Bioinformatic analysis of protein/DNA interactions

In this thesis, we focused on local structural features of the DNA backbone in protein-complexed DNA and non-complexed (naked) DNA, and its dependence on types of a base pairing in DNA, and on the base sequence. To reach this goal we analyzed about 1,400 crystal structures of DNA in complexes with proteins and more than 400 crystal structures of naked DNA. DNA local conformations were structurally classified into 38 dinucleotide conformers ntCs, which were described previously (Svozil et al. Nucleic Acids Res. 2008). The ntC were further clustered into 16 structural alphabet classes ntA to reduce the number of analyzed variables. We assembled base-paired dinucleotides from double helical DNA structures accord- ing to their assigned structural alphabet classes into so called Association matrices. Three basic Association matrices were analyzed; two compare ntA/ntA associations between dinucleotides forming only Watson-Crick base pairs in protein/DNA com- plexes and in naked DNA, respectively; the third one ntA/ntA associations between dinucleotides base-paired also by non-Watson-Crick pairs. We also analyzed As- sociation matrices of dinucleotides as a function of their sequences. The analyzes revealed differences in structural behavior of various ntA and their dependence on dinucleotide sequences