mosaicMPI: a framework for modular data integration across cohorts and -omics modalities.

Area of Science:

• Computational biology
• Genomics
• Systems biology

Background:

• Multi-modal molecular profiling generates large datasets for complex diseases.
• Integrating diverse data types and cohorts is challenging for biological discovery.

Purpose of the Study:

• To present mosaicMPI, a computational framework for mosaic integration of multi-modal datasets.
• To discover and integrate molecular programs representing cell types and states into a biological theme network.

Main Methods:

• Developed mosaicMPI for low to high-resolution molecular program discovery.
• Integrated single-cell and bulk data across multiple platforms using glioblastoma datasets.
• Propagated clinical and molecular annotations onto the network of programs.

Main Results:

• Robust integration of single-cell and bulk molecular programs across platforms.
• Generated a richly characterized landscape of biological themes by propagating annotations.
• Demonstrated deep understanding of individual tumor samples and modality relationships.

Conclusions:

• mosaicMPI enables systematic exploration of biological themes within and across datasets.
• Provides a reference map for rapid integration of new multi-modal datasets.
• Facilitates a deeper understanding of complex diseases at a molecular level.