Abstract:
|
Jointly-sequenced multiomic single-cell datasets of two modalities (for example, RNA and protein) offer biologists insight on cell-type identification at single-cell resolution via existing dimension-reduction methods that aggregate information from both modalities, but do not quantify how much information is common to both modalities or distinct to each modality. We develop a new dimension-reduction method called Tilted-CCA to fill this gap, where we formalize information as geometric features based on the nearest-neighbor graphs and build upon the statistical foundation of Canonical Correlation Analysis (CCA). We demonstrate the utility of Tilted-CCA in two ways by analyzing PBMC cells, sequenced via CITE-seq (for RNA and protein) and 10x (for RNA and ATAC). First, we show that Tilted-CCA offers insight on designing the smallest antibody panel where the resulting protein expressions provide cell-type separations not reflected in RNA. Second, we show for this biological system, RNA and ATAC are tightly intertwined such that neither modality contains much distinct information, whereas RNA and protein each contains more distinct information not reflected in other modality.
|