CellMeSH

CellMeSH performs probabilistic cell-type identification by mapping clusters from single-cell RNA sequencing (scRNA-seq) to cell types using a gene–cell-type database constructed from indexed literature such as MeSH.

Key Features:

• Automated Annotation: Maps scRNA-seq-derived clusters to cell types based on cluster-specific gene expression patterns.
• Comprehensive Database Construction: Constructs a database of gene–cell type associations by automatically linking information from millions of publications using indexed literature resources such as MeSH.
• Probabilistic Query Methodology: Employs a probabilistic method to query the constructed database and retrieve cell-type information robust to noise in literature-derived associations.
• High Accuracy Performance: Reports up to 60% top-1 and 90% top-3 accuracy on human datasets and up to 58.8% top-1 and 88.2% top-3 accuracy across three mouse datasets.

Scientific Applications:

• scRNA-seq cluster annotation: Annotates clusters from single-cell RNA sequencing to identify and quantify cell types for genomics and transcriptomics studies and to assess cellular heterogeneity across species.

Methodology:

Database construction via automated extraction of gene–cell type associations from scientific literature (e.g., MeSH) and a probabilistic querying framework for cell-type retrieval.