Exploring data visualization techniques in molecular dynamics: Analysis of dynamic cross-correlation maps as dynamic networks

Structural Bioinformatics is a subfield of Bioinformatics dedicated to studying biological mechanisms through the three-dimensional structures of biomolecules, particularly proteins. In this context, Molecular Dynamics is a technique that enables the computational simulation of molecular systems based on their chemical structures. However, this technique faces challenges related to high data dimensionality, high computational costs, and the significant storage requirements for the resulting data. Consequently, it is necessary to develop computational methods that facilitate the analysis, integration, and visualization of data derived from these simulations. Currently, a wide range of approaches exists to analyze such data, including metrics based on system energy or atomic displacement, such as Principal Component Analysis (PCA), Root Mean Square Deviation (RMSD), and Root Mean Square Fluctuation (RMSF). To identify informative patterns within these complex systems, a common analytical method is the Dynamic Cross Correlation Map (DCCM). DCCM analyzes the average correlation of Cartesian displacement vectors for all-versus-all amino acid residues, accounting for the various conformations adopted during the simulation. However, the average correlation done in DCCM can occlude specific transition correlations during the simulation. The DCCM can be interpreted as an adjacency matrix, as defined in Graph Theory. By incorporating the time dimension and simulation states, this matrix becomes dynamic, providing the data structure necessary to construct a dynamic network. Thus, the objective of this work is to explore the transformation, representation, and analysis of structural data from a computational perspective, aiming to bridge the gap between Data Visualization and Structural Bioinformatics. The results include the development of a tool that provides an alternative visualization of the DCCM method by incorporating time segmentation.