Finally, we showed how iterative coarse-graining (ICG) can identify informative neuronal reconfigurations related to task dynamics in the zebrafish.

I bet this approach will be helpful for future behavioural studies!

These task induced zebrafish reconfigurations are most flexible and informative at the mesoscale (thousands of cells).

Interestingly, mice showed two distinct peaks of reconfiguration at the micro- and mesoscales, supporting the theory of multiple scale-dependent information processing streams.

Not done there, we explored beyond rest into behaviour and found conserved static scaling but evidence of a privileged temporal scale in zebrafish aligned with the stimuli.

That is, a multiscale organisation permits flexibility to adapt to behaviourally relevant timescales.

The few multiscale networks that recapitulated the diverging info theory regimes/nonGaussianity also revealed more beneficial features such as maximising dynamic range and communicability.

Contrasting MANY network null models, we found that only hierarchical modular networks could controllably change the slope and reproduce the empirical static and dynamic scaling

Though geometric networks worked well up to a scale dependence. See some related recent Aussie work by Pang et al 2023

We wanted to understand the origin of this scaling.

The C. elegans recordings contained tagged cells and a known anatomical connectivity, which let us show cross-scale functional pairings were far more likely if the cells had a physical connection.

That is multiscale structure and function!

This multiscale organisation also provides a functional dial to operate across varying timescales!

Again, with a conserved slope (despite the noisier measure), from sparse and fast activity at the cellular scale to more continuous fluctuations at the macroscale.

How could this pluralistic information capacity emerge?

We found evidence for a precise self-similar coordination consistent across species (invertebrates, fish, mammals) and balanced between two extremes (identical/independent).

This was unique against various nulls!

We discovered a divergence in redundant information capacity at coarser scales.

This shows that the cellular scale is optimised for efficiency, while coarser ensembles support resiliency. This mirrored a shift from heavy-tailed to Gaussian correlation distributions.

We opted for a Kadanoff iterative coarse-graining approach among the growing multiscale techniques. This was dyadic (maximal log spread), comparable to standard analyses (correlations), and fast. Letting me track activity across scales

Crazily, I applied this to as many species as possible!

Hi BlueSky fam, for my first post and to celebrate our recent paper being physically published I thought I’d do a summary thread!

This has been my most favourite (and toughest) work to date.

Please help share around!!

www.cell.com/cell/abstrac...
(Reach out if you can’t access)