PERFORMANCE ANALYSIS OF RESTRICTION ENZYMES BASED IN RESULTS OF CLUSTER STABILITY

Abstract

This paper presents an application of a method aiming to compare a performance of the restriction enzymes with the results obtained by the analysis from cluster stability within a Brazilian collection of 119 Bradyrhizobium strains. The stability has been studied as a combination of six restriction enzymes used in the RFLP-PCR analysis and three ribosomal regions using three restriction enzymes per region, each combination forms a pair, thus there are nine pairs: pair 1 (Cfo I 16S), pair 2 (Dde I 16S), pair 3 (Dde I IGS), pair 4 (Hae III IGS), pair 5 (Hae III 23S), pair 6 (Hha I 23S), pair 7 (Hinf I 23S), pair 8 (Msp I 16S), pair 9 (Msp I IGS). The analysis of cluster stability is a way to validate the partitioning of data encountered through any conventional clustering algorithms. The aim is to compare a reference cluster obtained from all of samples with several clusters from subsamples of original dataset. For this study, the sampling ratio was 0.8 and 25 datasets were made from subsamples. The similarity was calculated between pairs of samples of the data and the stability was computed using the whole collection of similarities. For the system analyzed, were generated 511 experiments (all combinations from 1 up to 9 pairs = 9!-1 pairs) and the number of possible clusters varied from 2 to 10. The results indicated that groupings up to 7 clusters were sufficient to achieve the stable result with a reduction of time expense in simulations. The pairs 1 and 2 increased the similarities of the clustering process. The pair 3 increased the similarities of experiments when associated with pairs 7 and 8, and decreased the similarities of the experiments with pair 4. Pair 4 decreased the similarities of the experiments. Keywords: restriction enzymes, cluster stability, Bradyrhizobium.