A multi-relational learning framework to support biomedical applications

Abstract

The definition of tools able to extract knowledge from structured biological data in order to support scientists research is increasing as shown by the popularity reached in the field of bioinformatics. In particular we focus our attention on the domain of assisted reproduction techniques with particular interest on the field of intracytoplasmic sperm injection. In this paper we would provide a multi-relational learning framework able to discover hidden relationships between entities involved in this application domain. Our approach is based on a multi-relational partitional clustering algorithm followed by a multi-relational rule induction. Furthermore, the obtained rules can be represented in a easily comprehensible form and can be used as an advisor to the clinicians during their work in order to help them in determining what knowledge sources are relevant for a treatment plan. © 2011 Springer-Verlag Berlin Heidelberg.