Do living and nonliving packings differ in their neighbour topology?
Counting how neighbouring points connect in 3D is enough to tell most living tissues apart from nonliving packings like granular piles and foams.
Source
Topological packing statistics of living and nonliving matter
Study at a glance
- Design
- Computational / modelling — Delaunay tessellation motif counting plus a graph-based optimal-transport distance applied to published 3D point clouds
- N
- No single N; datasets include 15 biofilm colonies per species, nine zebrafish brain regions, 90 embryo time windows, and 40 nonliving-system points in the combined atlas
- Population
- 3D point clouds from bacterial biofilms, zebrafish brain and embryos, other tissues, simulated granular packings, foams and stars
- Outcome
- Topological diffusion distance between neighbourhood-motif distributions and clustering in a low-dimensional embedding
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
The authors tessellated 3D point clouds (cell nuclei, bacteria, particles, stars) into Delaunay tetrahedra and counted how often each local neighbourhood motif appears. They then measured distances between motif distributions using a topological diffusion distance defined on a flip graph linking motifs one neighbour exchange apart. They applied this to biofilms of four bacterial species, zebrafish brain regions and developing embryos, and a combined atlas of living and nonliving systems.
What they found
Biofilms from the four species separated cleanly, tracking cell aspect ratio. The embryo's time ordering could be recovered from topology alone, without using timestamps. In the combined atlas, of 40 nonliving points only an industrial foam fell inside the region occupied by living systems. Randomly shuffling brain cell positions and relaxing them produced a different topology closer to polyurethane foam, suggesting growth history shapes packing.
The limits
What it doesn't show
It is a descriptive comparison of existing datasets, so which physical features drive the separation is only partly tested (one randomisation control on one tissue). The atlas mixes very different data sources and imaging methods, and choices such as the boundary parameter alpha are set per dataset. The link between curvature of the developmental trajectory and biological transitions is suggestive and described mostly in supplementary material. Many details (motif labelling, convergence with sample size) are in supplements not included in this text.
Key terms
- Delaunay tessellation
- A division of space into tetrahedra connecting nearest-neighbour points; the dual of the Voronoi diagram.
- Motif
- The local pattern of tetrahedra around one point, formed by that point and its immediate neighbours.
- Flip graph
- A graph whose nodes are motifs and whose edges join motifs that differ by one elementary neighbour exchange.
- Topological diffusion distance
- A fast relaxation of earth mover's distance on the flip graph that measures how much probability must move to turn one motif distribution into another.
- Multidimensional scaling
- A method that places items in a low-dimensional map so that map distances approximate a given distance matrix.
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
What does a 'motif' represent in this framework?
Common questions
Why ignore particle shape and chemistry?
Using only neighbour relations lets the same method compare systems as different as bacteria and stars, looking for universal ordering principles.
Why does growth matter for the topology?
Cells inherit positional and orientational correlations when they divide; randomising positions removes that memory and changes the motif distribution.
Does the method need training data?
No. Unlike machine learning classifiers, it only needs the point clouds being compared.
More on Granular matter