How does a protein fold? A time series segmentation study

Abstract

In this work, we perform time series segmentation on average velocity time series of penta-alanine protein simulated in water. We then cluster the segments according to their fluctuation characteristics and discover periods of weak overall fluctuations surrounded by well-aligned segment boundaries. We speculate that these quiet periods may correspond to the folded state of the protein and that the well-aligned segment boundaries surrounding them may indicate the folding and unfolding processes. We then proceed to systematically identify such global transitions, measure the lifetimes of the quiet periods and the folding/unfolding events, and identify plausible precursor segments leading the transitions. We then check our findings of possible folding events against traditional methods of comparing molecular structures and measuring the radius of gyration. Finally, we extended the work by including atomic cross correlation analysis on various quiet periods and transitions to identify possible hydrogen bond formation and to better understand the folding dynamics of the protein.