Computational Drug Discovery and Design

algorithms: (1) systematic and (2) stochastic search methods

[11]. Systematic search strategies entail small changes in the struc-

tural parameters that alter the ligand conformation in a gradual

fashion [10]. By exploring the energy landscape of the conforma-

tional space, the algorithm converges to a conformation that cor-

responds to the minimum energy solution. Systematic search

algorithms are effective in probing the conformational space; how-

ever, a local minimum conformation can be provided instead of the

global minimum conformation. This shortcoming can be solved by

executing several searches simultaneously starting from a broad

range of different conformations. In addition, as systematic search

methods explore all combinations of the structural parameters, the

number of possible combinations grows exponentially with the

degrees of freedom of the ligand, leading to the so-called combina-

torial explosion. Docking packages such as FRED and DOCK

address this issue using incremental construction algorithms,

which gradually build the ligand structure into the binding site

[12, 13]. Incremental algorithms break down the ligand structure

in several fragments and then add each part sequentially in comple-

mentary regions of the binding site until the whole structure has

been reconstructed. The conformational search step is performed

individually for the added fragments, decreasing the degrees of

freedom to be probed and avoiding combinatorial explosion.

In contrast, stochastic methods explore the energy landscape by

randomly changing the ligand structural parameters [14]. Stochas-

tic algorithms generate sets of diverse solutions, exploring a broad

range of the conformational space. This approach is useful for

avoiding confining the conformations at local minima, thus increas-

ing the likelihood of generating global minimum solutions

[10, 14]. Genetic algorithms (GAs), one of the most successful

applications of stochastic search strategies, are implemented in

widely used programs such as AutoDock and GOLD

[15, 16]. Genetic algorithms apply the principles of natural selec-

tion by encoding the initial conformation of the ligand in a vector

termed the chromosome. Taking this chromosome as a starting

point, the algorithm generates an ensemble of conformations cov-

ering a wide range of the conformational space. Next, the chromo-

somes with the lowest energy values are selected as starting points

for the generation of the next ensemble of conformations. By

repeating the GA routine several times, the mean energy of the

population is diminished by transferring favorable structural fea-

tures from one generation of chromosomes to another, decreasing

the energy landscape to be probed [10, 16].

1.1.2 Estimation
of the Binding Energy

In addition to predicting the binding conformation of ligands,

molecular docking algorithms apply scoring functions to evaluate

the binding energy of the proposed solutions [17]. The energy

Molecular Docking and Structure-Based Virtual Screening 33