CellMeSH: probabilistic cell-type identification using indexed literature.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) is a powerful technique for analyzing cellular heterogeneity in complex biological samples.
• Current scRNA-seq workflows rely on manual annotation of cell types, which is time-consuming and requires specialized biological expertise.
• Automating cell type annotation is crucial for scaling scRNA-seq data analysis and uncovering cellular diversity.

Purpose of the Study:

• To develop an automated method for cell type identification in scRNA-seq data.
• To create a comprehensive and up-to-date database of gene-cell-type associations from scientific literature.
• To improve the accuracy and efficiency of cell type annotation in scRNA-seq experiments.

Main Methods:

• Introduced CellMeSH, an automated approach for cell type identification.
• Constructed a database linking gene and cell-type information from millions of publications.
• Employed a probabilistic query method for reliable information retrieval from the literature-derived database.
• Incorporated optional utilization of prior tissue or cell knowledge for enhanced annotation.

Main Results:

• CellMeSH demonstrates superior top-one and top-three accuracies compared to existing methods on mouse and human datasets.
• The automated database construction is more comprehensive and easily updatable than manual approaches.
• The probabilistic query method effectively handles noisy gene-cell-type associations from literature.
• CellMeSH successfully automates the laborious cell type annotation step in scRNA-seq analysis.

Conclusions:

• CellMeSH provides a robust and scalable solution for automated cell type annotation in scRNA-seq data.
• The system significantly reduces the reliance on expert biological knowledge for annotation.
• CellMeSH enhances the discoverability of cellular heterogeneity by improving annotation accuracy and efficiency.