Illumina’s bcl2fastq2 software (v2.19.1) was used to demultiplex the sequencing output to 12 plate-level FASTQ files (1 per 96-well plate). A python-based pipeline ( was used to (i) demultiplex each plate-level FASTQ file to 96 cell-level FASTQ files, trim 52 nucleotide reads to 35 nucleotides, and append UMI information from read 1 (R1) to the header of read 2 (R2); (ii) perform genomic alignment of R2 with Bowtie2 (v2.2.2) using a concatenated hg19/External RNA Controls Consortium (ERCC) reference assembly; and (iii) convert aligned reads to gene-level counts using a modified version of the HTSeq (v0.5.4p1) python library that identifies reads aligning to the same location with identical UMIs and reduces them to a single count. One UMI-corrected count was then referred to as a “transcript.” The pipeline was configured with the following settings: alignment quality (min_bc_quality) = 10, barcode length (bc_length) = 6, UMI length (umi_length) = 5, cut_length = 35.