First choose the wt amino
residue (1 letter code), the
corresponding residue number (1-99), and any one of the 19 possible
substitutions of the wt residue (1 letter code). Next, choose the
number of activity classes (2 or 3). For the two-class case, the
activity of a mutant is considered to be either "active" or "inactive".
Using three classes,
the mutant activity labels are "positive", "intermediate", or
"negative". Repeat for up to five single-point mutants of HIV-1
protease. Finally, click on the "Check Activity Levels" button to
obtain the predicted activity levels of the requested protease mutants.
All predictions are based on decision tree models generated by using
vectors for a training set of 536 protease mutants
with experimentally measured activity levels. Decision tree models
based on the 1-against-all method are used for making the three-class
predictions. Among the training set data (those with "known" activity
when queried above), the activity
levels for 336 of these mutants can be located in Loeb et al.3 The activity of
200 mutants constitute a currently unpublished personal communication
courtesy of Ronald Swanstrom.