Several comparative studies of docking programs show the poor accuracy of some of the commercially available packages. This failure results in part from the use of an inaccurate protein model. Although inhibitors are in general accurately docked back to their corresponding protein structure (self-docking), docking to other structures (cross-docking) usually performs poorly. Several docking programs treat the proteins as rigid object and do not account for conformational changes upon binding which results in poor performance in cross docking studies. The use of the rigid protein model usually causes an inaccurate prediction of ligand/affinities activities and low enrichment factors in virtual screening studies.
FITTED aims at improving the accuracy of existing molecule docking software program. It uses a more accurate protein models and is based on a pharmacophore-oriented docking method combined with a genetic algorithm based docking approach. The later takes advantage of more than one structure to dock compounds in virtually flexible proteins. A library of experimentally observed protein conformations is used and composite structures are created in order to model the protein flexibility and to explore a wide region of conformational space. The proteins, ligands and potential bridging water molecules are described as genes and a mixed Lamarckian/Darwinian evolution optimizes the whole complex.
Docking methods are computational techniques that are able to predict the binding mode of ligands (e.g., enzyme inhibitors, receptor antagonists) to their biological targets. Combined with a scoring function, these methods can be used to screen large libraries of compounds for their affinities for a specific pharmaceutically relevant target. FITTED is a program for the docking of small-molecule ligands onto flexible proteins.
The main module (FITTED) docks the ligands into the flexible protein in presence of displaceable water molecules. A complete description of the theory and modules of FITTED 1.0 can be found in the original reference .
The current version of the suite includes additional features and will be reported soon:
Automatic interaction site generation
Automatic binding site identification
Interaction site / force field consensus docking
Matching algorithm for initial population generation
Pharmacophore oriented docking
Force field parameter estimator
Generalized AMBER force field (GAFF) implementation
Two highly predictive scoring function (RankScore 2 and RankScore 4)
Filters for drug-like molecules
Improved atom typing
Improved atom charges for the ligands
Semi-flexible protein docking with flexible waters
United atom representation of the ligand protein non-bonded interactions
FITTED, an Algorithm To Account for Protein and Ligand Flexibility
FITTED is based on a genetic algorithm (GA) with an emphasis on speed. This docking software is unique in that it can deal with both the flexibility of macromolecules and the presence of bridging “displaceable” water molecules. Additional operators to the more traditional crossover and mutations were implemented and led to a significant increase in speed. These operators simulate the learning (through energy minimization at various stages) and the early selection of individuals based on a crude estimation of their fitness (e.g., is the ligand in the binding site?).
Several modes within FITTED allow it to solve various computational problems which may arise: Docking, Virtual Screening, Scoring and SAR. FITTED is able to dock using various modes of protein flexibility: rigid docking (Rigid), semiflexible docking (Semiflex), semiflexible docking with movable waters (Flex_Water) and fully flexible docking (Flex). Within FITTED there are two scoring functions used, one applied during the docking (DockScore) and another used for a final scoring (RankScore2 or RankScore 4).
A genetic algorithm is a global optimization technique. In the present case, it is used as a conformational search tool, allowing for the sampling of flexibility of complex systems. The first step is the creation of an initial population of poses, with many different conformations, also known as individuals. Each conformation can be described by a chromosome made up of genes; in this case each gene contains the value of a molecular descriptor such as torsional angles, position and orientation. Once the initial population is created the population evolves. Within this loop, parents (a pair of individuals in the population) are selected and coupled together using genetic operators, creating new conformations called children. With the children produced, the population is trimmed (some parents and kids are removed) and another reproduction cycle is started. This loop is continued until the population converges. This is the basis for the FITTED genetic algorithm.
