2000 年 15 巻 1 号 p. 169-176
The paper describes machine discovery of protein functional models from protein databases with an Inductive Logic Programming (ILP) method using conceptual hierarchy and search biases. The proposed ILP method discovers effectively protein functional models that explain the relationship between protein functions and structures from protein databases describing amino acid sequences and properties of proteins. The method is based on top-down search for relative least general generalization and uses domain knowledge defining the conceptual hierarchy of protein functions and search biases. The method succeeds in discovering protein functional models for forty membrane proteins, which coincide with conjectured models literature of molecular biology.