DRUG DISCOVERY TODAY DISEASE MODELS The use of dynamic computational models of neural circuitry to streamline new drug development Jeffrey E. Arle 1,2,3, * , Kristen W. Carlson 1 1 Department of Neurosurgery, Beth Israel Deaconess Medical Center, Boston, MA, United States 2 Department of Neurosurgery, Harvard Medical School, Boston, MA, United States 3 Department of Neurosurgery, Mount Auburn Hospital, Cambridge, MA, United States We propose a relatively novel use of dynamic models of neural circuitry in the process of new drug develop- ment for neurological disorders. A neural circuit model of depressive disorder was developed. Differences in synaptic activation represented variations in drug bind- ing affinity strength for ten putative molecules. Circuit dynamics led to changes in firing rates compared to normal and depressed baselines. Differing abilities to affect circuitry dynamics not linearly related to binding affinity were appreciated, allowing molecules to be prioritized earlier for future study. Animal and human trial planning can use this information to streamline drug development, conserving cost and time. Section editors: William W. Lytton – Department of Physiology and Pharmacology, SUNY Downstate Medical Center And Department of Neurology, Kings County Hospital Center, Brooklyn, New York, USA. Samuel A. Neymotin – Department of Neuroscience at Brown University, Providence, Rhode Island, USA. A recent comprehensive study by DiMasi et al. [1] of the cost of developing 106 drugs found that out-of-pocket research and development (R&D) costs averaged $1.4 billion per drug, while including the cost of capital employed raises the aver- age total to $2.5 billion, in line with other similar studies [1– 3]. Despite the advent of new drug development tools, this study found costs had increased at an average rate of 8.5% per year above the rate of inflation [1]. Key to their methodology and understanding of the under- lying reasons for high costs is tying the costs of unsuccessful drugs to the costs of the successful, eventually-approved drugs. The clinical success rate reported in their 2003 study was 21.5%, while the 2016 study found the rate had dropped to 11.8%, consistent with success rates reported in other studies [1,5]. Offsetting declining clinical success rates is the ability of companies to abandon drug candidates predicted to fail earlier in the process [1,4] and use computational models to provide early inexpensive predictions [5]. The average time from drug candidate synthesis to mar- keting approval is typically over 10 years [1]. In general, R&D costs rise throughout the development cycle and thus the earlier a company can eliminate candidates likely to fail, the greater the savings and lower the opportunity cost [1,6]. In silico neural circuitry testing in early drug development Given the staggering, increasing expense of drug R&D, the R&D timeline, and the probability that animal models may fail to predict human trial results, we propose the Drug Discovery Today: Disease Models Vol. 19, 2016 Editors-in-Chief Jan Tornell – AstraZeneca, Sweden Andrew McCulloch – University of California, SanDiego, USA Computational Models of Neurological Disorder *Corresponding author: J.E. Arle (jarle@bidmc.harvard.edu) 1740-6757/$ © 2017 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ddmod.2017.01.002 69