Robust Elastic Network Model: Precise Prediction of Atomic Fluctuations in Protein Crystal Structures

Abstract

Elastic network model (ENM) based normal mode analysis has become popular as its capability and suitability for the study of protein dynamics. However, the existing ENMs often fail to reproduce the experimentally observed B-factor (i.e., atomic fluctuation) because of oversimplification of their force-fields. In this work, we have proposed a robust ENM (RENM) in which, for reflecting inter-connections accused from surrounding molecules in a unit cell, symmetric constraints based on crystal space group are applied to the representative single molecule as well as its intra-connections are also represented by using lumped masses and specific spring constants depending on the types of amino acids and chemical bonds, respectively. More than 500 protein structures are tested by RENM. Their results show better agreement with experimental B-factor without additional computation burden compared to those of traditional ENM. Moreover, the global spring constant is quantitatively determined as a function of temperature at 100K and 290K, which enables us to compute directly atomic fluctuations and vibrational density of state without any fitting process. Thus, RENM is expected to play an important role in understanding protein dynamics based on its crystal structure information.