SSMD

SSMD estimates dataset-specific cell-type signature genes and relative cell-type proportions from mouse transcriptomic, DNA methylation, and ATAC-seq datasets to enable deconvolution of complex mouse tissues.

Key Features:

• Non-parametric dataset-specific signature identification: Employs a novel non-parametric approach to identify dataset-specific cell-type signature genes tailored to each experimental condition.
• Community detection for cell-type and marker refinement: Uses a community detection method to establish and refine cell types and their marker genes across variable genetic and phenotypic backgrounds.
• Constrained matrix decomposition for proportion estimation: Applies constrained matrix decomposition to determine relative cell-type proportions across diverse experimental platforms.

Scientific Applications:

• Handling variable cell types and markers: Manages variability in cell types and marker genes arising from different genotypic and phenotypic conditions in mouse models.
• Cross-platform deconvolution: Operates across diverse experimental platforms to maintain consistent deconvolution performance regardless of data source.
• Small-sample studies: Functions with limited training data, enabling application in studies with small sample sizes.
• Comprehensive cell-type estimation: Estimates proportions for over 35 cell types across systems including blood, inflammatory tissues, the central nervous system, and hematopoietic systems.

Methodology:

SSMD implements a semi-supervised framework combining non-parametric identification of dataset-specific signature genes, community detection to define and refine cell types and markers, and constrained matrix decomposition to estimate relative cell-type proportions; it is applied to transcriptomics, DNA methylation, and ATAC-seq data.