Fragment-Based Protein Structure Prediction, Where Are We Now?

Abstract

In the past decade, there has been an extensive advancement in the creation of methods for the design and prediction of protein structures. Expeditious growth in protein structure and sequence databases has charged the development of computational approaches for the prediction of structures. This review focuses on fragment-based strategy, a computational approach for the prediction of the three-dimensional structure of proteins. Fragment assembly has immensely improved protein structure prediction accuracy, especially of the single-domain proteins at the fold level. Fragment libraries are generated using the dihedral angles along with local structural information of known protein structures. This leads to the construction of a full-length polypeptide chain of a query protein using the fragments present in these libraries. The energy function of the proteins is minimized contributing to multiple conformations considering the backbone atoms and “centroid” side-chain pseudo-atoms using conformational sampling. Lastly, Monte Carlo simulation is performed for the sampling of the side-chain rotamers and reduction of energy for more precise and refined model construction. The quality of the fragments determines whether the native-like conformations generated are accurate or not. The future direction as well as tools like ROSETTA, QUARK, FRAGFOLD, M-TASSER, and AlphaFold2 that use fragment assembly for optimal structure prediction have also been described and compared in this review.