The lung has a lot of structures and diverse cell types working as different divisions and offices, to work highly coordinately and perform their functions to avoid disease. But some essential and functional cells are hidden among the whole cell population, so what we did is to grab them out of the population and figure out how they contributed to the immune response in NSCLC (non-small cell lung cancer).
With the advance of spatial and single cell technology, RNA-seq data with higher resolution is now starting to become available, which is greater dimensionality with both single cell resolution and now with spatial information. With bulk transcriptomic and scRNAseq data still widely used and prevalent, there is still a need to get accurate cellular proportions for spatial bulk RNA data (e.g., Visium or GeoMx data) and cell annotation for scRNA-seq data.
Here we also developed a workflow to create reference and generate cell type-specific markers from specific reference dataset(s). This can then be applied for conducting reference-based cell annotation and marker-based cell annotation for the dataset(s) of interest. At this stage, six annotation methods commonly used for single cell analysis are integrated into our workflow. An ensembled majority vote strategy is currently implemented to get consensus annotation and cell identification confidence levels.
Our workflow has been applied to both traditional scRNA-seq data and spatial multi-omics Single-Cell imaging data (Nanostring CosMx Spatial Molecular Imager). Our initial results suggest reliable annotation can be generated for scRNA-seq data in agreement with the published work. The agreement is to a lesser extent for CosMx SMI data, and the reason for that is being investigated. This work will present the exploratory investigations and cell annotations analysis results using the proposed ensemble approach for the new CoxMx SMI data.