This toolbox is for processing and analyzing 2P imaging data aquired during behavior. It should give you everything you need to analyze how behavioral events are related to your imaging data.
- neural data: processed .mat files from Suite2P
- behavioral event data: a block.mat file from Rigbox
- experimental metadata: a Timeline.mat file from Rigbox
- video data: an eye_proc.mat file from facemap (if you want to analyze videos)
- Initialize your experimental session by creating the
expInfo
struct. This will hold most of the experimental metadata (e.g., stimulus contrasts & onset times) and some behavioral data (e.g., trial-by-trial choices)
expInfo = initExpInfo({{'LEW031'}},{{'2020-02-03',1,[1]}});
- Process your experiment. This does three things: (1) loads the
block
andTimeline
structs intoexpInfo
, (2) generatesbehavioralData
to hold trial-by-trial wheel movements and event times, and (3) generatesneuralData
to hold full traces as well as trial-by-trial event-aligned responses for each cell
[expInfo, neuralData, behavioralData] = processExperiment(expInfo);
- Retrieve data from video ROIs that were processed in Facemap
eyeData = getEyeData(expInfo);
NB: The identity/indexing of video ROIs will depend on how you decide to process your videos in Facemap. The conventions for toupee are described below
This code uses cross-correlation to synchronize video frame times and global Timeline times, by comparing the SVD motion energy of an ROI containing the lick spout and optical beam events that were logged at the lick spout itself. An output plot lets you inspect the sync by overlaying video motion energy and optical lick events.
We use this analysis since UDP signal times were not logged between eye & experiment computers at the B2 imaging rig. If you have UDP times available in your Timeline
struct, these can be used to sync the two time series instead.
- Generate trial-by-trial event-aligned activity for each video ROI and store it in
eyeData
[eyeData] = alignFace(expInfo, eyeData, behavioralData);
A few scripts for plotting the choice performance over a session are available.
Plots a psychometric curve for a 2AFC task (one curve) or a biased-block 2AFC task (two curves). It reads the block file to automatically determine which task was run. A single session can be plotted by calling either:
plotPsychometric({{'LEW025'}}, {{'2019-11-15',1}})
plotPsychometric(expInfo)
Multiple sessions are concatenated by calling plotPsychometric(mouseList, expList)
, where:
mouseList = {{'Mouse1'}} or {{'Mouse1'},{'Mouse2'},{MouseN'}}
expList = {{'2018-06-10',2,[2 3]},{'2019-03-27',1,[1]}}
(NB: the final vector typically holds the same value as the second integer. There are occasional old files that concatenate over multiple sessions (hence the vector), but these are rarely encountered)
Multiple sessions can also be plotted by calling plotPsychometric(expInfo)
, if expInfo
is a struct of multiple experiments (see initExpInfo.m
)
In the plot, green curves correspond to high-value left blocks, and orange curves correspond to high-value right blocks.
Plots a rolling mean of the proportion of right choices over time. This is useful when inspecting a biased-block 2AFC where you expect the mouse to go through epochs of more left or more right choices. It also outputs pRight
, which is a vector of the rolling mean of p(right) over time.
In the plot, green corresponds to high-value left blocks, and orange corresponds to high-value right blocks.
This function can be called the same way as plotPsychometric(varargin)
, above.
Most data analyses will require comparing some types of trials to other types of trials (e.g., correct vs. incorrect, left vs. right). Generally, the repository calls these trial conditions and there are a few scripts that will let you index the trials of your choosing.
Prepares specific name-value pairs to identify the trial conditions you want to include in your analysis. Available pairs are:
'repeatType'
:'all'
,'random'
, or'baited'
('baited' trials are those that were repeated after an incorrect response)'movementDir'
:'all'
,'cw'
, or'ccw'
('cw' refers to the movement a mouse would make to correctly report a left-side stimulus)'movementTime'
:'all'
,'early'
, or'late'
(refers to when the mouse made its first movement with respect to the cue delay)'highRewardSide'
:'all'
,'left'
, or'right'
(refers to which stimulus side had a high-value reward when reported correctly)'responseType'
:'all'
,'correct'
, or'incorrect'
'rewardOutcome'
:'all'
,'rewarded'
, or'unrewarded'
(this is useful for a 2AUFC task or a task where the reward valve fires probabilistically)'pastStimulus'
:'all'
,'left'
,'right'
, or'zero'
(refers to stimulus side)'pastMovementDir'
:'all'
,'cw'
, or'ccw'
'pastResponseType'
:'all'
,'correct'
, or'incorrect'
'trialsBack'
: integer (used in conjunction with 'past' conditions; default = 0)'switchBlocks'
:'all'}
,'beforeLeft'}
,'beforeRight'
,'afterLeft'
, or'afterRight'
(selects the last (first) 50 trials before (after) a switch to a different reward block)'whichTrials'
: indexing vector or'all'
(lets you select a custom range of trials)'specificRTs'
: 1 x 2 vector ([min max]) or'all'
(selects trials that fall within the range you specify)
By default, running trialConditions = initTrialConditions()
chooses 'all' for each trial condition. Name-value pairs can be concatenated with a comma and can be listed in any order.
Example: trialConditions = initTrialConditions('responseType','correct', 'movementDir','cw')
selects trials where the mouse was correct AND moved the wheel clockwise.
Specifying contrast conditions is called outside of trialConditions
; see below.
Generates a 1 x n vector of trial numbers that pass your conditions.
contrasts
can be called as an integer or a vector, but must include a value that was used in the task. To check which contrasts were used in the task, run the helper function contrasts = getUniqueContrasts(expInfo)
, which generates a vector of all contrasts used in the current session
Example: [~, trialIDs] = selectCondition(expInfo, [-1 1], behavioralData, trialConditions)
generates the trial IDs for all trials with contrast = 1 or -1 AND passing any conditions you specified earlier in trialConditions
A quick way to generate lots of different groups of conditions for use later on
Trials are automatically segregated by contrast, movementDir, responseType, and highRewardSide'. The only extra condition you can specify from the command line is movementTime
('early', 'late', or 'all'). Trial history is not supported.
Trials are organized by single conditions, interacting conditions, and contrast-corrected interacting conditions (the number of trials in each 'bin' is equalized, e.g., same number of correct vs incorrect -100% contrast trials)
One of the early steps in data loading/processing is to determine how a trace of continuous activity (e.g., neural data trace, pupil area, etc.) relates to events of interest (stimulus onset, first wheel movement, valve onset, etc.). We do this by extracting segments of activity and compute their timing with respect to the event (instead of with respect to, say, computer time or Timeline time). We can use these segments to compute the event-triggered average (ETA).
Customarily, event-aligned activity is held in a struct called neuralData.eta.alignedResps
(neurons) or eveData.eta.alignedFace
(video ROIs), and contains four cells:
- activity aligned to stimulus onset (cell {1})
- movement onset (cell {2})
- feedback onset (cell {3})
- go-cue onset (cell {4})
Within each cell is a 3D matrix of size trials x time x unit, where a unit is either a single neuron or a video ROI. time is a fixed time window around the time of the event and expressed in seconds. The length is determined by the time sampling, and the time labels can be found in the 1 x n vector eta.eventWindow
. Timepoint 0 corresponds to the event onset, and will usually be the middle element in the vector (ceil(length/2)).
If you have generated a vector of trial IDs that you are interested in (see 'Indexing trial types', above), you can use this vector to index directly into the first dimension of the 3D matrix (i.e., eta.alignedResps{1}(whichTrials, :, :)
)
Examples:
- To extract the stimulus-aligned activity trace of Cell 40 on trial 10, call
eta.alignedResps{1}(10, :, 40)
- To extract Cell 40's stimulus-aligned activity trace on a subset of trials, call
eta.alignedResps{1}(whichTrials, :, 40)
wherewhichTrials
is a vector of trial IDs - To find the stimulus ETA for Cell 40 across all trials, call
mean(eta.alignedResps{1}(:, :, 40), 1)
- To extract a value for Cell 40's activity precisely at movement onset on trial 10, call
eta.alignedResps{2}(10, eventWindow == 0, 40)
- To extract an activity value from ALL cells precisely at movement onset on trial 10, call
eta.alignedResps{2}(10, eventWindow == 0, :)
Stimulus contrast is expressed on a scale of 0–1, and is signed to denote the screen it appeared on. Left = negative; right = positive.
Animal choices are either -1 ('chose left') or +1 ('chose right').'Chose left' means that the mouse reported a stimulus on the left by turning the wheel CW. 'Chose right' means the mouse reported a stimulus on the right by turning the wheel CCW.
This describes the task's block structure, reporting whether high-volume rewards for correct choices were delivered on the left (-1) or the right (+1).
These are wheel directions, determined from the perspective of the mouse. CW is the wheel action that moves a stimulus to the right; CCW is the wheen action that moves a stimulus to the left. These designations can be used independently of correct/incorrect. CW turns increase the rotary encoder value; CCW turns decrease the rotary encoder value (raw encoder values aren't used much but it's good to bear in mind).
- Green vs orange: This color pair is used to compare blocks of high-value left choices (green) to blocks of high-value right choices (orange).
- Blue vs red: This color pair is used to compare stimulus position or brain hemisphere. Blue means left (or contralateral); red means right (or ipsilateral).
- Green vs brown: This color pair is used to denote correct (green) versus incorrect (brown) trial outcomes.