On the role of mismatches in DNA denaturation

In the framework of the Poland Scheraga model of DNA denaturation, we derive a recursion relation for the partition function of double stranded DNA, allowing for mismatches between the two strands. This relation is studied numerically using standard parameters for the stacking energies and loop entropies. For complementary strands (of length 1000), we find that mismatches are significant only when the cooperativity parameter $\sigma$ is of order one. Since $\sigma \sim O(10^{-5})$ in DNA, entropic gains from mismatches are overwhelmed by the energetic cost of opening a loop: mismatches are therefore irrelevant (and molecular recognition of the strands is perfect). Generating random mutations with probability $p$ on one strand, we find that large values of $\sigma$ are rather tolerant to mutations. For realistic (small) values of $\sigma$, the two strands do not recombine, even for small mutation fractions. Thus, molecular recognition is extremely selective.