Why do our free energy numbers predict fairly well melting temperatures of oligos longer than 15 bp but do worse for shorter ones? Discrepancies between predicted and measured melting temperature for short oligos have been already reported in bulk measurements  and attributed to differences in analytical methods used to extract melting temperatures. For instance, there are a few definitions of the melting temperature that not always coincide: maximum heat capacity of the sample, maximum derivative of UV absorbance, etc. Another possible explanation is that short oligos ( bp) might not have the double helix perfectly formed and the formation energies involved in the duplex are slightly different from the energies for longer sequences (i.e., typical boundary effects of small systems). Although there is no conclusive answer to this question, it is worth underlining that UO free energy values are obtained in order to correctly predict the melting temperatures for all oligo lengths. This might lead to error compensation between the melting temperature data sets corresponding to short and long oligos. Let us stress that with increasing length, the prediction of melting temperatures is more tolerant of errors in the details of the NNBP energies, where sequence effects are averaged out. In addition, deviations from the bimolecular model (Eq. 3.14) arise for sequences with , as their melting process begins to shift toward pseudo-monomolecular behavior . Still, our predicted melting temperatures for oligos with agree well for the sequences reported in ref. .
JM Huguet 2014-02-12