Back

Sample-level modeling of single-cell data at scale with tinydenseR

Milanez-Almeida, P.; Schildknecht, D.; Linder, M.; Brachmann, S. M.; Weiss, A.; Adler, F.; Lenticchia, S. C.; Meistertzheim, M.; Wild, S.; Cuttat, R.; Jayaraman, P.; Lee, L. H.; Mulvey, T.; Hassounah, N.; Crafts, G.; Quinn, D. S.; Orlando, E. J.

2025-11-30 bioinformatics
10.1101/2025.11.26.690752 bioRxiv
Show abstract

Single-cell studies now routinely encompass hundreds of samples and millions of cells, offering unprecedented opportunities to link sample-level phenotypes with cellular and molecular states. However, current workflows often depend on cell-level inference and rigid clustering, which can distort significance and obscure subtle, continuous variation, in particular for complex experimental designs. Here, we present tinydenseR, a clustering-independent framework that enables robust, scalable, and statistically sensitive detection of differential cell states, outperforming existing workflows in speed, memory usage, and biological resolution. Technology-agnostic at its core, tinydenseR works seamlessly on scRNA-seq, flow, mass and spectral cytometry. Across synthetic benchmarks, a preclinical xenograft model, two immuno-oncology trials and a multi-study atlas, tinydenseR uncovers disease and treatment history-associated effects, including subtle within-cluster heterogeneity. Designed to accelerate discovery in clinical, preclinical, and translational research, the open-source package is available at GitHub.com/Novartis/tinydenseR.

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.