AEMVC: anchor enhanced multi-omics cancer subtype identification

Abstract

The discovery of cancer subtypes has helped researchers gain deeper insights into the study of oncology heterogeneity. However, since cancer complexity exists in various omics levels, extracting and adaptive combining complementary information across multi-omics are still challenges in cancer subtype prediction approaches. Based on the subspace learning of multi view clustering, we propose a new multi group cancer subtype recognition model based on anchor enhancement. Firstly, we generate anchors for each view's local similarity graph structure to enhance the connectivity between samples. Secondly, the graph convolution module is used to learn the consistency similarity features and specific features of patient samples in each view. Finally, the corresponding cancer subtype clustering results can be calculated according to the self-expressive coefficient matrix of the consistency similarity features obtained in the previous step.