Cytomulate: accurate and efficient simulation of CyTOF data.

Area of Science:

• Computational Biology
• Bioinformatics
• Immunology

Background:

• Cytometry by time-of-flight (CyTOF) generates complex, high-dimensional single-cell data.
• Numerous analysis tools exist for CyTOF data, but objective evaluation is lacking due to unknown ground truth in real datasets.
• Existing evaluation methods often rely on real-world data, hindering robust assessment of new algorithms.

Purpose of the Study:

• To develop Cytomulate, a novel, accurate, and reproducible simulation algorithm for CyTOF data.
• To provide a reliable benchmark for evaluating existing and future CyTOF analysis tools.
• To facilitate the objective assessment of computational methods for high-dimensional single-cell data.

Main Methods:

• Developed Cytomulate, a simulation algorithm specifically designed for CyTOF data characteristics.
• Validated Cytomulate's ability to capture key features of real CyTOF datasets.
• Compared Cytomulate's data distribution learning capabilities against established single-cell RNA-seq and generative models.

Main Results:

• Cytomulate accurately simulates various characteristics inherent to CyTOF data.
• The algorithm demonstrates superior performance in learning overall data distributions compared to other methods.
• Cytomulate outperforms single-cell RNA-seq tools (scDesign2, Splatter) and generative models (LAMBDA) in data simulation accuracy.

Conclusions:

• Cytomulate offers a robust platform for the development and objective evaluation of CyTOF analysis methods.
• The simulation approach addresses the need for ground truth in assessing bioinformatics tools.
• This work establishes a foundation for advancing computational strategies in single-cell data analysis.