Staged Multi-Omics Data Classification for Breast Cancer Diagnosis

Breast cancer is a complex and heterogeneous disease with a significant genetic component, characterized by diverse molecular subtypes that influence its progression, treatment response, and prognosis. Advances in high-throughput omics technologies, such as genomics, transcriptomics, proteomics, and epigenomics, have greatly improved our understanding of the genetic and molecular landscape of breast cancer. Conventional AI approaches typically require integrating all omics data at the outset to achieve an optimal breast cancer diagnosis, which can be inefficient and often unnecessary. To reduce clinical costs and improve diagnostic accuracy using multi-omics data, we propose a novel Staged Graph convolutional network with Uncertainty Quantification (SGUQ). SGUQ begins with mRNA data and progressively incorporates DNA methylation and miRNA data only when necessary, reducing overall costs and minimizing exposure to potentially harmful tests. Experimental results demonstrate that SGUQ achieved an accuracy of 0.814, a weighted F1 score of 0.795, and a macro F1 score of 0.755 for breast cancer diagnosis using the TCIA dataset, significantly outperforming peer methods. Furthermore, within the test cases, 14.38% of patients require only mRNA, and 7.19% of patients require both mRNA and DNA methylation data. The proposed SGUQ can be applied not only to breast cancer diagnosis using multi-omics data but also to clinical decision-making involving multi-view data.