Molecular docking using genetic algorithms

Genetic algorithms are applied to dock molecules to binding siies of known structure. Intermolecular interaction energy is minimized using the genetic algorithm approach for the docking process. A software package has been developed especially for handling intennolecular energy minimization problems by GA optimization. Interrnolecular interactions of both small organic dimers and a larger molecular complex of an anticancer drug have been investigated. The performance of GAs on molecular docking calculations is discussed and compared with numerical methods. The results of implementation indicate that the GA approach is superior to conventional methods used in energy minimization when there exist many local minima as well as a global minimum. Significant improvement over conventional computational methods for molecular docking can be seen when applying GA to large molecular systems. I n t r o d u c t i o n Molecular docking has proved to be a useful tool for molecular recognition studies, yet it is a difficult problem both in terms o f understanding the primary determinants and developing computational methods. A good docking procedure may have practical applications in rational drug-design, identification of functionally related surface structures, prediction of molecular complexes, computer simulation of molecular association between substrate and enzyme, and so on. One molecule can be docked into another molecule in a lock-and-key fashion due to the attractive forces between two molecules with matched georfletry shapes, l.ntermolecular forces, in principle, can be obtained from quantum mechanical calculations,J1] but usually, these calculations are prohibitively lengthy. At present this quantum mechanical approach is not practical for systems of significant size because of computer limitations. Instead, the intermolecular energy may be represented by an empirical "force field" from which the ensemble intermolecular energy may quickly be calculated. [2] An additional serious problem has to be solved in finding optimum molecular docking: the energy hypersurface for systems beyond the simplest has many subsidiary energy minima as well as a global minimum. The major difficulty to be overcome in predicting the structure of molecular clusters and docking through energy minimization is the challenge of searching the large and complex energy hypersurface to find the most stable or global energy minimum. Various approaches to optimization of functions are well documented in literature.J3] The commonly used methods in potential energy minimization are numerical methods (steepest descent, Newton-Raphson). These conventional methods generally fail to accomplish the task of finding a global minimum of the intermolecular energy Permission to copy without fee all or part of this material is granted provided that the copies are not made or dixtributeal for diroct commercial advantagc, the ACM copyright notice and the title of the publication and its date appear, and notice is given that copying is by permission of the .Association for Computing Machine~. To copy otherwise, or ,to republish, requires • fee ~ndlor specific permission. O 1994 ACM 0~791-647-6/ 94/ 0003 $3.50 function.J4] Recently, genetic algorithms (GAs) [5, 6, 7] have been showing broad promise in many application areas. In this paper we report our work on efficiently finding global intermoleeular energy minima of docking using a genetic algorithm approach. First we will briefly address molecular docking problems related to rational drug design. Then the current computational approaches to minimization of intermolecular energy are discussed. Next the implementation of our genetic algorithm is described. Also the results of our application . are discussed in some detail. Meanwhile, comparison is made between these results and pi-evious work done by numerical methods. And finally, the conclusion of our study is provided. Molecular Docking in Rational Drug Design Drug developers have long sought out new approaches for improving the drug discovery process. The industry now spends multi-billions of dollars a year on drug research and development. As the drug development pipeline has become an even more critical factor in drug industry, improving the productivity of drug R&D by incorporating new approaches in drug discovery has been a major means. Much of the emphasis in improving the drug R&D process is placed on rational methods of drug discovery and design, encompassing techniques of computational chemistry, NMR spectroscopy and X-ray crystallography. For instance, drugs find their own binding sites in hemoglobin and influence the equilibrium between its two alternative structures in unexpected ways.[8] Drugs dock into niches of the protein which fit their van der Waals volume and take up positions that minimize the intermolecular potential energy. The major intermolecular interactions between drug and protein are electrostatic including strong hydrogen bonds and weak interactions between aromatic quadrupoles and nonpolar interactions between aliphatic hydrocarbons. Elucidation of the binding sites of drugs in proteins is apparently important when designing new drugs. Molecular docking, the integral part of molecular recognition between two molecules, is achieved through complementarity of molecular surface structures and energetics. This complementarity can take many forms: charge-charge interaction, hydrogen bonding, van der Waals' interaction and the size and shape of surfaces. Among these features, welldefined geometrical relationships among the individual molecules of the formed protein complexes are the fundamental considerations of the binding or association. Many works done for protein docking problem are based on this assumption.J9, 10] In a purely geometry docking, the procedure involves analysis of molecular surfaces and shapes, matching the complementary shape and size. This is an optimization of the position of molecules relative to each other; an optimization procedure can reduce the computational problems associated with the docking. However, a purely geometry optimization procedure may