Rapid, accurate, and memory-frugal sub-gene-level quantification of probe-based, multiplexed single-cell data
- 1University of Maryland
Abstract
Probe-based, multiplexed single-cell assays enable accurate transcriptome profiling while processing multiple samples in parallel. Their predefined probe sequences and additional sample barcodes introduce computational requirements that differ from those of poly(A)-based assays, for which many existing single-cell quantification tools were developed. We extend simpleaf to enable efficient demultiplexing and quantification of probe-based single-cell libraries. A new, cache-friendly dictionary specializes lookup to small probe references, while a shared thread budget balances parallel decompression and mapping. We extend RAD and downstream tools with reusable sample-scoped correction plans and external-memory collation of hierarchically barcoded data, retaining probe identities through quantification to enable rapid mapping and quantification with a small working memory budget. Across five 10x Genomics Flex datasets comprising 44 samples, simpleaf had the shortest runtime and lowest peak memory use of the three tools on every library, while producing gene-expression estimates, cell clusters, and marker genes concordant with cyto and Cell Ranger. Compared with cyto, simpleaf was on average 2 times faster and used approximately one-thirteenth as much memory. simpleaf also exports probe-level quantification results, allowing a range of analyses at sub-gene resolution. Here, we show that individual probe counts reveal expression changes obscured by gene-level aggregation, identify the transcript groups contributing to gene-level changes, and provide insight into splicing dynamics across cell types. These results demonstrate the computational and biological value of retaining probe-level information in multiplexed singlecell analysis.
Related
- Accurate quantification of protein concentrations and state populations by 2D 1H-N Q-BEST NMR
- Brain-wide neuron quantification toolkit reveals strong sexual dimorphism in the evolution of fear memory
- Accurate and scalable demultiplexing of single-cell RNA sequencing using BEACON
- Efficient and accurate mixed model association tool for single-cell eQTL analysis
- VIGAT: VLM-guided gene activity quantification from single-cell ATAC-seq
BibTeX
@misc{he2026rapid,
title = {{Rapid, accurate, and memory-frugal sub-gene-level quantification of probe-based, multiplexed single-cell data}},
author = {D. He and Y. Gao and R. Patro},
howpublished = {bioRxiv},
year = {2026},
month = nov,
doi = {10.64898/2026.11.05.764594},
url = {https://www.biorxiv.org/content/10.64898/2026.11.05.764594v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Rapid, accurate, and memory-frugal sub-gene-level quantification of probe-based, multiplexed single-cell data
AU - D. He
AU - Y. Gao
AU - R. Patro
PY - 2026
DA - 2026/11/05
PB - bioRxiv
DO - 10.64898/2026.11.05.764594
UR - https://www.biorxiv.org/content/10.64898/2026.11.05.764594v1
AB - Probe-based, multiplexed single-cell assays enable accurate transcriptome profiling while processing multiple samples in parallel. Their predefined probe sequences and additional sample barcodes introduce computational requirements that differ from those of poly(A)-based assays, for which many existing single-cell quantification tools were developed. We extend simpleaf to enable efficient demultiplexing and quantification of probe-based single-cell libraries. A new, cache-friendly dictionary specializes lookup to small probe references, while a shared thread budget balances parallel decompression and mapping. We extend RAD and downstream tools with reusable sample-scoped correction plans and external-memory collation of hierarchically barcoded data, retaining probe identities through quantification to enable rapid mapping and quantification with a small working memory budget. Across five 10x Genomics Flex datasets comprising 44 samples, simpleaf had the shortest runtime and lowest peak memory use of the three tools on every library, while producing gene-expression estimates, cell clusters, and marker genes concordant with cyto and Cell Ranger. Compared with cyto, simpleaf was on average 2 times faster and used approximately one-thirteenth as much memory. simpleaf also exports probe-level quantification results, allowing a range of analyses at sub-gene resolution. Here, we show that individual probe counts reveal expression changes obscured by gene-level aggregation, identify the transcript groups contributing to gene-level changes, and provide insight into splicing dynamics across cell types. These results demonstrate the computational and biological value of retaining probe-level information in multiplexed singlecell analysis.
ER -