MOFS: Multimodality Fusion Subtyping
Year: 2025
Type: article
Abstract: The MOFS (Multimodality Fusion Subtyping) framework is a comprehensive approach for integrating and analyzing multi-layer biological data to achieve clinically relevant disease subtype classification. The core principle of MOFS lies in combining diverse biological modalities—such as genomic, transcriptomic, proteomic, histopathological, and radiological data—into a unified analytical framework. By integrating these heterogeneous data sources, MOFS offers a more holistic and nuanced view of disease biology, enabling the identification of underlying patterns that might be overlooked in single-modality analyses. Integrating multimodal data reveals causal features that may be obscured in single-modality analyses, offering a more complete understanding of diseases. Multimodal data fusion can be categorized into early, intermediate, and late fusion based on the timing of integration. Intermediate fusion integrates data during clustering, allowing for the identification of multimodal joint clusters, capturing dependencies between different omics layers, and revealing underlying biological mechanisms. Intermediate fusion is generally considered more advanced than both early and late fusion, but it demands more sophisticated integration algorithms. In the MOFS framework, multiple intermediate fusion strategies are employed based on diverse principles to maximize the interpretative power and reliability of the derived subtypes. Following this, late fusion is applied to combine the results obtained from the different algorithms, culminating in a final, consensus-driven clustering outcome. It is important to note that the framework does not rely on any specific intermediate fusion algorithm. Rather, it evaluates the overall consensus from multiple clustering results, ensuring robustness and reliability. The MOFS framework represents a powerful tool for disease subtype classification by integrating multiple biological data modalities. Through the integration of intermediate fusion and late fusion strategies, MOFS uncovers novel disease subtypes and mechanisms that may be missed in single-modality approaches. This integrated approach not only enhances the accuracy and robustness of disease classification but also provides deeper insights into disease biology, potentially guiding more personalized diagnostic and therapeutic strategies.
Author: Xu, Hui
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W6930656827
FWCI:
Cites:
Cited by:
Related to: