MLR–ANN and RTO Approach to l-opioid Receptor-binding Affinity. Pooling Data from Different Sources GuillermoRamı´rez-Galicia 1,4, *, Ramo ´n Gardun ˜ o-Jua ´ rez 1 , Omar Deeb 2 and Bahram Hemmateenejad 3 1 Instituto de Ciencias Físicas, Universidad Nacional Autónoma de MØxico, PO Box 48-3, 62250 Cuernavaca, Morelos, MØxico 2 Faculty of Pharmacy, Al-Quds University, PO Box 20002, Jerusalem, Palestine 3 Department of Chemistry, Shiraz University, Shiraz, 71454 Shiraz, Iran 4 Current address: Universidad del Papaloapan, Circuito Central 200, Parque Industrial, 68301 Tuxtepec, Oaxaca, MØxico *Corresponding author: Guillermo Ramírez-Galicia, memorgal@gmail.com One hundred and six morphinan derivatives were taken from the Drug Evaluation Committee reports to propose several quantitative structure– activity relationship models to describe the l-receptor-binding affinity. After several proce- dures to reduce the descriptor number, 21 descrip- tors were selected for the descriptor pool by a complete Multiple Linear Regression methodology. In this procedure only three molecules were con- sidered as outliers. Several tests changing the relation between training:predicted sets were con- sidered to find the best relation between these sets. The higher the number of molecules in the predicted set the higher the predictive power was observed. The optimal number of descriptors was established using the Akaike’s information crite- rion and Kubinyi fitness function parameters. The Artificial Neuron Network methodology was applied to improve the Multiple Linear Regression best result. Finally, the regression through the ori- gin methodology was applied to establish the best model from the Artificial Neuron Network method- ology. The best quantitative structure–activity relationship model was proven to be independent of chance correlation. Key words: Artificial Neuron Network, cheminformatics, Multiple Linear Regression, quantitative structure–activity relationship, structure- based drug design, l-receptor-binding affinity Received 21 August 2007, revised and accepted for publication 2 Janu- ary 2008 Three opioid receptor subtypes, designated by l, d, and j, have been identified in the central nervous system and periphery (1,2), and they have been involved in a variety of physiological processes especially analgesia (3). There is abundant biological and structural information for known ligands. The action of the narcotic analgesics now avail- able can be defined by their activity at these three specific opiate receptor types. Narcotic analgesics are also classified as agonists, mixed agonist–antagonists, or partial agonists by their activity at the opioid receptors. Antagonists selective for the d-receptor modulate the development of tolerance and dependence induced by l-agonists such as morphine (4). Furthermore, other studies demonstrate that d antagonists, when delivered in concert with l-agonists, result in effective pain modulation without the negative side-effects usually associated with l-receptor activation (5). The l-receptor shows the highest affinity for morphine because its desensitization and internalization show two regulatory mechanisms thought to contribute to the development of tolerance to opioids (6,7). In the morphinan-based opioids the N-substituent effect domi- nates the efficacy at the l-receptor. There are several methods available for the development of novel opi- oid analgesics lacking the undesired effects of traditional l-mediated analgesics such as morphine. Coop and MacKerell suggested (8) that this success will depend on multidisciplinary research incorporating medicinal chemistry, molecular modeling, and pharmacology. A major factor in drug design is the field of quantitative structure–activity relationships (QSAR), which are mathematical equations relating chemical structure to their biological activity (9,10). A wide variety of descriptors including constitutional, topological, and quantum chemi- cal have been reported for QSAR analysis (11,12). Peng et al. (13) constructed 3D-QSAR models using the CoMFA methodology on a series of opioid receptor antagonists. Their mod- els showed excellent internal predictability and self-consistency. CoMFA analysis demonstrated that variations in the binding affinity of opioid antagonists are dominated by steric rather than electro- static interactions with the three opioid receptors. On the other hand, fentanyl analogs are selective agonists of the l-opioid receptor and several studies on their structure–activity relationship have been performed (14,15). These results led to a distinct pharmacophore of the binding site of the l-opioid receptor. However, these theoretical results cannot give a direct feature about how ligands bind with the l-opioid receptor. 260 Chem Biol Drug Des 2008; 71: 260–270 Research Article ª 2008 The Authors Journal compilation ª 2008 Blackwell Munksgaard doi: 10.1111/j.1747-0285.2008.00626.x