Prediction of melting profiles and local helix stability for sequenced DNA

The utility and limits of the current theory of helix-coil transition of DNA were intensively studied. Comparison of calculated and observed melting profiles affords solid criteria of a theoretical model and parameter values. Our model differs from others in that heterogeneity in stacking interactions is taken into consideration (Section II). The parameter values were estimated by fitting calculated to observed profiles for several known-sequenced DNA fragments (Section III). Calculated profiles agree quite well with observed, not only for DNAs used for the fitting but also for those not used. Such good agreement was never obtained with traditional two-term (A X T and G X C) stability assignment. It was shown that the all-or-none approximation for local melting transition (Azbel's model) provides informative but somewhat erroneous estimates of local stability. The stabilities of 10 kinds of stacking doublets estimated by the fitting accord well with known properties of DNA double helix (Subsection III-4); three-dimensional structure, calculated stacking energies, and binding specificity of some intercalative drugs are all consistent with the estimated order of stacking stabilities. Relationship between triplet stability and genetic code was discussed. The mechanisms of ionic-strength dependence of melting profiles were discussed in Section IV. Four different mechanisms were proposed and their relative contributions in various salt ranges were considered. It was suggested that at low ionic strengths melting processes deviate significantly from equilibrium, which fixes a limit to the applicability of the present theoretical mode. Locally unstable regions and frequently opening regions were located on several natural DNAs by calculating probabilities of individual base-pair doublets being in an unstacked state. With few exceptions, origins of DNA replication on various DNAs fall on frequently opening regions, indicating their functional importance. Positive correlations were also obtained between frequently opening regions and transcriptional promoters, terminators and gene boundaries (Section V).