The abstract is read at the publisher; the summary is JClub's.
GraphHDBSCAN* recovers biologically meaningful hierarchical cell populations from scRNA-seq data, matching curated references more closely and resolving monocyte subpopulations.
- Why it matters: scRNA-seq needs hierarchical organization, but most methods produce flat clusters and struggle in very high-dimensional, noisy spaces.
- What they did: The authors propose GraphHDBSCAN*, a graph-based extension of HDBSCAN* with an effectively hyperparameter-free hierarchical stage, evaluated on multiple scRNA-seq datasets.
- The result: Across datasets, the method agrees more closely with curated biological references than other hierarchical clustering approaches and also produces flat partitions that outperform community-detection methods.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.