Cell Type Classification and Discovery across Diseases, Technologies and Tissues Reveals Conserved Gene Signatures of Immune Phenotypes
Author(s): Mathew Chamberlain, Nima Nouri, Andre Kurlovs, Richa Hanamsagar, Frank O. Nestle, Emanuele de Rinaldis, Virginia Savova.
The classification of immune cell phenotypes in single cell data is a major challenge in biology research today. Here, we present a novel machine learning approach, SignacX, which uses neural networks trained with flowsorted gene expression data to classify immune cellular phenotypes in single cell RNA-sequencing data. We demonstrate that SignacX accurately classified single cell RNA-sequencing data across diseases, technologies, species, and tissues, and outperformed other leading methods in immune phenotype classification, particularly for classification of CD8 and CD4 T cell subsets. We used the annotations generated by SignacX to identify conserved and tissue-specific gene expression-based signatures of immune cell types. Next, we defined immune-relevant precision medicine candidate drug targets in rheumatoid arthritis using single cell data from human synovium. A full release of this workflow together with detailed vignettes, an interactive data portal and freely accessible software that is integrated with Seurat and is easy to use can be found at our GitHub repository (https://github.com/Sanofi-Public/PMCB-SignacX).