search

brainhack-school2020 / haqqeez_bhs2020_project

Determine relevant statistical and machine learning tools best suited to consolidate neural data with behavioural data.
0 stars 0 forks source link

Interesting project! #1

Open MarkNelson86 opened 4 years ago

MarkNelson86 commented 4 years ago

Hi Zeeshan,

Your project sounds very interesting and well suited for this summer school. Do you have an ROI/cluster of interest? If not, you might be able to sort the mice into groups by behavioral measures and then compare across groups? This could be e.g. correct vs incorrect choices on small sets of frames, or % correct performance over larger time windows/the whole task. How many mice do you have btw?

Thanks, Mark

haqqeez commented 4 years ago

Hi Mark,

Thank you! Indeed, one method to interpret the neural data is to look only at the time points when a decision was being made on the touchscreen by the mouse. Unfortunately, I've only collected pilot data on about 2 animals for this task, so it wouldn't be ideal to do between-animal comparisons for now.

What I could do, however, is compare activity during correct/incorrect attempts within the same session for the same animal, as they attempt the task numerous times in a single session. I can also compare this across days for the same animal to see if correlations strengthen or change later on, when the animal has had good practice on the task, vs. early on when it was still learning the task. Does that make sense to you?

Zeeshan

MarkNelson86 commented 4 years ago

Hi Zeeshan,

Yes, sounds like a good plan. I am excited to see how it turns out.

Where are these 100 cells located in the mouse brain? Is the behavioral data also time-stamped by stimulus presentation?

You mention wanting to consolidate the data, perhaps some phases of the trial could be averaged out? E.g. maybe the drinking behavior is not so interesting or doesn't vary with learning.

haqqeez commented 4 years ago

Hi Mark,

These cells are located on the dorsal surface of the hippocampus (region CA1). And yes, the behavioural data is also time-stamped and can be synced to match the frames of the neural recordings.

As for averaging out the data, I think I'd like to select 'windows' of frames to look at that would be of interest, yes. Maybe a 2 second window centered around when the mouse makes a decision, and discarding the other data when the mouse is walking around and waiting for the next trial. Although, perhaps I'd like to include some less interesting windows as well (such as drinking behaviour) as a control to show that perhaps pairwise correlations between cells during these uninteresting windows (or randomly selected windows) don't show and changes across days of learning.

MarkNelson86 commented 4 years ago

Solid plan. Very interesting project. Look forward to seeing it develop.