A latent unknown clustering integrating multi-omics data (LUCID) with phenotypic traits.

Area of Science:

• Genomics
• Biostatistics
• Translational Research

Background:

• Multi-omics data generation is increasing in epidemiologic, clinical, and translational studies.
• Integrating high-dimensional data types is crucial for discovering novel associations and patient subgroups.

Purpose of the Study:

• To propose an integrative model, Latent Unknown Cluster Identification (LUCID), for multi-omics data integration.
• To distinguish unique omic effects and jointly estimate outcome-relevant subgroups.
• To enable future prediction of risk subgroups and phenotypes.

Main Methods:

• Developed an integrative statistical model (LUCID) for multi-omics data.
• Utilized simulation studies to validate estimation consistency and subgroup accuracy.
• Applied the model to real-world genomic, exposure, and metabolomic datasets.

Main Results:

• Simulation studies demonstrated accurate estimation of parameters and subgroup identification.
• The model successfully recapitulated subgroup-specific effects.
• The approach was applied to real data, integrating genomic, exposure, and metabolomic information.

Conclusions:

• LUCID provides a robust framework for integrating diverse high-dimensional omics data.
• The method facilitates the identification of novel biomarkers and patient subgroups.
• LUCID enables accurate prediction of disease risk and phenotypes.