Thread by @acnavasolive

[1/8] My first paper is finally out at @NatureComms!

nature.com/articles/s4146…

We use evolutionary algorithms, simulations and unsupervised machine learning clusterization to study the dynamics of CA1 pyramidal cells during theta. See the thread! 👇🏽

[2/8] Previous works from the @LMPrida lab noted functional differences between deep & superficial CA1 pyramidal cells:
nature.com/articles/nn.40…
We have seen that during theta, they tend to segregate their firing and shift on demand

[3/8] To understand the mechanisms, we implemented a biophysically realistic model based on reconstructed morphologies from @NeuroMorphoOrg. Intrinsic and synaptic features were selected by a genetic algorithm.

Check the model here!
> github.com/acnavasolive/L…

[4/8] That gave us thousands of intrinsic and synaptic features that we then expressed in different morphologies, giving rise to heterogeneous synthetic cells. Then we submitted these individuals to a realistic collection of theta modulated glutamatergic and GABAergic inputs.

[5/8] We found that a given individual phase-locks differently to theta depending on intrinsic, synaptic and morphological factors and that this can change dynamically in a predictable way.

See here the same set of intrinsic factors expressed in four different morphologies!

[6/8] With a logistic regression model we were able to evaluate the influence of different factors and made several predictions, some of which we successfully validated experimentally. Others are open to the community for confirmation! ;)

[7/8] One of my favorite results: use Self-Organizing Maps to relate the firing preference with dynamical changes of input pathways. We actually predict mechs for reversed theta sequences!

See our #Matlab tool RhythSOM:
github.com/acnavasolive/R…
Thanks @EnrRodSeb! :D

[8/8] To sum up! We have used a combination of advanced computational and experimental techniques to understand how intrinsic, synaptic and morphological factors interact in a very complex, non-linear way to determine firing dynamics...

...which is, in fact, very low dimensional (peak-trough). Beautifully, neurons implement a kind of dimensionality reduction, from an incredibly complex system to a simplified response that dynamically changes depending on cognitive demands.

I can feel now my pyramidal cells activating while I wander around the house full of excitement, all the more knowing what factors determine their phase-locking preference!