Rm 25a @10:45 to hear about first full cell assembly reactivation in sleep in primates (human or nonhuman), comparing rvn for newly-learned and old memoranda. Implications for memory linking, integration. @sabbaspoor.bsky.social and I would love your feedback! #sfn25 🧠🧪

A brain region's importance is based on change in dynamics upon removing it. The brain is dynamic and so is "region importance". So while we can think of multivariate spatiotemporal patterns, it's also possible to focus on regions if one wishes to. Though in my view multi-region is the way to go.

After careful curation of single units from CA1 and connected structures, we identified cell assemblies - repeated patterns of synchronous firing (typically N~100, some approaching 300). Some assemblies were more active in NEW trials, others OLD. But did either or both groups reactivate in sleep?

= massive labor of love, built over yrs of foundational work, made possible with the laudable contributions from international students @sabbaspoor.bsky.social (lead, now w/ @jordan-farrell.bsky.social) and @aymanaljishi.bsky.social, and with BRAIN Initiative NINDS + Whitehall Foundation funding.

BIC/AIC across a range of states feels inefficient. Tried using clustering (KMeans + silhouette analysis) to pre-estimate and then fine-tune within a range around the given numbers, but it systematically overestimates states on a simulated dataset. Any suggestions (2/2)?

Can you add me to the list please?

Thank you for sharing this work. Can I use SIMPL for nonspike data like calcium transients? If so, do you recommend changing the Poisson distribution for the neural signal?