A regularized Bayesian Dirichlet-multinomial regression model for integrating single-cell-level omics and...

Area of Science:

• Computational biology
• Genomics
• Translational medicine

Background:

• Cell type abundance varies significantly between patients, influenced by phenotype, age, gender, and lifestyle.
• Existing methods struggle to effectively integrate single-cell omics data with clinical variables.
• Understanding these relationships is crucial for disease research and personalized medicine.

Purpose of the Study:

• To develop a novel computational framework for integrating single-cell RNA sequencing (scRNA-seq) data with patient-level clinical data.
• To investigate the associations between cell type abundance and clinical variables across different diseases.
• To identify cell-type-specific relationships at various hierarchical levels.

Main Methods:

• A regularized Bayesian Dirichlet-multinomial regression framework was developed.
• A novel hierarchical tree structure was incorporated to analyze cell types at different granularities.
• The model was applied to scRNA-seq data from patients with pulmonary fibrosis, COVID-19, and non-small cell lung cancer.

Main Results:

• The proposed framework successfully identified significant associations between specific cell types and clinical variables.
• These associations were observed across three distinct disease cohorts.
• The hierarchical structure enabled the discovery of relationships at multiple cell-type resolution levels.

Conclusions:

• The developed Bayesian regression model provides a robust method for integrating scRNA-seq and clinical data.
• This integrative analysis yields novel biological insights into disease mechanisms.
• The findings have the potential to inform future clinical interventions and personalized treatment strategies.