String Clustering and Statistical Validation of Clusters

摘要

In this article we present a new string clustering algorithm and a statistical validation of discovered clusters. They are obtaiend by searching for common structures of strings and by grouping those sharing wide words. The application of a sttistical test, estiamting cluster homogeneity, allows to find automatically the number of classes. We apply our method to extract key-structures in 40 biological sequences of about 100 characters each in order to build clusters, and to find three original families again.