Molecular design can be understood as the application of molecular modeling and simulations, among other HPC-based techniques of computational biology, to produce atom-level models of biological macromolecules.
Molecular modeling
Figure 1. Typical workflow to produce a comparative molecular model of a biological macromolecule.
In spite of the availability of experimental techniques to model the 3D structure of biological macromolecules, computational biology techniques to produce comparative models are very helpful when dealing with proteins of unknown 3D structure whose sequence identity to proteins with known 3D structure is above 30% (the safe zone). Figure 1 shows a typical workflow to produce comparative models of proteins whose 3D structure is not known.
Figure 2. Effective usage of comparative models must take into account the resolution of the template structure and the sequence identity with the target sequence (extracted from Baker and Sali, 2001)
Molecular modeling tools by satisfaction of spatial restrains, such as Modeller, are widely used to produce models of proteins whose 3D structure is not known. However, the effective usage of those models depends on the resolution of the used template and the sequence identity, between other important factors such a those depicted in Figure 2.
Molecular simulations
Using mathematical approximations such as Hamiltonian dynamics and force-fields, biological molecules to be simulated are depicted, atom by atom, into a three-dimensional box on the computer where the set of forces -defined in the force-field- interact with the atomic structure to produce a dynamical behavior of the system through time. Due to the high-end computer requirements to produce molecular simulations, HPC techniques such as parallel and programming are commonly used to implement molecular simulation engines such as NAMD. Figure 3 represents the CHARMM classical force field.
Figure 3. The classical CHARMM force field is composed by terms that takes into account bonding and non-bonding interactions.
Using the CHARMM and other classical force fields, the vectorial field F is obtained through the computation of the potential energy gradient that depends on the coordinates r1 to rn. The summation of the different energy terms expressed in the functional form of the force field (Figure 3), generates the potential energy of the molecular system’s conformation. The dynamical behavior of biological macromolecules can provide insight into the function of such complex systems. Applications of molecular simulations range from protein engineering to computer-based drug design and nanobiotechnology.
Computer Based Drug Design
Figure 4. Drug discovery phases. Drug discovery is a highly complex and very expensive process. With a total cost of near a thousands millions dollars during 10 years, attrition costs on the last phases are important to be diminished as much as possible. Modified from Rawlins, M. D., Nature Reviews Drug Discovery 3: 360-4 (2004)
Modern drug design efforts are focused towards reducing the attrition costs that occur during the late phases of the development of drug discovery. Thus, during the first phases of drug discovery (Figure 4), computational biology tools such as comparative modeling, docking and molecular dynamics are broadly used to the discovery of therapeutical targets, to identify hit compounds, to produce lead compounds and to evaluate important pharmaceutical properties such as ADME/TOX (Adsorption, Distribution, Metabolism, Excretion and Toxicity).
Docking: Evaluating the interaction energy between molecules
In order to determine the feasibility that two molecules can interact by complementarity of their physicochemical properties, computational biology tools such Docking can be used to get an estimation of the delta G involved during the interaction procedure (Figure 5).
Figure 5. Molecular docking allows the evaluation of the delta G interaction between two molecules. Modified from Kitchen, D. B., et. al., Nature Reviews Drug Discovery, 3: 935 – 949 (2004)
Computational biology programs such as Autodock are commonly used to produce different conformations of ligands interacting with proteins. Once those molecular poses are obtained, the delta G of interaction can be computed by producing the cycle shown in figure 5.