Protein Structure Classification and Loop Modeling Using Multiple Ramachandran Distributions

• DOI: 10.1016/j.csbj.2017.01.011
• Publication Date: 2017-02-08T14:15:30Z
• Authors: Seyed Morteza Najibi; Mehdi Maadooliat; Lan Zhou; Jianhua Z. Huang; Xin Gao
• Keywords: Ramachandran plot; Representation

Recently, the study of protein structures using angular representations has attracted much attention among structural biologists. The main challenge is how to efficiently model continuous conformational space based on differences and similarities between different Ramachandran plots. Despite presence statistical methods for modeling data proteins, there still a substantial need more sophisticated faster tools large-scale circular datasets. To address this need, we have developed nonparametric method collective estimation multiple bivariate density functions collection populations backbone angles. proposed takes into account nature trigonometric spline which efficient compared existing methods. This approach widely applicable when estimate from with common features. Moreover, coefficients adaptive basis expansion fitted densities provide low-dimensional representation that useful visualization, clustering, classification densities. provides novel unique perspective two important challenging problems in structure research: structure-based angular-sampling-based loop prediction.