The analysis of multilinear regression has been used to search a mathematic
model, able to forecast water solubility of an heterogeneous group formed
by 97 organic compounds. The generation of the descriptors which characteri
ze different structural aspects of the group has been realised using Molecu
lar Connectivity. A linear model formed by four descriptors foresees satisf
actorily the solubility of all the compounds except the raffinose.