KAVLI Suite2P with Datajoint

Forked from Suite2p: fast, accurate and complete two-photon pipeline on 05.04.2018. Added OASIS on same date. Algorithmic details in http://biorxiv.org/content/early/2016/06/30/061507.

Last commit

Last commit taken over from master in original repo: 08003ee6ec9a4a4ad87e600dcc053063b02ddac9

Changes

Made independent of the original "forced" folder structure (Mouse/Date/) to be able to call .tif folders directly. If mouse_name does not exist in make_db, process with non-prefixed folder structure.

What changed are the export filenames (changed throughout different .m files). Consider the following example:

OLD:

foldr = fullfile(ops.RegFileTiffLocation, ops.mouse_name, ops.date, ...
         ops.SubDirs{k}, sprintf('Plane%d', iplane));

NEW:

mouse_name = getOr(ops, 'mouse_name', []);
if isempty(mouse_name) % file names are fixed in make_db and no folder structure has to be assumed.
    foldr = fullfile(ops.RegFileTiffLocation, sprintf('Plane%d', iplane));  
else
    foldr = fullfile(ops.RegFileTiffLocation, ops.mouse_name, ops.date, ...
             ops.SubDirs{k}, sprintf('Plane%d', iplane));    

This gets rid of the pre-defined folder structure that contains both mouse_name and date in the path.

Standard setup

To run Suite2P (and omit the mouse_name/date folder structure):

• It is now assumed that the data folder (RootDir) contains all the .tif files that should be analyzed together and they have some logical ordering such that natsort can take care of putting them into the right order.
• I keep the following folders all the same: RootDir = RootStorage = RegFileRoot = ResultsSavePath = RegFileTiffLocation: Those are all network/fileserver locations. This saves everything (including registered tif files) to the original data folder. temp_tiff is set to a local drive.
• The tif files are either scanimage bigtifs with different channels being chained together like channel1-channel2-channel1-channel2-... or mimic this structure (either uint16 or int16).
• If the data contains a second (red, non-signal) channel, then the setup should be:

db.expred                 = 1;
db.nchannels              = 2;
db.nchannels_red          = 2;
db.AlignToRedChannel      = 0; % Change this to 1 if red channel has nice structural signal
db.REDbinary              = 1;
db.redMeanImg             = 1;

This will create bin files plane1.bin and plane1_RED.bin in the export folder and create separate motion corrected tif stacks for each channel. It will calculate a mean red image (after motion correction) and calculate probabilities of cells being red, reflected in the export as dat.stat.redcell, dat.stat.notred, dat.stat.redprob (it runs identify_redcells_sourcery for that).

Careful! Assumes now that the experiments all have the same number of green and red channels (within one data folder). To do: (1) Test normal suite2p folder structure.

Datajoint

This version integrates a datajoint based job handler (see +datajoint). Documentation will follow.