6 DOF haptic feedback for molecular docking using wave variables Bruno Daunay, Alain Micaelli Commissariat ` a l’Energie Atomique (CEA) LIST / DTSI 92265 Fontenay Aux Roses, France bruno.daunay@cea.fr, alain.micaelli@cea.fr St´ ephane R´ egnier Laboratoire de Robotique de Paris CNRS - UPMC BP 61, 92265 Fontenay Aux Roses, France regnier@robot.jussieu.fr, Abstract— This paper presents a new method for a six degrees of freedom haptic feedback in molecular docking simulations in virtual reality. The proposed method allows real-time haptic interaction even in the case of classical molecular simulation which implies notoriously long computation time. These sim- ulations are classically used by the pharmaceutical industry (Sanofi-Aventis) and are based on the energetic description of atoms to estimate the interaction between a ligand and a protein. The haptic control scheme uses wave variables for a stable and robust teleoperation, and a transcription of the calculated energy into forces and torques for the manipulation of a flexible ligand around the binding site of a flexible molecule. This method can then be used with any energetic force field using a minimization process, thus avoiding the fastidious optimization of molecular simulation programs. I. INTRODUCTION Drugs are made of small molecules (ligands) which in- teract with proteins in order to inactivate them through a specific pocket (binding site or active site). The computa- tional process of searching for a ligand that is able to fit the binding site of a protein is called molecular docking. The conformation of the ligand in the binding site has the lower potential energy. The only informations provided by the used softwares during the simulation, are a visual return of the conformation of the molecules and the value of the involved energy. Because of the relatively low success rates of the docking for fully automated algorithms, including a human operator in the loop appears as a solution [1]. Interactive haptic feedback for molecular docking can give additional information on the behaviour of the forces present inside the receptor. The operator would then be able to feel the repulsive or the attractive areas and define the best geometry of the ligand. Note that this scenario can only function in the case of a real-time simulation. The method we use for describing proteins is called empirical. All the molecular interactions are approximated by the Newtonian theory, therefore this method allows to simulate big proteins in an acceptable computational time. In order to simulate their behaviour, several methods are used and differ according to their applications. The method we use is based on the minimization of the force field during the ligand manipulation. The goal is to reach the potential minimum but independently of time (to the contrary of the molecular dynamic simulation technniques). Fig. 1. Manipulation scene. The ligand (green molecule) has to be moved through the protein to the binding site. The protein will search for a stable conformation during the docking. The aim of our work is not to optimize the molecular simulators (as proposed in some other works [2], [3]) but to conceive a method that takes into consideration their speci- ficities. Indeed, the pharmaceutical engineers use softwares which are not real-time but which describe the interatomic interactions very precisely. Moreover, during their research, they use several force fields, each one being specific to a molecular property. Knowing that several force fields need to be minimized, that energetic interactions need to be described, and that the computing time for conformational changes is important, we developed a method allowing to feel the forces during a molecular docking using any molecular simulator based on a force field minimization process. This article is structured as follows: the first paragraph describes the force field and the simulation we use in order to evaluate both the interaction energy between the ligand and the protein and the conformational change of these two molecules. The second paragraph describes a simple force/position bilateral coupling in order to specify the different problems to overcome. Then we propose a stable method for the control scheme of such a simulation and show how the forces can conveniently be felt in order to make the operator “feel” the binding site’s force field. 2007 IEEE International Conference on Robotics and Automation Roma, Italy, 10-14 April 2007 WeC4.2 1-4244-0602-1/07/$20.00 ©2007 IEEE. 840