资 源 简 介
To better exploit the information Gene Ontology has to offer, we design a
clustering validation index that integrates its graph structure and the intrinsic features of GO terms. The proposed index is made up of
two metrics for measuring intra-cluster functional compactness and inter-cluster functional similarity, respectively. Instead of using GO
as functional categories, the proposed index seeks to compare clusters on the same level of specificity. A major difficulty in assessing
GO-driven validity indices lies in the fact that the ground truth of the biological structure is unknown. Therefore, the evaluation involves
six clustering algorithms, three distinct biological datasets, as well as existing GO-driven and data-driven validity indices. Results from
three experiments, each assessing the validity indices from a different perspective, have confirmed that the proposed index has the
best overall performance. In summary, both theoretical and experimental analyses show that the pro
