Loop Modeling

Loops are among the most difficult parts of protein structures to predict because of their conformational heterogeneity and flexibility. Often, their dynamics prevent them from being resolved entirely, leading to conspicuous gaps in publicly available structures. This lack of information can prevent critical structure/function relationships from being established. To address this, we developed a computational method, Hash/RCD, implemented in the BCL that rapidly generates diverse libraries of possible loop conformations in proteins with missing amino acid residues[1]. Our approach combines and assembles loop conformations that have been previously determined and deposited in the PDB using hash map, followed by fast minimization and optimization using a previously published algorithm called random coordinate descent. By combining conformational hashing with RCD, the strengths of each are maintained while their weaknesses are ameliorated.