In this talk I will first describe how stresses on DNA are regulated in living organisms. Then I will sketch the computational methods used to predict the duplex destabilization properties of stressed DNA sequences. These properties do not depend only on local sequence attributes. Rather, the imposed stress couples together all sites within the molecule, making the destabilization behavior of a sequence competitive and interactive in character.
When one applies these methods to the analysis of DNA sequences on which experiments have been performed, the sites and extents of predicted stress-induced strand separation agree precisely with experimental measurements. This allows these computational methods to be used to predict the destabilization properties of molecules for which experimental data is not available. One finds that sites of predicted destabilization do not occur at random, but instead are closely associated with specific types of DNA regulatory regions. Implications of these results regarding possible mechanisms of activity of these regions will be discussed. The strengths of these associations suggest a new strategy to search DNA sequences for specific types of regulatory regions. This approach augments string-based searches for consensus sequences with evaluations of this essential physical chemical correlate of activity - susceptibility to stress-induced DNA duplex destabilization.