Keswani, J. and W. B. Whitman. 2001. Relationship of 16S rRNA sequence similarity to DNA hybridization in prokaryotes. Int. J. System. Evol. Microbiol. 51:667-678.

The relationship between 16S rRNA sequence similarity (S) and the extent of DNA hybridization (D) was well described by the equation ln(-lnD) = 0.53[ln(lnS)] + 2.201 when D was determined by either the S1 nuclease or membrane filter methods.  When the presence of nonultrametric rRNA sequences and differences between genera or families were controlled, this relationship accounted for 78% of the variability of D given S, and it was possible to estimate the distribution of D form S with a known precision.  Thus, D < 0.70 was expected to occur 50, 95, and 99% of the time when S was 0.998, 0.992, and 0.986, respectively.  The relationship between D and S varied between prokaryotic taxa even within the same subphylum, and more precise estimates of D could be made when the relationship for a particular taxon was known.  The relationship between D and S was not significantly different between the prokaryotic domains, and S appeared to be a quasi-molecular clock of approximately constant rate when averaging effects and stochastic factors were taken into account.  The relationship between logD and logS was nonlinear, and D provided a very poor measure of relatedness for distantly related organisms.  For instance, within the range 1.0 > S > 0.95, D decreased from 1.0 to 0.15; and within the range 0.95 > S > 0.90, D decreased from 0.15 to 0.06.  Lastly, at least some of the rRNA sequence from about one-third of the taxa examined had nonultrametric properties where S was much lower than expected from the value of D.  For these taxa, S was a poor indicator of relatedness for closely related strains.  Thus the ultrametric properties of rRNA sequences should be tested before making taxonomic or phylogenetic conclusions based upon S.