AT domain classification on basis of HMMs.

The first column lists the different substrates and between brackets the number sequences that were analyzed. The Table lists the number of correctly (c, bold) and falsely (f) classified sequences and the number of sequences that scored above threshold (at, grey and in italics). The values in columns 2, 3 and 4 were derived from the use of a single HMM per substrate, and the columns 5, 6 and 7 relate to the prediction made using an ensemble of multiple HMMs per substrate. The values in columns 8, 9 and 10 relate to the Leave One Out cross validation.
$ The set contained 167 non redundant sequences. See the legend of Figure 2 for the systematic name of the various substrates.
# The category ‘other’ sequences includes those specific for 3MbuC, AC, CH and CP as only one sequence has been experimentally identified and thus no reliable model could be made.