Hi all,

Fminer algorithms for the calculation of frequent and correlated 
subgraphs (BBRC and LAST-PM) now support (a) automatic hit calculation 
per chemical (the number of times it occurs in it) and (b) the 
processing of not just binary- and numerically-labelled datasets, but 
also the processing of up to five different labels in the data 
(multinomial setting).

Property (a) may be of interest for instantiated models, where 
patterns can now have weights instead of just being perceived as 
present or not.
Property (b) allows to efficiently derive patterns correlated to any 
of the multiple target classes (implemented by a direct generalization 
of the binomial case).

Please read the following articles for some theory and usage 
information: 
http://www.maunz.de/wordpress/opentox/2011/support-calculation-fminer 
and http://www.maunz.de/wordpress/opentox/2011/multinomial-fminer.

Thank you
Andreas