BMCL Digest Strategies to improve in vivo toxicology outcomes for basic candidate drug molecules Tim Luker a,⇑ , Lilian Alcaraz a , Kamaldeep K. Chohan a , Niklas Blomberg b , Dearg S. Brown c , Roger J. Butlin d , Thomas Elebring b , Andrew M. Griffin e , Simon Guile a , Stephen St-Gallay d , Britt-Marie Swahn f , Steve Swallow c , Michael J. Waring d , Mark C. Wenlock g , Paul D. Leeson a a AstraZeneca R&D, Medicinal Chemistry, Charnwood, Loughborough, Leicestershire LE11 5RH, UK b AstraZeneca R&D, Medicinal Chemistry, Mölndal, Sweden c AstraZeneca R&D, Safety Assessment, Alderley Park, UK d AstraZeneca R&D, Medicinal Chemistry, Alderley Park, UK e AstraZeneca R&D, Medicinal Chemistry, Montreal, Canada f AstraZeneca R&D, Medicinal Chemistry, Södertälje, Sweden g AstraZeneca R&D, DMPK Screening and Profiling, Alderley Park, UK article info Article history: Received 4 July 2011 Accepted 18 July 2011 Available online 27 July 2011 Keywords: Candidate drugs Attrition Predictive models Calculated descriptors Toxicology abstract A valid PLS-DA model to predict attrition in pre-clinical toxicology for basic oral candidate drugs was built. A combination of aromatic/aliphatic balance, flatness, charge distribution and size descriptors helped predict the successful progression of compounds through a wide range of toxicity testing. Eighty percent of an independent test set of marketed post-2000 basic drugs could be successfully classified using the model, indicating useful forward predictivity. The themes within this work provide additional guidance for medicinal design chemists and complement other literature property guidelines. Ó 2011 Elsevier Ltd. All rights reserved. It has been estimated that around 96% of nominated candidate drugs fail at some stage in development and that within these troubling statistics, toxicity has become the major cause of drug attrition. 1 Attention is now increasingly focused on this issue and significant effort has been devoted to bringing predictive toxicity screens earlier into the drug discovery process. There have been some successes (e.g., in vitro hERG potency, phospholipidosis and Ames), 2 however it is generally difficult to develop predictive mod- els for whole animal toxicity given the diversity of mechanisms that produce toxic outcomes. Nevertheless, once mechanism- based toxicity (from the primary biological target) is removed from the analysis, the quality of a candidate drug is largely a conse- quence of its physical and structural properties. These properties significantly impact not only on affinity for the intended biological target(s) and pharmacokinetic properties, but also non target-re- lated toxicological effects such as reactive metabolite formation, receptor promiscuity, and organ toxicity. The control of these compound related risks due to physical properties and molecular structure is at the discretion of the medicinal chemist within the confines of obtaining potency at the desired target. Several potential strategies have emerged to minimize general compound attrition and these have been recently reviewed. 3 A number of groups and companies have suggested preferable phys- icochemical spaces that enrich successful progression of oral drugs via control of properties such as lipophilicity, hydrogen bonding, molecular weight, molecular size and polar surface area (PSA). These provide guidelines and property boundaries that include the ‘rule of five’, 4 ‘4/400’, 5 ‘3/75’ 6 and more general concepts such as molecular obesity 7 and potency-lipophilicity separation (ligand lipophilicity efficiency (LLE)). 8 Whereas several of the guidelines refer to ADME properties, the ‘3/75’ guideline from Pfizer was specific to in vivo animal toxicol- ogy attrition. Interestingly, bases were at particular risk within the Pfizer compound set when the physicochemical properties c log P >3 and PSA <75 Å 2 were present. The high interpretability of these findings are attractive, particularly as some link to promiscuity rates were provided. 6 As part of a global compound attrition initiative within Astra- Zeneca, the ‘3/75’ guidelines were applied to a set of internal oral candidate drugs of the basic charge type from the last decade that 0960-894X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.bmcl.2011.07.074 ⇑ Corresponding author. E-mail addresses: Tim.Luker@astrazeneca.com, fxxtim@gmail.com (T. Luker). Bioorganic & Medicinal Chemistry Letters 21 (2011) 5673–5679 Contents lists available at ScienceDirect Bioorganic & Medicinal Chemistry Letters journal homepage: www.elsevier.com/locate/bmcl