Similarity search in datasets of short (5-15 amino acid) protein fragments has importance in both providing a building block to more complex (full protein search) algorithms and for possible use in direct biological investigations.

In this talk I will present PFMFind, a system that helps indicate possible biological functions of individual protein fragments by using similarity-based search to find close neighbors of the fragments of a given query sequence and then retrieving the descriptions of the search hits from the publicly available databases. The results are displayed in a way that allows a user to quickly focus on their most significant features.

PFMFind consists of three main components: a user interface client, an efficient indexing scheme called FSindex and a relational database. About half of the talk will be about FSindex; the theory behind it, implementation issues and its performance. I will also present a few typical exploration experiments using PFMFind.