standalone deconvolution questions about suite2p HOT 1 CLOSED

Reply from Marius via email:
i) Is the output of the algorithm particularly sensitive to the parameters provided?

It is sensitive to the timescale of the Calcium indicator, but not much to anything else. There is an option to estimate the timescale parameter directly (or even the full kernel), but it assumes an underlying Poisson process, and it can be severely biased otherwise (although not many people recognize this problem...). As such, I currently recommend setting this timescale by hand, to a value you believe your Calcium indicator has. I venture to say, the output won't be significantly different if you miss the true timescale by a factor of +/- 50% or so.

Also, don't set the threshold too low, or it will take forever and you will get a noise floor.

ii) When using Suite2P, how do you suggest handling missing data points? Would setting them to zero Vs stripping them out (shortening the recording) bias the output of the deconvolution particularly?

Setting them to 0 would introduce artifacts in the output. The best solution seems to be interpolating between the nearest available timepoints, and it should not impact the deconvolution much.

from suite2p.