Neural computing in pharmaceutical products and process development - Computer-Aided Applications in Pharmaceutical Technology - page 129

Application of tree induction to the data base yielded a set of
hierarchical rules, which were represented as decision trees
(Figure 5.6). Developed methodology was proven to be successful
in classifi cation of test compounds in appropriate formulation
classes.
Results obtained suggest that it is possible to develop tools
for prediction of the optimal formulation approach for novel
compounds on the basis of knowledge of their physicochemical
properties. A relatively small set of information, most of which
is usually available at candidate drug pre-nomination stage,
was suffi cient to establish a useful map of the relationships
between compound features and formulation technology
categories. This kind of approach would signifi cantly reduce
resources needed for formulation development and increase
knowledge of the formulations being developed in the
pharmaceutical industry.

￿

￿ ￿

Decision trees generated using tree induction
methodology for selection among (a) lipidic/
surfactant and solid dispersion formulation
classes

Figure 5.6

( Continued )