Multicenter comparison of three different analytical systems for evaluation of DNA banding patterns from Cryptococcus...

The enormous improvement of molecular typing techniques for epidemiological and clinical studies has not always been matched by an equivalent effort in applying optimal criteria for the analysis of both phenotypic and molecular data. In spite of the availability of a large collection of statistical and phylogenetic methods, the vast majority of commercial packages are limited by using only the unweighted pair group method with arithmetic mean algorithm to construct trees and by considering electrophoretic pattern only as migration distances. The latter method has serious drawbacks when different runs (separate gels) of the same molecular analysis are to be compared. This work presents a multicenter comparison of three different systems of banding pattern analysis on random amplified polymorphic DNA, (GACA)(4), and contour-clamped homogeneous electric field patterns from strains of Cryptococcus neoformans var. neoformans isolated in different clinical and geographical situations and a standard Saccharomyces cerevisiae strain employed as an outgroup. The systems considered were evaluated for their actual ability to(i) recognize identities, (ii) define complete differences (i.e., the ability to place S. cerevisiae out of the C. neoformans cluster), and (iii) estimate the extent of similarity among different strains. The ability to cluster strains according to the patient from which they were isolated was also evaluated. The results indicate that different algorithms do indeed produce divergent trees, both in overall topology and in clustering of individual strains, thus suggesting that care must be taken by individual investigators to use the most appropriate procedure and by the scientific community in defining a consensus system.