Extended Data Fig. 1 Schematic of a microSPLiT sequencing construct.

Only the polyA-primed construct is shown as an example. The random hexamer-primed construct would harbor six random nucleotides in place of the dT15 sequence.

Extended Data Fig. 2 Example data before filtering, shown as a barcode-rank plot.

A red line indicates a chosen threshold for filtering out low-quality cells and debris. a , A high-resolution experiment; data were filtered at 400 total UMIs/cell. Before filtering: 884,736 barcode combinations; after filtering: 60,831 combinations, corresponding to cells. b , A poor-resolution experiment; data were filtered at 350 total UMIs/cell. Before filtering: 884,736 barcode combinations; after filtering: 644 combinations.

Reporting Summary

Supplementary Tables 1–4

Supplementary Table 1: cost estimate of microSPLiT. Supplementary Table 2: cost estimate and comparison of high-throughput bacterial scRNA-seq methods. Supplementary Table 3: sequences of oligonucleotides and source barcode plates used in microSPLiT. Supplementary Table 4: barcode sequence files for STARsolo analysis.