Using machine learning technology to identify platelet and megakaryocyte types in single-cell transcriptome data

Platelets play crucial roles in blood clotting, wound healing, and the tumor microenvironment. However, identifying them in single-cell transcriptome data is challenging. This study integrated different data sources to construct a rapid and accurate platelet identification model called Pltscanner. The model demonstrated high sensitivity and specificity during validation, capable of recognizing platelets by their characteristic gene expression profiles. Pltscanner also effectively clustered with self-isolated platelet data and accurately predicted vascular distribution in spatial transcriptome data. Compared to existing tools, Pltscanner is faster and provides more accurate annotations. Furthermore, the research revealed that platelets are not a single type but can be categorized into three classes, each potentially playing different roles in the tumor microenvironment. This study contributes to a deeper understanding of platelet functions in the tumor microenvironment and lays the foundation for investigating the mechanisms underlying tumor development.