TimeFlow 2: An Unsupervised Cell Lineage Detection Method for Flow Cytometry Data.

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Cell lineage detection is crucial for understanding cell differentiation.
• Existing methods often require prior knowledge or temporal data.
• Large flow cytometry datasets present computational challenges for lineage inference.

Purpose of the Study:

• To develop a novel computational method, TimeFlow 2, for cell lineage inference.
• To enable lineage detection from static, unordered flow cytometry data.
• To accurately model cell differentiation pathways and marker dynamics.

Main Methods:

• TimeFlow 2 utilizes cell orderings and defines coarse cell states along pseudotime segments.
• It constructs cell state paths and groups them using an optimal transport-based cost function.
• The method was applied to healthy bone marrow samples and compared against established techniques.

Main Results:

• TimeFlow 2 accurately assigned monocytes, neutrophils, erythrocytes, and B-cells to distinct differentiation pathways.
• Inferred marker dynamics showed high correlation across corresponding lineages in multiple patients.
• TimeFlow 2 demonstrated superior performance on flow cytometry data and competitiveness on mass cytometry data.

Conclusions:

• TimeFlow 2 provides a robust, data-driven approach for cell lineage inference.
• The method facilitates modeling and comparison of marker dynamics across diverse cell lineages.
• Accessible source code and tutorials support the adoption of TimeFlow 2 in biological research.