Data analysis

The barcodes and UMIs in the reads were extracted using UMI-tools (version 1.0.0) with the following command: umi_tools extract -I r1.fastq --read2-in=r2.fastq --bc-pattern=NNNNNNCCCCCC --read2-stdout. The reads were mapped by the aligning software HISAT2 (version 2.1.0) to the reference genome (GRCh38 and GRCm38). Read counts per gene were determined with featureCounts (version 1.6.4) and UMI-tools with the following command: umi_tools count --method=directional --per-gene --per-cell --gene-tag=XT. The duplicates in the unique molecular identifiers were discarded. Using the resulting counts, differentially expressed genes (FDR = 0.1) were extracted by the R library DESeq2 (version 1.20.0), which was also used to transform count data into regularised log data before clustering (heatmap3, version 1.1.6) and dimension reduction (UMAP, version 0.2.2.0).