EADock

Article Name : EADock: Docking of Small Molecules Into Protein Active Sites With a Multiobjective Evolutionary Optimization

DOI: https://doi.org/10.1002/prot.21367 (opens in a new tab)

Abstract

EADock dock is new molecular docking software which focuses on the accuracy which is needed in structure based drug design. EADock is interfaced with the CHARMM package for energy Calculation and cordinated handling

Structure and Drug design

The number of resolved protein structure is growing rapidly due to the improvement in the crystallographic techniques.

Research in the structure based drug design is increased in past several years. Presently the most common techniques is VHTS and Complementary to that,in silico rational drug design suggest-based modification of a lead compound

The docking Problem

Both VHTS and RDD rely on the structure based drug design. Once the energeticallu most favourable binding mode for several ligand is identified, they can be ranked according to their estimated affinity / activity .This step is extremly sensitive to accuracy of the predicted binding model

Existing Approaches

All methods primarily rely on heuristic sampaling techniques to generate binding mode. Their ranking is then achived using a scoring function. several heuristics are use the can be classified into three categories

• Approaches derived from systemic Research
• molecular dynamics stimulations techniques
• stochastic methods

Standard molecular mechanics simulation techniques (molecular dynamics and minimization) are appealing because of their physical foundations, but are time consuming and not effective at crossing high free energy barriers within accessible simulation time.8 However, the reduction of van der Waals and electrostatic repulsions was found to improve the sampling by lowering conformational transition energy barriers.9,10

Stochastic methods (Monte Carlo, genetic algorithms, and tabu search) are general optimization techniques with a limited physical basis, and are able to explore the search space ignoring energy barriers.

Genetic algorithm / Evolutionary Algorithms are iterative stochastic optimization procedure where an initial population of solution is generated and evaluated with respect to a set of constrains, described by fitness functions. In the docking problem, the fitness function described the intermolecular action between the ligand and the receptor

Refer genetic algorithm of Coding train on youtube

Genetic algorithm / Evolutionary Algorithms

Scoring functions can be classified into three families

• Empirical scoring function
• Knowledge based scoring function
• Force-field based scoring functions

1. Emperical scoring function

Empirical scoring function are expressed as a weighted sum of terms arising from given molecular interactions, such as hydrogen bonds, ionic, and van der Waals interactions.The weighting factors are fitted on a database of complexes with known structures and binding free energies. Their transferability to complexes outside the training database is thought to be more limited when compared to force field-based scoring functions

2. Knowledge based scoring functions

The second family is based on potential of mean force that are derived from large datasets of experimental 3D structures

3. Force Field based scoring functions

Force Field based scoring functions is based on molecular mechanics force fields, summing the interaction energy and the internal energies of both partners. If the protein is kept rigid, its internal energy does not change and can be ignored, speeding up the evaluation of a binding mode. These scoring functions are usually sensitive to atomic coordinates, limiting their applications in cross-docking experiments. Softened van der Waals potentials have the advantage of being less sensitive to atomic coordinates in these cases, but also suffer from being less selective. As force field-based scoring functions are not trained on a set of complexes, a good trans-ferability to in real world applications can be expected. Two docking approaches using the CHARMM31 package were published previously.

EADock Algorithm

Most of the published algorithm on docking is based on VHTS which can be very fast and less accurate. The docking algorithm for RDD is proposed in this research paper, Genetic algorithm for Docking (EADock) provides a unique combinations of four methods.

EADock uses the combinations of two fitness functions

The first one, which neglects solvent effects, is used to drive the search toward local minima because of its efficiency and speed. These minima are then exposed to a more selective and computationally demanding fitness function, which includes the solvation free energy. This approach thus relies on the assumption that minima of the second fitness are also minima of the first, though their rank may be different

Second, a mechanism inspired by tabu search restricts the search space as the evolution proceeds, by storing a list of previously visited unfavorable docking poses and preventing the search from revisiting these poses, thus facilitating the exploration of new conformational space. This continuous update of the search space also ensures that the evolution does not converge to complexes that do not correspond to a minimum of the second and more selective fitness.

Third, the sampling is performed with operators that combine a broad and a local search of the conformational space. Some of these operators are semi-stochastic, dealing with rotations and translations. Other operators, called ‘‘smart operators,’’ aim at crossing energy barriers by transiently modifying the fitness landscape, in a physical and deterministic manner.

Fourth, aside from this flexible sampling framework, coordinates handling and energy calculations are delegated to the CHARMM package, for which a Java API was developed. EADock is thus able to use the latest improvements available in CHARMM, especially sophisticated solvation models such as GB-MV2.

Methods

Keyword / Abbrevations

• VHTS : Virtual high-throughput screening
• RDD : rational drug design