DERIVATION OF A SCALE-INDEPENDENT PARAMETER WHICH CHARACTERIZES GENETIC SEQUENCE COMPARISONS

A new method of quantifying the similarity between genetic sequences is presented. The method makes use of the finding that sequence comparisons expressed in binary vector form have an associated scale-independent parameter, D. This parameter is represented in the function M(S, n) = (N) (en/enD), where S is the vector, n represents the window size which is allowed to vary, N is a constant, and D is the scale-independent measure of homology. By comparing two sequences using this method, a unimodal, symmetric distribution of D values associated with the frameshifted vectors is obtained. The degree of sequence similarity is determined by the distribution of these parameters. A set of sequences of evolutionary interest coding for glyceraldehyde-3-phosphate dehydrogenases and mammalian insulins iscompared using this methodology. The results confirm evolutionary tree distances calculated using different procedures. Since a z score can be calculated for each comparison, the method allows forthe rapid identification of sequence homologies ranked according to the probability of occurrence. This unique scale-independent measure of similarity allows contrasts and comparisons between any two sequence fragments using all available order information. © 1993 Academic Press, Inc.