Metropolis-Monte Carlo algorithms are developed to analyze the strand
separation transition in circular superhelical DNA molecules. Moves th
at randomize the locations of unpaired regions are required in order t
o diminish correlations among the sampled states. This approach enable
s accurate simulations to be performed in reasonable computational tim
es. Sufficient conditions to guarantee the formal correctness of the c
omplete algorithm are proven to hold. The computation time required sc
ales at most quadratically with molecular length, and is approximately
independent of linking difference. Techniques are developed to estima
te the sample size and other calculation parameters needed to achieve
a specified accuracy. When the results of Monte Carlo calculations tha
t use shuffling operations are compared with those from statistical me
chanical calculations, excellent agreement is found. The Monte Carlo m
ethodology makes possible calculations of transition behavior in cases
where alternative approaches are intractable, such as in long molecul
es under circumstances where several runs of open base pairs occur sim
ultaneously. It also allows the analysis of transitions in cases where
the base pair separation energies vary in complex manners, such as th
rough near-neighbor interactions, or the presence of modified bases, a
basic sites, or bound molecules. (C) 1995 American Institute of Physic
s.