TOAST

TOAST performs reference-free deconvolution and cell-type-specific and cross-cell-type differential analysis on gene expression microarrays, DNA methylation microarrays, and RNA-seq to estimate cellular composition and detect differential signals in heterogeneous tissue samples.

Key Features:

• Reference-free deconvolution enhancement: Employs an iterative computational method that identifies cell-type-specific features to refine composition estimates without relying on external reference datasets.
• Cross-cell-type differential analysis: Detects both cell-type-specific and cross-cell-type differential signals in complex tissue samples, enabling detection of subtle variation in gene expression and DNA methylation across cell types.
• Variable selection improvement: Improves variable selection during deconvolution by iteratively focusing on cell-type-specific features to increase the accuracy of identified biological signals.

Scientific Applications:

• Genomics and epigenetics studies: Estimation of cellular composition and attribution of expression or methylation signals to specific cell types in heterogeneous tissues.
• Multi-platform analyses: Applicable to gene expression microarrays, DNA methylation microarrays, and RNA-seq datasets for cross-technology tissue deconvolution.
• Validation and benchmarking: Performance evaluated using simulation studies and application to six real datasets.

Methodology:

The core methodology alternates iteratively between identifying cell type-specific features and estimating the composition (proportions) of those features within samples.