The Fractal Properties Of Protein Energy Time Series Based On Molecular Dynamics Simutation

By molecular dynamics simulation, we get the energy data of seventeen pro-teins from four protein classes: α+β α/β α and β, then regard them as timeseries. Firstly we convert these time series to non-negative, and use the multifrac-tal detrended fluctuation analysis method(MF-DFA)to analyze them. We fndtheir h(2) in the generalized Hurst index curves are greater than1, h(q) curvesdecrease with q, which indicates all energy time series are non-stationary andhave multifractal feature. In order to carry out traditional multifractal analysison them, we use the empirical mode decomposition (EMD) method to removethe trend of the protein energy time series, which have been converted to be non-negative. Then we use the MF-DFA to analyze these new series. The resultsshow that their h(2) in the generalized Hurst index curves are smaller than1,which indicates detrended energy time series are stationary. Then we use tradi-tional multifractal method to analyze them. The results show that their D(q)distribute nearby1and the multifractal spectra f (α) curve like a bar, whichimplies detrended energy time series are close to random series and no longerhave the multifractal feature.