Large volumes of microarray data are registered daily in public repositories such as SMD (Belkin and Niyogi, 2003) and GEO (Ashburner et al., 2000). Such repositories have quickly become a community resource. However, due to the inherent heterogeneity of the microarray experiments, the data generated from different experiments could not be directly integrated and hence the resources have not been fully utilised. To address this problem, we propose a new microarray integration framework that achieves high-quality integration through exploiting invariant features such as relative information among genes. We also show how the proposed approach generalises the previous frameworks.