MOSCATO

MOSCATO performs regularized tensor regression to integrate multi-omic single-cell data and identify features associated with clinical outcomes.

Key Features:

• Multi-Omic Data Integration: Organizes genomics and transcriptomics single-cell data into multi-dimensional tensors with each dimension corresponding to an omic type.
• Regularized Tensor Regression: Performs regularized tensor regression using the clinical outcome as a predictor to identify clinically associated networks across omic types.
• Penalization Techniques: Applies penalization methods, including an elastic net constraint on coefficient vectors, to select non-zero coefficients corresponding to selected features.
• Robustness and Flexibility: Validated via simulations based on established single-cell simulation methods and accommodates alternative distributional assumptions and covariate adjustments.

Scientific Applications:

• Personalized Medicine: Identifies genetic patterns at the cellular level that associate with clinical outcomes to support personalized medicine analyses.
• Disease Research: Applied to real-world comparisons of healthy subjects and leukemia patients to identify genes linked to disease.

Methodology:

Organizes multi-omic single-cell data into tensors; performs regularized tensor regression using the clinical outcome as a predictor; applies elastic net and other penalization techniques for feature selection; validates performance via simulations based on established single-cell simulation methods.