一种基于广义相似性的共调控基因聚类算法

Recently, more and more reported studies focus on pattern/tendency-based co-regulated genes clustering, however, most of which cannot be directly applied to find the negative regulated gene clusters. Even for the positive regulated gene clusters, the pattern/tendency -based approaches may also risk missing some results which are potentially of high biological significance. In this paper, in order to group the positive and negative co-regulated genes together, we propose a novel model, namely g-Cluster, based on the general similarity, where two genes are clustered into the same co-regulated gene cluster if and only if they have the same code. Further, a tree-based clustering algorithm with several efficient pruning and optimization strategies, called FBTD, is designed to identify all maximal g-Clusters in a "First Breadth-first and Then Depth-first" manner. We conduct extensive experiments on real (e.g. Yeast and AML-ALL) and synthetic datasets and commit the results to Gene Ontology. Both theoretic analysis and experimental results show that our algorithm is effective and efficient, and outperforms the existing approaches.