missRows

missRows implements Multiple Imputation–Multiple Factor Analysis (MI-MFA) to estimate positions of missing individuals across multi-omics datasets and quantify the uncertainty of those estimates.

Key Features:

• MI-MFA: Implements Multiple Imputation–Multiple Factor Analysis combining multiple imputations with Multiple Factor Analysis.
• Multiple imputations: Performs multiple imputations to fill missing rows, producing M completed datasets.
• Multiple Factor Analysis (MFA): Applies MFA to each completed dataset to derive individual coordinates across data tables.
• Consensus solution: Combines the M MFA configurations into a single consensus solution estimating positions of missing individuals on the first MFA components.
• Diagnostic visualizations: Provides confidence ellipses and convex hulls to visualize and assess uncertainty introduced by missing values.
• Comparative performance: Shows superior solution accuracy relative to regularized iterative MFA (RI-MFA) and mean variable imputation (MVI-MFA).
• Empirical validation: Evaluated on two real omics datasets and corresponding incomplete artificial datasets with varying patterns of missingness.
• Uncertainty quantification: Quantifies reliability of estimated coordinates by considering variability across imputed configurations.

Scientific Applications:

• Multi-omics integration: Integration of multiple omics layers in studies where certain individuals are absent from some datasets.
• Coordinate estimation: Estimation of coordinates for absent biological units on MFA components for downstream multivariate analyses.
• Uncertainty assessment: Assessment of uncertainty and robustness of integrative analyses under varying patterns of missingness using diagnostic visualizations.

Methodology:

Performs multiple imputations to create M completed datasets, applies Multiple Factor Analysis to each completed dataset, and combines the M resulting configurations into a single consensus solution estimating positions on the first MFA components, with diagnostics provided as confidence ellipses and convex hulls.