A computational method that combines information-driven docking and genetic algorithm.

Method Source code Server

Here we use genetic algorithms to theoretically determine the three-dimensional structure of a protein-protein complex. In gdock we apply the tried-and-trusted information-driven docking paradigm; in which researchers input residue numbers that are known (or expected) to be participating in the interaction. This information is taken into account when generating the initial complex, and when scoring the generated complexes, thus making gdock a local-docking algorithm that uses restraints à posteriori.

The input consists of a protein receptor and a protein ligand PDB files and the aforementioned lists of residues; each individual in a generation of the genetic algorithm is a possible complex between the receptor and the ligand, in which the ligand position is represented by its euler angles and the cartesian coordinates distributed around the geometric center of the initial position.

The fitness function is the interacting energy of the individual, and the number of generations, rate of mutation and crossover, parameters that can be specified by the user. Once the epoch is done, the possible complexes are scored and clustered. On it’s latest benchmarking (v1.1.0) the software was able to obtain acceptable conformations for 119/227 complexes considering the top 1000 scored models. For more information about the method, the source code and the benchmarking, check the project repository.

Below you can see a complex obtained with gdock, the sticks represent the true-interface, used as restraints to guide the fitness function. This complex has an interface RMSD of 1.14 Å and was ranked as #317

The representation is interactive, rotate it! (:
2021 gdock. Some rights reserved.
Creative Commons License