MONTI: A Multi-Omics Non-negative Tensor Decomposition Framework for Gene-Level Integrative Analysis

Multi-omics data is frequently measured to enrich the comprehension of biological mechanisms underlying certain phenotypes. However, due complex relations and high dimension multi-omics data, it difficult associate omics features traits interest. For example, clinically valuable breast cancer subtypes are well-defined at molecular level, but poorly classified using gene expression data. Here, we propose a analysis method called MONTI (Multi-Omics Non-negative Tensor decomposition for Integrative analysis), which goal select that able represent trait specific characteristics. demonstrate strength integrated in terms subtyping. The first biologically meaningful manner form three dimensional tensor, then decomposed non-negative tensor method. From result, selects highly informative subtype features. was applied case studies 597 cancer, 314 colon 305 stomach cohorts. all studies, found classification accuracy significantly improved when utilizing available detect sets showed be strongly regulated by omics, from correlation between types could inferred. Furthermore, various clinical attributes nine were analyzed MONTI, some well explained We demonstrated integrating centric improves detecting other features, may used further understand characteristics software this study at: https://github.com/inukj/MONTI .