Browsing by Subject "restraint"

We have developed a software for homology modelling by satisfaction of distance restraints using MODELLER back-end. The protocols used extend exploration of distance restraints and conformational space. We drive the models in optimization cycle towards better structures as assessed by the used metrics on DOPE score, retrospective distance restraint realization and others. Hydrogen bond networks are optimized for their size and connectivity density. The performance of the method is evaluated for its ability to reconstruct GPCR structures and an extracellular loop 2. The software is written in object-oriented Python (v.2.7) and supports easy extension with additional modules. We built a relational PostgreSQL database for the restraints to allow for data-driven machine and deep learning applications. An important part of the work was the visualization of the distance restraints with custom PyMOL scripts for three-dimensional viewing. Additionally, we automatically generate a plethora of diagnostic plots for assessing the performance of the modelling protocols. The software utilizes parallelism and is computationally practical with compute requirements on an order of magnitude lower than those typically seen in molecular dynamics simulations. The main challenges left to be solved is the evaluation of restraint goodness, assigning secondary structures, restraint interconditioning, and water and ligand placement.