Similarity pattern analysis in mutational distributions

The validity and applicability of the statistical procedure - similarity pa ttern analysis (SPAN) - to the study of mutational distributions (MDs) was demonstrated with two sets of data. The first was mutational spectra (MS) f or 697 GC to AT transitions produced with eight alkylating agents (AAs) in the loci gene of Escherichia coli. The second was a recently summarized dat a on the distributions of 11562 spontaneous, radiation- and chemical-induce d forward mutations in the ad-3 region of heterokaryon 12 of Neurospora cra ssa. They were analyzed as large two-way contingency tables (CTs) where two kinds of profiles were compared: site (or genotypic class) profiles and or igin (or mutagen) profiles. To measure similarity (homogeneity) between any pair of profiles, the relevant sufficient statistics, Kastenbaum-Hirotsu s quared distance (KHi(2)), was used. Collapsing the similar profiles into di stinct internally homogeneous clusters named 'collapsets' revealed their si milarity pattern. To facilitate the procedure, the computer program, COLLAP SE, was elaborated. The results of SPAN for the lacI spectra were found com parable with the results of their previous analysis with two multivariate s tatistical methods, the factor and cluster analyses. In the ad-3 data set, five collapsets were revealed among origin profiles (OPs): (I) ENU = 4NQO = 4HAQO = FANFT = SQ18506; (II) AF-2 = EI = MMS = DEP; (III) ETO = UV; (IV) AHA = PROCARB; and (V) He ions = protons. Moreover, the previous observatio n that MDs are dose-dependent was confirmed for X-ray-induced MDs. Profiles induced with the low doses of X-rays are similar to that induced with Sr-8 5, and profiles induced with the medium X-ray doses to those induced with p rotons and He ions. Evaluated similarities appear to be rather reasonable: mutagens with similar mode of action induce similar MDs. Similarity pattern revealed among genotypic class profiles (GCPs) seems to be also interpreta ble. When supplemented with descriptive cluster analysis, SPAN appears to b e a fruitful methodology in MS analysis. (C) 1999 Elsevier Science B.V. All rights reserved.