Comments on: 'Classification of images of biomolecular assemblies: a study of ribosomes and ribosomal subunits of Escherichia coli'.

In a recent paper Frank et al. (1988; J. Microsc. 150, 99-115) proposed the use of a hybrid clustering technique to sort macromolecular images. In contrast to the claims made in that paper, however, the proposed technique is less sophisticated than the enhanced hierarchical ascendant classification technique proposed by us a few years earlier. The hybrid approach will lead to partitions which are less optimal in terms of minimizing the total intra-class variance for a predetermined number of classes, irrespective of whether the data show distinct clusters or, rather, possess a quasi-continuous intrinsic structure.