Can topology find real cell types across scales?
HiDeF finds structures that persist across resolutions: it recovered 89/136 mouse cell types (Jaccard >0.5), proposed 273 cell-type communities from 100,605 cells, and linked SARS-CoV-2 Nsp13 to TLE/WNT proteins.
Source
HiDeF: identifying persistent structures in multiscale 'omics data
Study at a glance
- Design
- Computational / modelling — HiDeF persistent-homology community detection on Tabula Muris scRNA-seq and protein-interaction networks, including SARS-CoV-2–human BioPlex
- N
- N=100605 · 100,605 mouse single cells; 136 reference cell types; 273 HiDeF communities; 332 SARS-CoV-2-interacting human proteins expanded to 1,948 proteins
- Population
- Tabula Muris mouse single-cell transcriptomes and human/yeast protein-protein interaction networks (BioPlex/STRING), including SARS-CoV-2 host proteins
- Outcome
- Recovery of reference cell types (Jaccard/F1) and persistent multiscale protein communities, including a TLE/WNT community targeted by viral Nsp13
Structured fields used in claim comparison tables when every cited study has a complete layer.
Inspect source passages
Can topology find real cell types across scales?
These passages occur verbatim in the stored paper text. They support the points listed below; they do not verify every statement in the explanation.
HiDeF analyzed mRNA profiles of 100,605 mouse single cells.
mRNA expression profiles of 100,605 single cells gathered across the organs and tissues of mice
65% (89/136) of reference cell types had Jaccard >0.5 overlap with a HiDeF community.
65% (89/136) having a highly overlapping community (Jaccard index > 0.5) in the HiDeF hierarchy
What they did
Applied persistent-homology community detection (HiDeF) to Tabula Muris cell-similarity networks and to protein interactomes, comparing recovery of 136 reference cell types against TooManyCells and Conos, then clustered a BioPlex neighborhood of 332 SARS-CoV-2 host proteins.
What they found
HiDeF matched reference types better than comparators (65% with Jaccard >0.5), yielded 273 putative mouse cell types, and among 251 persistent SARS-CoV-2-proximal communities highlighted TLE proteins that inhibit WNT—suggesting viral hijacking of that pathway.
The limits
What it doesn't show
Better overlap with existing annotations is not proof of new biology; SARS-CoV-2 WNT hijacking is a network inference, not a virology experiment, and mouse atlas labels still constrain what ‘new cell types’ mean.
Key terms
- HiDeF
- Hierarchical community-detection method that keeps structures persisting across resolution scales.
- Persistent homology
- Topological idea that intrinsic data structures are those that survive many scales.
- Resolution parameter
- Tuning knob that makes clustering prefer larger or smaller communities.
- Jaccard index
- Overlap score; >0.5 defined a highly overlapping match to a reference cell type.
- Tabula Muris
- Mouse organ/tissue single-cell RNA-seq atlas used here (100,605 cells).
- TLE/WNT
- TLE proteins inhibit WNT; HiDeF placed them in a community with SARS-CoV-2 Nsp13 interactors.
Flashcards
0 of 10 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
HiDeF’s mathematical hook is:
Common questions
How many mouse cells were clustered?
100,605 single cells from Tabula Muris.
How well were reference types recovered?
65% (89/136) had a HiDeF community with Jaccard index >0.5.
How many putative cell types did HiDeF call?
A hierarchy of 273 putative mouse cell types.
What did the viral network suggest?
A persistent TLE community consistent with hijacking of WNT.
More on Gene expression