Multivariate statistical analysis of mutational spectra of alkylating agents.

A series of multivariate statistical methods were used to explore the current knowledge on the mutational spectra of alkylating agents (AA) in bacterial and mammalian cells. The data relative to lac I and gpt genes of Escherichia coli were considered. The analysis focused on the distribution of GC to AT transitions, which account for the majority of AA-induced mutations. The statistical analysis of 15 different mutational spectra obtained by various laboratories pointed to a number of biological factors involved in the mutational process. First of all, factor and cluster analyses demonstrated that the mutational profiles obtained in mammalian cells form a homogeneous cluster different from the cluster formed by the bacterial cell mutational spectra. SN1-type AAs give rise to classes of mutational spectra statistically different from the spectra induced by the SN2-type AAs. The analysis of the mutated sequences of both genes pointed to a correlation between mutation induction by SN1 AAs, which react through a positively charged alkylating intermediate, and the occurrence of mutations at guanines preceded 5' by a purine. Moreover, our statistical analysis showed that the distribution of AA-induced mutations is not affected by the transcriptional activity of the target gene, but is strongly determined by the sequence specificity of AA-induced mutagenesis and by the structure of the target proteins. The agreement of our results with the findings of previous studies indicates that the multivariate data analysis methods are a sensitive and reliable tool for exploring the mechanisms underlying complex biological processes. The novelty of the present results lies in their quantitative character, and in the clarity of the graphical displays. We propose the use of this methodological approach to explore the large bulk of information available on mutational spectra.