Proteinlerde bilgi transferinin çok yönlü analizi

"Sequence-structure-dynamic-function" is a series of interconnected steps that explains working mechanism of proteins. This thesis aims to reveal dynamics of a protein from sequential and structural data, and use obtained dynamics to detect functional significant sites of the protein. Various formulations of mutual information —including exact anisotropic, multivariate Gaussian, isotropic, isotropic Gaussian, and Gaussian Network Model mutual information (GNM-MI) methods—are applied to structural data to reveal information exchange along a protein. Structural data contains molecular dynamics (MD) simulations of Ubiquitin for 1ms and Papain-like protease for 100μs, and their crystal structures. Multivariate Gaussian method achieves to capture long range dynamics of proteins using very short MD simulations. Gaussian distribution of fluctuations assumption well represents residue-level dynamics. GNM-MI method successfully reveals long-term dynamics of both cases, highlighting longer MD simulation are needed to capture the slowest movements. As an extention to GNM-MI, integration of transfer entropy (TE) formulation, sequence based Gaussian Network Model Transfer Entropy (SeqGNM-TE) method well replicates the results of structure based method. Collective behavior of functional residues and information transfer between distant functional sites are observed. Residues that are collective but don't have known function send information to functional residues, making them potential functional sites. These results make SeqGNM-TE a promising method for detection of functional sites of proteins. Results of SeqGNM-TE strongly support that certain dynamic features of protein dynamics and allosteric mechanism are evolutionarily conserved. SeqGNM-TE make a significant contribution to ``sequence-structure-dynamic-function" paradigm from perspective of ``evolution-allostery-function".