Fig. 4

AMULET is robust to cell type similarity and improves increased with read depth. a Total number of nuclei and multiplets detected by each method. Differences in number of nuclei are due to differences in inputs (i.e., alignment (BAM) files and Cell Ranger cell annotations for AMULET, fragment files (Cell Ranger output) and ArchR cell annotations for ArchR). Overall, ArchR detects more nuclei as multiplets using default parameters than AMULET. b Reference annotations for islet 1. Islet 1 annotations correspond to alpha, beta, delta, and ductal cell types. c Multiplets detected by AMULET and ArchR for islet 1. Majority of multiplets detected were not shared between the two methods. Note: ArchR detected the majority of delta cells as multiplets. d Average recall for detecting artificially introduced multiplets in immune and islet samples with respect to median read count per nucleus. AMULET reaches its peak performance for ~25K median valid read pairs per nucleus. Differences between immune and islet cells likely stem from the overall differences in their accessibility levels and patterns