Rosetta Ligand-Protein Docking with Self-Adaptive Differential Evolution

Molecular Docking is a growing area of research due to its importance in structural-based drug design. Along the years, many computational methods have been proposed as a possible solution for the molecular docking problem, but none of them genuinely solved it. In computational theory, this problem is treated as an NP-Hard one, meaning that there are no efficient algorithms to find the best solution in polynomial time. In this way, metaheuristics became an option for finding feasible solutions in a viable computational time. Despite that, metaheuristics are parameter-dependent methods which must be tuned to find the balance between exploitation and exploration capabilities. In this sense, we used a Self-Adaptive Differential Evolution algorithm with four different mutation techniques in the flexible-ligand rigid-receptor molecular docking problem and thoroughly compared them with the single mutation versions - something not yet explored. Also, we used the Rosetta energy function as a score function for binding evaluation, which is not broadly used in most of the related works. Our results showed that Self-Adaptive Differential Evolution could reach competitive results in comparison with different algorithms that used the same energy function and molecular flexibility.