CytoSpill

CytoSpill reduces spillover noise in mass cytometry (CyTOF) data by quantifying and compensating channel-to-channel signal spillover to improve accuracy of cell population analysis.

Key Features:

• Implementation: Implemented as an R package for processing CyTOF data.
• Control-independent compensation: Quantifies and compensates spillover effects independently of single-stained controls and normalization beads.
• Knowledge-guided modeling: Incorporates knowledge-guided modeling to inform spillover quantification.
• Finite mixture modeling: Uses finite mixture modeling for component separation in signal distributions.
• Sequential quadratic programming: Applies sequential quadratic programming for optimization of spillover compensation.
• Validation datasets: Validated on five public CyTOF datasets including human peripheral blood mononuclear cells (PBMCs), C57BL/6J mouse bone marrow, healthy human bone marrow, chronic lymphocytic leukemia patient samples, and healthy human cord blood.
• Performance: Produces results comparable to single-stained control-based compensation in datasets with available ground truth.
• Subpopulation discovery: Reduces spillover effects on affected markers and facilitates identification of cluster-specific markers and novel subpopulations.

Scientific Applications:

• Large-scale immune profiling: Enables large-scale immune profiling by removing spillover artifacts that confound cell clustering in CyTOF datasets.
• Tumor immune microenvironment analysis: Applicable to profiling tumor immune microenvironments to improve identification of immune cell phenotypes.
• Immunotherapy and biomarker discovery: Supports discovery of immune-specific biomarkers and development of novel immunotherapies through more accurate CyTOF data.
• Comparative sample analysis: Facilitates comparative analyses across sample types such as PBMCs, bone marrow, cord blood, and patient-derived samples.

Methodology:

Knowledge-guided modeling combined with finite mixture modeling for signal component separation and sequential quadratic programming for optimization, implemented in R.