Machine learning predicts putative hematopoietic stem cells within large single-cell transcriptomics data sets.

Area of Science:

• Hematology
• Computational Biology
• Genomics

Background:

• Hematopoietic stem cells (HSCs) are vital for blood formation and are implicated in various blood disorders.
• Single-cell RNA sequencing (scRNA-seq) is a powerful tool for studying HSCs, but lacks traditional flow cytometry validation.
• There is a need for methods to identify HSCs within large scRNA-seq datasets.

Purpose of the Study:

• To develop a computational tool for identifying mouse bone marrow HSCs using scRNA-seq data.
• To establish a robust and broadly applicable method for HSC identification in single-cell transcriptomic studies.

Main Methods:

• Machine learning approaches were tested to score single-cell transcriptomes.
• The hscScore tool was developed to identify HSCs based on gene expression similarity.
• The tool was evaluated across multiple scRNA-seq datasets from different laboratories and technologies.

Main Results:

• The hscScore tool accurately identifies mouse bone marrow HSCs from scRNA-seq data.
• The method demonstrates robustness across different scRNA-seq technologies and experimental batches.
• The trained model and code are publicly available to the research community.

Conclusions:

• hscScore provides a fast and reliable method for identifying HSCs in scRNA-seq data.
• This tool facilitates the analysis of single-cell gene expression data in hematopoiesis research.
• The availability of hscScore supports broader adoption and advancement in the field.