepts that minimally acceptable values are context dependent with respect to other characteristics of the molecule, alternative methods of using the data to make advancement decisions are also necessary. Use of imprecise data evaluated in an inappropriate context runs the risk of eliminating potentially successful drug candidates. How many times has this happened already and has it contributed to the lack of approvable drug candidates? A company’s intellectual property deserves the best, careful consideration and evaluation using the best technologies available in making decisions.

How is ADMETRx meeting these challenges?

ADMETRx is a high value solution provider focused on multi-criteria decision-making models to facilitate more effective use of data, and providing high quality ADME data to support them. We have developed a multi-criteria decision-making tool to facilitate better use of data. The model considers both context dependence and interrelationships of the properties. Recognizing the importance of the highest quality data to make the advancement decision, we provide ADME profiling services to support the model – although a client’s own data may also be used. These constitute the ADMETRx ‘toolkit’ that is used to provide solutions to our client’s discovery challenges. The objective is to identify the winners earlier and advance only those into clinical development. This will ultimately lead to improved drug approvals and decreased overall cost.

Philip Burton is currently Chief Executive and Scientific Officer of ADMETRx. He received the Meritorious Manuscript Award in both 1994 and 1998 from the American Association of Pharmaceutical Scientists (AAPS), and was elected Fellow of the AAPS in 1999. He is currently a member of the American Chemical Society, AAPS, New York Academy of Sciences, AAAS and ASPET. He has served as Associate Editor for the Journal of Pharmacology and Experimental Therapeutics since 2000, and on the editorial boards of the Journal of Pharmaceutical Sciences and Chemical Biology and Drug Design. Dr Burton has published and lectured extensively and is internationally recognized as a thought leader in drug absorption mechanisms.