Predicting interaction networks of breast cancer risk genes using multiple microarray data

Introduction

Global expression profiling by microarray can provide invaluable information about biological properties of breast cancers. Here I report predicted gene regulatory networks of known breast cancer risk genes using multiple microarray data, in order to understand how the risk genes interact with each other and how the interaction may be related to the pathogenesis of breast cancer.

Methods

I used microarray data of breast cancer samples from four published studies. By combining the data from three smaller studies, I obtained two datasets with 548 and 684 samples, respectively. For each dataset, Pearson coefficients of expression levels between 74 known breast cancer risk genes were calculated first. The gene association network was also obtained by a new correlation metric called asymmetric correlation, which quantifies the nonlinearity of the correlations. Finally the results from two analyses were combined to obtain predicted gene regulatory networks.

Results

I found in both datasets that ESR1, GATA3 and FOXA1 formed a close cluster and each of them had interactions with a number of genes. In particular, FOXA1 showed positive interactions with ERBB2 and IGF1R while ESR1 and GATA3 were positively associated with NAT2 in both datasets. Positive associations were also found between AGTR1, FOXA1 and GATA3, and between CDH1, NAT2 and FGFR2. Moreover, FGFR2 and AGTR1 had negative associations with ERBB2, indicating that they have overwrapping but distinct gene network.

Conclusions

Transcription factors ESR1, GATA3 and FOXA1 were found to form a core network, which was connected by plasma membrane signal transducers ERBB2, IGF1R and AGTR1. FGFR2 and CDH1 are associated with this network, but they seem to play distinct roles in breast cancers.