![conference banner](/images/conference-banner.jpeg)
Abstract
Protein models are useful in structure-based protein engineering applications, including designing drugs, redesigning enzyme specificity, and designing new folds in proteins. Predicting loop structures is considered the main challenge in protein 3-D structure modeling. The flexibility of loop regions dictates the need for special attention to their conformations. In this paper, we report on the implementation of iterative stochastic elimination (ISE) optimization technology for the ab initio modeling of protein variable regions (loops). The ISE algorithm was tested on a benchmark of 70 structurally refined loops. The median root-mean-square deviation (RMSD) of the loop residues was 1.5Å, with 80% of the targets conforming to the native with RMSD lower than 2.0Å. The median-backbone, heavy-atom global RMSD of the loop predictions were 0.67Å for short loops (4-6 residues), 0.88Å for medium loops (7-9 residues), 1.68Å for long loops (10-12 residues) and 2.76Å for very long loops (13-16 residues). In addition to the accurate modeling of short, medium and long loops, the current method provided us with ensembles of conformations, which are crucial for studying the dynamic nature of loops, mainly on the surfaces of proteins. The proposed technique could be incorporated into modules for generating homologybased models and for flexible docking.
Keywords: Loop prediction, iterative stochastic elimination, ensemble of loop conformations, optimization algorithm, comparative modeling.