Chand lab is at Cosyne! (and I’m here to look for postdocs on cell types and dynamics; reach out if you want to chat!).

It's hard to argue everything is everywhere once you unconfound your task variables more carefully (posting your fig for others)! Unfortunately, I haven't seen many that do so in mice because presumably it is difficult to train. I think Shin's work is one www.nature.com/articles/s41... of the few

Other changes from the previous: PhysMAP has been improved and now outperforms classifiers trained directly on data. It also matches other approaches like beta-VAEs while being several orders of magnitude more performant. We also now look at three new datasets including NPix Ultra and Allen VisBeh.

Third result: PhysMAP's visualization-centric approach can detect modality-specific batch effects! Distributional shifts in modalities can occur due to differences in experiment and this has a profound effect on classification. This has been completely unaddressed but we view as crucially important.

Second result: Combined with high-density recordings via Neuropixels and monosynaptic connectivity detection (GLM-CC, Kobayashi et al. Nat. Comms. 2019), this can be used to monitor several cell types at once plus their interactions (cortical microcircuit), simultaneously, without optotagging.

First new result: We demonstrate (for the first time in neuro?) simultaneous interactions of multiple cell types during behavior in an experiment! This is through label transfer: we use joint embedding and graph nearest neighbor classifier to take ground truth and identify cells in another dataset.

Picked a great weekend to visit NYC!

This is exactly the crucial slide on all of my talks :)

We interpret this as a way to connect structural properties (waveform, ISI, amp., multi-chan. waveform) and function of cell types (direction/orientation selectivity) in vivo. In addition, this also allowed us to study the functional interactions between these types and thus the neural circuit.

3/4

Applying WaveMAP to Neuropixels recordings, we found putative cell types that exhibited distinct physiological, functional, and laminar distribution properties closely matching known cell types such as narrow-spiking exc. cells and exc. projection neurons with high direction selectivity.

2/4

Not sure what everyone is panicking about 🫤 but my portfolio is doing great!

It’s GRFP day! My take every year: Congrats to the winners! unfortunately for most, today they received disappointment. I was rejected from GRFP/CSGF/NDSEG likely bc my GPA was awful. Here are scores for my F31, years later. 1 = best (1 to 9). If you’re good, they’ll see it eventually; keep at it.

I have proof that I was once great 😭

Saw this incredible joke at the dry cleaners. They're not wrong!

Got the charades win at the joint BU/Harvard/MIT neuro social. Sorry losers, maybe next year.