Finding multiple target optimal intervention in disease‐related molecular network

Abstract

Drugs against multiple targets may overcome the many limitations of single targets and achieve a more effective and safer control of the disease. Numerous high‐throughput experiments have been performed in this emerging field. However, systematic identification of multiple drug targets and their best intervention requires knowledge of the underlying disease network and calls for innovative computational methods that exploit the network structure and dynamics. Here, we develop a robust computational algorithm for finding multiple target optimal intervention (MTOI) solutions in a disease network. MTOI identifies potential drug targets and suggests optimal combinations of the target intervention that best restore the network to a normal state, which can be customer designed. We applied MTOI to an inflammation‐related network. The well‐known side effects of the traditional non‐steriodal anti‐inflammatory drugs and the recently recalled Vioxx were correctly accounted for in our network model. A number of promising MTOI solutions were found to be both effective and safer. ### Synopsis Common disorders such as cancer, inflammation, diabetes and cardiovascular disease often result from multiple molecular abnormalities. Single‐target drugs may not help in such cases because they cannot exert an effect on all abnormal molecules. Combination drug therapy and traditional remedy, the fruit of a long history of clinical experience, are major treatment measures to control these complicated diseases. They regulate multiple drug targets that are involved in the disease and gain some successes in the treatment ([Ahren, 2008][1]; [Xiao and Yang, 2008][2]). These successes and the development of high throughput experiments promote the idea of a system‐oriented drug design. This novel strategy, taking into account the intrinsic properties of biological systems, placed importance on restoring the normal balance of the system by regulating multiple drug targets. Growing attention has been paid to this field, and several marketing successes of multicomponent therapies have already been reported, for example, salmeterol/fluticasone (Advair; GlaxoSmithKline) ([Nelson, 2001][3]), nicotinic acid/lovastatin (Advicor; Kos Pharmaceuticals) ([Gupta and Ito, 2002][4]; [Bays et al , 2003][5]) and AZT‐3TC (Combivir; GlaxoSmithKline) ([Larder et al , 1995][6]). Although high‐throughput experiments have gained eximious achievements in system‐oriented drug design ([Zimmermann et al , 2007][7]), theoretical approaches call for practical and systematic computational strategies for multi‐target drug design based on disease network modeling. In this study, we have developed a computational method for finding multiple‐target optimal intervention (MTOI) solutions in a disease network. For a given disease network, the method tries to identify effective drug targets and the combination of multiple invention that can best restore the disease network to a desired normal state. Instead of focusing on a single drug target, the MTOI method analyzes the relevant network as a system to extract information about the inter‐conversion of the disease and the normal state of the network. The flowchart of MTOI is shown in [Figure 1][8]. As a concrete example, we have applied MTOI to an inflammation‐related network—the arachidonic acid metabolic network (AAnetwork). Inflammation is a tough disease in which its major medicaments, non‐steroidal anti‐inflammatory drugs (NSAIDs), have severe side effects on the gastrointestinal tract ([Ambegaonkar et al , 2004][9]). The later introduced selective cyclooxygenase‐2 (COX‐2) inhibitors are found to have cardiovascular side effects ([Kusmak, 2005][10]). The withdrawal of Vioxx (rofecoxib) is a great shock to the pharmaceutical industry ([Arellano, 2005][11]). Notwithstanding these failures, the long history of the anti‐inflammatory struggle has accumulated abundant experimental and clinical data, which allowed us to construct a mathematical model for inflammation. With this model, we simulated the drug effects of popular anti‐inflammatory medicines, such as aspirin, Vioxx, etc. The known bleeding or cardiovascular side effects of these drugs were correctly reproduced ([Table III][12]). We applied MTOI to the AAnetwork and identified five anti‐inflammatory drug targets. A number of optimal multi‐target strategies were found that are both effective in controlling the inflammation mediators and safe with minimized side effects, such as the simultaneous control of COX1/2 and leukotriene A4 hydrolase (LTA4H). Given sufficient knowledge about the disease‐related network, our results suggest that computational methods can play an important and unique role in the new era of network‐based drug design. We have demonstrated the utility and power of MTOI with the robust identification of the drug targets and the multi‐target intervention solutions that are both effective and safe. With the fast accumulation of experimental data and more disease networks revealed, we expect an increasingly critical role for computational approaches in system‐based drug design. Mol Syst Biol. 4: 228 [1]: #ref-2 [2]: #ref-61 [3]: #ref-36 [4]: #ref-20 [5]: #ref-8 [6]: #ref-32 [7]: #ref-63 [8]: #F1 [9]: #ref-4 [10]: #ref-30 [11]: #ref-7 [12]: #T3