Compare Neural Activity Across States¶

This tool uses 1.0 compute credits per hour.

Overview¶

The Compare Neural Activity Across States tool during states tool calculates the average modulation during user specified states. Specifically, this tool calculates the modulation of neurons during each state and validates modulation by permutation testing. It then uses the significance of the modulation scores to identify whether each neuron is up or down regulated during these states.

Input Data¶

This tool accepts either a single state, or two opposing states. If only one state is specified, the function will compare activity levels against a placeholder state of random periods of the recording not during the specified state the same length as the specified state. If two states are specified, the function will compare the activity levels of the two states against each other.

Valid Inputs¶

Source Parameter	File Type	File Format
Cell Set File(s)	cell_set	isxd
Annotations File	experiment_annotations	parquet

Generating synchronized annotation files

You can generate the synchronized annotation file by running one of these synchronization tools:

Parameters¶

Parameter	Required?	Default	Description
Cell Set File(s)	True	N/A	ISXD cell set file(s)
Annotations File	True	N/A	Experiment Annotations file
State A	True	N/A	Name of first state to analyze. This state should exist in the annotations file.
State B	False	N/A	Name of second state to compare against (optional). If blank, comparison will be made to times not in state A
Column Name	True	state	Column name in annotations file with states
Number of Shuffles	False	1000	Number of shuffles to determine if a neuron is significantly modulated by state
Significance Threshold	False	0.01	Significance threshold (alpha) below which p-values are considered significant.

Parameter	Required?	Description
Input files: Cell Set File	Yes	Cell set file `.isxd` to analyze, or series of cell set files.
Input files: Annotations File	Yes	corresponding synchronized `annotations.parquet` file generated by previous synchronization analysis
States: State A	Yes	Name of the first state to compare
States: State B	No	Name of the second state, if any to compare against the first state
Parameters: Number of shuffles	No	Value that defines the number of shuffled to perform during the permutation testing. Default value is 1000
Parameters: Column Name	No	Name of the column where the state data is. Default value is "state" from the synchronization tool

Algorithm details¶

graph TD A[Run population activity workflow] --> B[calculate modulation for each cell in each state]; A[Run population activity workflow] --> C[Plot cell traces]; B[calculate modulation for each cell in each state] --> D[Calculate significance of modulation by permutation testing]; B[calculate modulation for each cell in each state] --> E[Plot comparison of modulation of states against each other for each individual cell]; D[Calculate significance of modulation by permutation testing] --> F[Classify neurons as up/down/non-regulated by significance value]; E[Plot comparison of modulation of states against each other for each individual cell] --> G[Plot comparison of mean neuron response per state]; F[Classify neurons as up/down/non-regulated by significance value] --> H[Plot Cell Maps with modulated neurons colored]

The tool uses several metrics to measure and analyze population activity, described in the following sections

Average measures¶

When the activity of a cell is measured, the mean of the calcium trace is used.

Modulation score¶

The modulation score is calculated using a difference/sum formula. Specifically

\[ \frac{ \text{mean trace activity during state} - \text{mean trace activity not during state} }{ \text{mean traceactivity during state} + \text{mean trace activity not during state} } \]

Significance level¶

The significance level is determined using circular permutation, where the state labels are shuffled and the modulation score is recalculated. The p-value is calculated as the minimum between the proportion shuffeled scores greater than the original score and the proportion of shuffled scores less than the original score.

This means that the specificity of the p-value is determined by the number of shuffles performed. (e.g. 1000 shuffles will give a p-value with a resolution of 0.001)

Outputs¶

The tool will return a .csv file with individual cells as rows, and the columns for each state analyzed. If only one state is provided, then state 2 will be a placeholder state for "not during state". The columns will contain the average activity of each cell in each state, the modulation score, p-value, and modulation identity for each cell in each state.

An example output file looks like this:

name	State 1 raw modulation score	State 1 p-value	State 1 modulation identity	State 1 mean activity	State 2 (or placeholder) raw modulation score	State 2 p-value	State 2 mean activity
C000	-0.057537501	0.256	0	8.013278525	0.055948157	0.204	10.88235528
C001	0.321526855	0.006	1	14.86207092	0.140212216	0.058	8.404072915
C002	-0.218330082	0.002	-1	-3.856631176	-0.020360241	0.24	0.290256447

Previews¶

The tool will also generate two previews of the output data:

Cell Maps¶

This preview will display two panels. Each panel will show cell footprints from the cell set, and highlight footprints of up- or down-regulated cells during each state.

Neuron traces¶

This preview consists of three panels. The top panel displays the traces of all neurons (gray). The two states are shown in red and black. The bottom left panel shows a scatter plot of neuron activity in the two states. Each dot is a single neuron. The bottom right panel shows a bar plot of average neuron response in each state.