Application Of Graph-based Data Mining To Biological Networks
Abstract
A huge amount of biological data has been generated by long-term research. It is time to start to focus on a
system-level understanding of bio-systems. Biological networks are networks of biochemical reactions,
containing various objects and their relationships. Understanding of biological networks is a starting point
of systems biology.
Multi-relational data mining finds the relational patterns in both the entity attributes and relations in the
data. A widely used representation for relational data is a graph consisting of vertices and edges between
these vertices. Graph-based data mining, as one approach of multi-relational data mining, finds relational
patterns in a graph representation of data.
This thesis will present a graph representation of biological networks including almost all features of
pathways, and apply the Subdue graph-based data mining system in both supervised and unsupervised settings.
This research will also show that the patterns found by Subdue have important biological meaning.