Structural bioinformatics of cyclic AMP-DNA binding sites and molecular dynamics simulations of DNA oligonucleotides

Regulatory binding proteins often bind to multiple target DNA sequences in a genome, affecting the transcription of downstream genes. Predicting where the proteins bind in an effort to understand gene regulation has proven a difficult task. Consensus sequence is insufficient because biological sites span a large sequence space. Using the CAP (cyclic AMP receptor protein) system in E. coli as a well studied simple prototype, identifying protein binding sites has been addressed with sequence, DNA structure, and free energy.; HMMs (Hidden Markov Models), tools specialized in probabilistic pattern recognition, are used to generate probabilistic sequence by position along the DNA. The sequence specific dynamic structure of DNA is taken from a library of state of the art MD (Molecular Dynamics) simulations, which give time resolved, atomistic level dynamics of molecules consistent with Newton's laws of motion. MD structure is incorporated into a sequence model by way of a transformation procedure, yielding hybrid models including both sequence and MD derived dynamic structure that have important advantages over sequence based models, indicating the dynamic structure of DNA is important to the binding event.; The additive free energy model is parameterized from the best experimental binding information. These models provide a means of estimating the free energy for the contact regions of the binding interface and indicate substantial contributions to binding come from the flanking and interhelix regions.; The success of this project hinges on the existence of state of the art MD simulations and the existence of a database of known binding sites. Two issues affecting DNA structure investigated by MD simulations in this study are the effects of flanking DNA bases and ions on DNA local structure, which have been addressed in the context of advancing the state of the art of MD. A literature search of CAP binding sites yielded a database of binding sites guaranteed to be functional in their promoters, a set suitable for promoter modeling.; This research provides an alternate means of determining binding sites avoiding reliance on sequence alone. The ideas uncovered in this research may be generally applicable for identifying binding sites for other systems.