PathMolD-AB: Spatiotemporal pathways of protein folding using parallel molecular dynamics with a coarse-grained model. (August 2020)

• Record Type: Journal Article
• Authors: Hattori, Leandro Takeshi; Pinheiro, Bruna Araujo; Frigori, Rafael Bertolini; Benítez, César Manuel Vargas; Lopes, Heitor Silvério
• Is Part Of: Computational biology and chemistry. Volume 87(2020)
• Publication Date: 2020-08
• Subjects: Canonical ensemble -- CUDA -- 3D-AB off-lattice -- Protein folding dataset; Chemistry -- Data processing -- Periodicals; Biology -- Data processing -- Periodicals; Biochemistry -- Data processing; Biology -- Data processing; Molecular biology -- Data processing; Periodicals; Electronic journals; 542.85
• DOI: 10.1016/j.compbiolchem.2020.107301
• Languages: English
• ISSNs: 1476-9271

Abstract

Highlights: A parallel canonical ensemble MD using a CG model running in a master-slave CPU-GPU architecture. Analysis of the impact of different protein sizes in the performance of the proposed approach. A dataset of spatiotemporal pathways of protein folding, suitable for the study of the PFP. A biological analysis was accomplished to predict structure of crystallized proteins. All the software developed are made freely available. Abstract: Solving the protein folding problem (PFP) is one of the grand challenges still open in computational biophysics. Globular proteins are believed to evolve from initial configurations through folding pathways connecting several thermodynamically accessible states in a free energy landscape until reaching its minimum, inhabited by the stable native structures. Despite its huge computational burden, molecular dynamics (MD) is the leading approach in the PFP studies by preserving the Newtonian temporal evolution in the canonical ensemble. Non-trivial improvements are provided by highly parallel implementations of MD in cost-effective GPUs, concomitant to multiscale descriptions of proteins by coarse-grained minimalist models. In this vein, we present the PathMolD-AB framework, a comprehensive software package for massively parallel MD simulations using the canonical ensemble, structural analysis, and visualization of the folding pathways using the minimalist AB-model. It has, also, a tool to compare the results with proteins re-scaled from the PDB. We simulate and analyze, as case studies, the folding of four proteins: 13FIBO, 2GB1, 1PLC and 5ANZ, with 13, 55, 99 and 223 amino acids, respectively. The datasets generated from simulations correspond to the MD evolution of 3500 folding pathways, encompassing 35 × 10 6 states, which contains the spatial amino acid positions, the protein free energies and radii of gyration at each time step. Results indicate that the speedup of our approach grows logarithmically with the protein length and, therefore, it is suited for most of the proteins in the PDB. The predicted structures simulated by PathMolD-AB were similar to the re-scaled biological structures, indicating that it is promising for the study of the PFP study.