Rapid, accurate, and memory-frugal sub-gene-level quantification of probe-based, multiplexed single-cell data

simpleaf demultiplexes and quantifies probe-based multiplexed single-cell libraries across five 10x Genomics Flex datasets (44 samples) with the shortest runtime and lowest peak memory.

• Why it matters: Probe-based assays add predefined probe sequences and sample barcodes, creating computational demands that existing poly(A)-focused tools may not handle efficiently.
• What they did: The method uses a cache-friendly dictionary for probe lookup, a shared thread budget for parallel decompression and mapping, and extends RAD with reusable sample-scoped correction plans plus external-memory collation.
• The result: Across every library, simpleaf matched cyto and Cell Ranger in gene-expression estimates, cell clusters, and marker genes, while being ~2× faster and using ~1/13 the memory; probe-level counts reveal expression and splicing dynamics hidden at gene level.

AI summary of the abstract · may contain errors · not medical advice · read the paper.