Analysis of protein/DNA interactions by novel bioinformatic tools

Bohdan Schneider¹, Jiří Černý¹, Daniel Svozil², Jean-Christophe Gelly³,
Alexandre G. de Brevern³

^1Institute of Biotechnology AS CR, CZ-142 20 Prague, Czech Republic

²Prague Institute of Chemical Technology, Prague, Czech Republic

³ INSERM - University Paris Diderot, Paris 7, France

We analyzed interactions between proteins and DNA from over a thousand crystal structures of their complexes. To this end, we built a database of more than 50 thousand protein/DNA contacts that can sorted by a wide range of criteria such as the identity of the interacting atoms and residues, type of contacts including water-mediated ones, crystallographic resolution, and protein Pfam classification.

Novelty of the analysis lies in our fine-grained categorization of protein and DNA local conformations. Protein structures were classified into “peptide blocks” [de Brevern et al. Proteins 41, 271 (2000)] and DNA structures into dinucleotide conformers [Svozil et al. NAR 36, 3690 (2008)]. We determined how distributions of peptide blocks and dinucleotide conformers differ at and outside the protein/DNA interface and discussed variability of the distributions between various groups of structures (DNA complexes of enzymes, transcription factors, structural proteins) overall and also for contacts to the DNA minor and major grooves and phosphates. We examined how different are occurrences of peptide blocks and dinucleotide conformers at the interface and whether they are statistically over- or under-represented. We concentrated on analysis of direct polar contacts (mostly hydrogen bonds) and water-mediated contacts and observed e.g. a unique behavior of the water-mediated contacts in the DNA complexes of transcription complexes. Analysis of temperature displacement factors (“B-factors”) of the analyzed complexes showed the fundamental difference between behavior of proteins and DNA at the interface.

This work is supported by the Czech Science Foundation (P305/12/1801). BS and JC are supported by the institutional grant AV0Z50520701.