Enhancing Rare Cell Type Identification in Single-Cell Data: An Innovative Gene Filtering Approach using Bipartite Cell-Gene Relation Graph

Maziyar Baranpouyan, Hossein Mohammadi, Hojjat Torabi Goudarzi, Krishnaprasad Thirunarayan, Lingwei Chen

Abstract: A useful tool for examining cellular diversity is single cell RNA sequencing (scRNA-seq). However, the high dimensionality and technical noise of scRNA-seq data make analysis difficult. To address this issue, gene filtering has been widely adopted to minimize single cell data noise and enhance the quality of subsequent analyses. Nonetheless, existing gene filtering techniques may inadvertently omit critical but rare genes which are necessary for identifying rare cell types that play a pivotal role in comprehending many biological processes. A novel graph-based gene selection technique is suggested in this study with the aim of preserving the informative genes to better identify rare cell types. Our findings demonstrate that this technique enhances the identification of rare cell populations, providing a refined approach for scRNA-seq data analysis and potentially enabling earlier and more reliable disease detection.

External IDs:dblp:conf/bhi/BaranpouyanMGTC23