Efficient Matching of Single Cells and Cell Types

Next-generation RNA sequencing enables investigation of cellular diversity within tissues at the single-cell level, allowing unprecedented precision in discovering new cellular functions or monitoring disease progression. There is a need for interpretable, scalable, and interactive methods to explain cell types in a single-cell RNA -seq dataset. Here, we design a rank statistic-based annotation method, which simultaneously computes a visual projection and assigns cell types unbiasedly and independent of dataset size. We implement an interactive exploration tool, scAnnotator, as a part of the Orange data mining suite. We elucidate and benchmark annotation of diverse cells in human pancreas tissues blood, achieving up to 84.5 % accuracy compared to manually annotated cells. Easy-to-use data exploration with scAnnotator within Orange has the potential to aid discovery and diagnostics, as well as bring scRNA-seq data closer to decision-makers.