PROJECT TITLE :

RGFinder: A System for Determining Semantically Related Genes Using GO Graph Minimum Spanning Tree

ABSTRACT:

Biologists often need to know the set $S^{prime}$ of genes that are the most functionally and semantically related to a given set $S$ of genes. For determining the set $S^{prime}$, most current gene similarity measures overlook the structural dependencies among the Gene Ontology (GO) terms annotating the set $S$, which may lead to erroneous results. We introduce in this paper a biological search engine called RGFinder that considers the structural dependencies among GO terms by employing the concept of existence dependency. RGFinder assigns a weight to each edge in GO graph to represent the degree of relatedness between the two GO terms connected by the edge. The value of the weight is determined based on the following factors: 1) type of the relation represented by the edge (e.g., an “is-a” relation is assigned a different weight than a “part-of” relation), 2) the functional relationship between the two GO terms connected by the edge, and 3) the string-substring relationship between the names of the two GO terms connected by the edge. RGFinder then constructs a minimum spanning tree of GO graph based on these weights. In the framework of RGFinder, the set $S^{prime}$ is annotated to the GO terms located at the lowest convergences of the subtree of the minimum spanning tree that passes through the GO terms annotating set $S$. We evaluated RGFinder experimentally and compared it with four gene set enrichment systems. Results showed marked improvement.