Abstract: A molecular retrieval system and method answering to similarity queries for retrieving molecular structures, stored into a source database (22) and having a required similarity with an input structure, which can contain a set of property regions. A target database (23) stores the molecular structures described in hierarchical way and a knowledge base (24) stores well-known molecular fragments at different levels of description together with a set of physical and chemical properties associated to each fragment. A fragment recognizer (21) analyzes the fragments of the input structure to represent them in a hierarchical way. A query analyzer (25) analyzes the similarity queries and selects the appropriate level of molecular representation on the basis of the required similarity.