Collective behaviour changes with age

Yushi Yang has recently completed his PhD project in Bristol focusing on the collective behaviour of a specific living organism: Zebrafish, a small, semi-transparent fish coming from tropical fresh waters employed extensively in biology as a model organism. Quantitative studies of the collective behaviour of animals have been performed in many different contexts, from midges to starlings and fish are not excluded. However, Yushi has been one of the first to concretely realise an experimental setup allowing for the three-dimensional tracking of the trajectories of a conspicuous number of individuals (about 50). 

Together we looked into some simple models making physical sense of the emerging patterns of behaviour. This approach has serious epistemological challenges: what justifies the reduction of the social behaviour of an animal to a minimal physical model? what do we actually learn in the process of reduction of complexity? what predictive power is associated with such results?

In our approach, we pursued the identification of quantitative variables describing the evolving physical features of groups of fish: their average orientation, their relative distances, and their speed. We actually did not attempt to measure any specific form of interactions between the fish: these are certainly taking place by means that can be complex (vision, hydrodynamic feedback, maybe some form of signalling) but, in our description, they only appear as effective terms. In fact, what we seek is a small number of physical properties that allow us to organise the data: for example, we identify a scaled persistence length that appears to control the degree of polarisation of groups of fish of different ages: incoherent groups of older fish and well-coordinated groups of younger fish. It is indeed a key result of Yushi’s work that the behaviour of zebrafish changes markedly with age and that non-trivial correlations appear to be present only for the younger groups. We are able to map these different behaviours on a single master curve and model it with a simple physical model dominated by alignment interactions and delay (or inertia) in reorientation. 

While the different physical ingredients of the simplified model do not have an immediate biological interpretation, they allow us to narrow down the spectrum of relevant variables controlling collective behaviour and justify further research aimed at providing a causal, biological link between the behaviour of the group and the phenotypical characteristics of the individuals.

This research has been published in PLoS Computational Biology 

Yushi Yang , Francesco Turci, Erika Kague, Chrissy L. Hammond, John Russo, C. Patrick Royall, Dominating lengthscales of zebrafish collective behaviour, PLoS Comput Biol 18(1): e1009394, (2022), doi: 10.1371/journal.pcbi.1009394

Recontructed 3D positions and orientations of a group of young zebrafish (left) and original video (right).

Morphology analysis of bone malformation

Soft matter is a broad field, ranging from colloidal particles to micelles, from proteins to cellular tissues. As I mentioned in a previous post, a physics-based approach to characterising the morphology of soft biological matter can be very insightful. It provides simple, geometrical and structural metrics to identify variation in tissues. We recently demonstrated this approach in a rich article in Nature’s Bone Research, a journal dedicated to the quantitative study of bone properties in different species.

A few years ago, Erika Kague, from Physiology at the University of Bristol, asked me to think about her problem. By inducing genetic mutations, she can generate individual zebrafish (Danio rerio) – a common model organism in biology – with various malformations. These she can investigate to understand the effect of bone mineral density on the insurgence of osteoporosis. The issue is that one may want to rapidly, systematically and quantify the amount of malformation.

To do that, I worked with talented PhD student Yushi Yang to automatise a computer vision workflow of segmentation of three-dimensional images, detection of relevant anchor points and determination of several geometrical and structural metrics. Some of these quantities have been inspired by thermal soft matter analogues, such as the porosity of gels; others rely on graph theory metrics. The entire manually tagged database can be augmented using deep learning U-nets that Stephen Cross (Wolfson Bioimaging Facility, Bristol) optimised for our use case.

The result is a detailed characterisation of which genes promote certain kinds of changes in the bone structure. The data reveal a delicate balance between too little and too much bone mineral density to minimise the chances of osteoporosis.

Bone structure in wild type (wt) and mutant (sp7) zebrafish. The mutant clearly displays multiple signs of malformations.

The full article is here

Erika Kague, Francesco Turci, et al., 3D assessment of intervertebral disc degeneration in zebrafish identifies high and low bone density linked to disc disease, Bone Research 9, 39 (2021), doi: 10.1038/s41413-021-00156-y

Segmenting 3d biological data

I have recently been given the opportunity to study the segmentation of 3D data.  The group of Dr. C. Hammond of the School of Physiology, Pharmacology and Neuroscience in Bristol studies malformation in tissues of  Zebrafish  a model organism which can be genetically manipulated relatively easily .

A major task is to identify bone malformation or osteoarthritis. Hammond’s group manages to image hundreds of Zebrafish in three dimensions so that bone structures can be visualised. Identifying bone deformations in the spine, for example, is key to associate them to specific genetic marker. To do so, a quantitative analysis of the structure of the individual vertebrae is necessary.

It turned out that it is possible to do this via image analysis techniques that are publicly available in Python: the key libraries that I employed are scipy. ndimage and scikit-image.  Identifying the vertebrae in 3d means to perform  a segmentation of volumes and surfaces in 3d images.

An example of the vertebrae, individually resolved, can be visualised in 3d here below: