Evidence from nuclear sequences that invariable sites should be considered when sequence divergence is calculated

It has long been known, from the distribution of multiple amino acidreplacements, that not all amino acids of a sequence are replaceable. Morerecently, the phenomenon was observed at the nucleotide level inmitochondrial DNA even after allowing for different rates of transition andtransversion substitutions. We have extended the search to globin genesequences from various organisms, with the following results: (1) Nearlyevery data set showed evidence of invariable nucleotide positions. (2) Inall data sets, substitution rates of transversions and transitions werenever in the ratio of 2/1, and rarely was the ratio even constant. (3) Onlyrarely (e.g., the third codon position of beta hemoglobins) was it possibleto fit the data set solely by making allowance for the number of invariablepositions and for the relative rates of transversion and transitionsubstitutions. (4) For one data set (the second codon position of betahemoglobins) we were able to simulate the observed data by making theallowance in (3) and having the set of covariotides (concomitantly variablenucleotides) be small in number and be turned over in a stochastic mannerwith a probability that was appreciable. (5) The fit in the latter casesuggests, if the assumptions are correct and at all common, that currentprocedures for estimating the total number of nucleotide substitutions intwo genes since their divergence from their common ancestor could be low byas much as an order of magnitude. (6) The fact that only a small fractionof the nucleotide positions differ is no guarantee that one is notseriously underestimating the total amount of divergence (substitutions).(7) Most data sets are so heterogeneous in their number of transition andtransversion differences that none of the current models of nucleotidesubstitution seem to fit them even after (a) segregation of coding fromnoncoding sequences and (b) splitting of the codon into three subsets bycodon position. (8) These frequently occurring problems cannot be seenunless several reasonably divergent orthologous genes are examinedtogether.