Clustering protein conformations using a dynamic programming based similarity measurement

Understanding the structure and dynamics of proteins is essential in order to understand their function. In particular, it is important to detect clusters of highly populated regions which could correspond to intermediate structures or local minima. The conformational space of proteins is complex and high dimensional, which makes its analysis a highly challenging task. We present a Dynamic Programming (DP) method for clustering and classification of protein conformations, based on their lower-dimensional representation. Previously, we used the similarity method to identify pairs of co-regulated genes based on their microarray expression data. In this paper we demonstrate our method on trajectories obtained by a coarse grained protein conformational search of three different proteins. Our clustering method was extremely fast, and was able to produce compact, well separated clusters for all the tested examples, showing that both the DP-based method and the dimensionality reduction technique were able to preserve the inter-molecular distances and provide clusters that correspond to experimentally determined intermediates when such are available.