this clustering approach,
although not used to predict class, was capable of grouping samples
according to their known identification reasonably well, with fairly
reproducible results as the dimensionality of the data is reduced,
although notably with some overlap between the populations. It is
interesting to recognize however that of the three samples which the
ANN incorrectly identified using the top 1000 ions (highlighted in
Figure 4a by arrows), two are in clear clusters of the opposite group,
i.e. the sample on the extreme right of the incorrect classifications
was identified using 16SrDNAmethods as Kingella denitrificans, but
clearly showed more similarities to the N.meningitidis samples than
other species