A general framework for integrative analysis of incomplete multiomics data.

Area of Science:

• Genomics
• Proteomics
• Metabolomics
• Systems Biology
• Biostatistics

Background:

• Multi-omics studies are crucial for understanding complex diseases but face challenges with missing data and detection limits.
• Genomic data is often complete, while other omics data (RNA, protein, metabolite) may be measured on subsets.
• Quantitative omics data can have values below or above detectable thresholds, leading to censored measurements.

Purpose of the Study:

• To develop a rigorous and powerful statistical approach for integrative analysis of multi-omics data.
• To effectively handle arbitrary patterns of missing values and detection limits in quantitative omics measurements.
• To enable robust statistical inference in multi-omics studies with incomplete data.

Main Methods:

• Linear regression models relate omics variables to genetic variants and other covariates.
• Generalized linear models link phenotypes to omics variables and covariates.
• A joint-likelihood function is derived to accommodate missingness and detection limits.
• Maximum-likelihood estimation is performed using computationally efficient and stable algorithms.

Main Results:

• The proposed method provides statistically efficient and approximately unbiased estimators.
• The approach successfully integrates diverse omics data types, even with missingness and detection limits.
• Application to a chronic obstructive lung disease study revealed novel biological insights.

Conclusions:

• The developed method offers a powerful tool for integrative multi-omics data analysis.
• It addresses critical data challenges, improving the reliability of biological discoveries.
• This approach facilitates deeper understanding of complex diseases through comprehensive data integration.