Combining Non-Negative Matrix Factorization with Molecular Energy Landscape Analysis for Structure Quality Estimation of Proteins
摘要
Determining the intrinsic organization of the molecular structure space is crucial in various computational analyses of proteins. Identification of the inherent structural groups plays a vital role in this regard. The grouping of tertiary structures of a protein has broad applicability that includes rendering structure-energy landscape, capturing stable and quasi- stable structural states, modeling structural dynamics, and relating them to biological function. Its applicability can be extended further by utilizing the structural groups as the main ingredients to develop methods for single structure selection as well as for structure quality assessment. Exploring along the direction, this paper puts forward a novel method that aggregates structure proximity matrix factorization with structure-energy landscape analysis. Even though the method has several utilities, this article focuses on demonstrating its effectiveness in addressing the structure quality assessment problem.