Breast cancer (BC) is a complex disorder characterized through unchecked cell growth, requiring precise classification for effective treatment planning. Traditional approaches have relied on single-omic data, including gene expression, for subtype analysis. However, recent advancements highlight the advantages of integrating multi-omic data employing machine learning (ML) and deep learning (DL) approach. In parallel, Graph Neural Networks (GNNs) have attracted significant attention for their ability to model complex interrelationships in multi-omic datasets. The paper, we propose a GNN-based multi-omic frame work for breast cancer subtype prediction. Our methodology comprises two phases: (1) Feature selection and graph generation using clinical, gene expression, and epigenetic data, followed by feature extraction through a Graph Convolutional Network (GCN) layer; and (2) Representation learning using a Graph Autoencoder with a fully connected classifier for subtype prediction. The potential of GNNs in precision oncology is illustrated by the experimental results, which indicate that our model outperforms existing predictive models.