process_data.m

clear all;


% where data is stored (that you download from figshare)
dataroot = '/media/carsen/DATA2/grive/10krecordings/imgResp/';        
% where processed data and results are saved
matroot  = '/media/carsen/DATA2/grive/10krecordings/stimResults/';
useGPU = 1;

addpath(genpath('.'));

%% SCRIPTS FOR MAIN ANALYSES AND FIGURES

% compute two repeats from sequence of stimuli and response 
% and subtracts spontaneous components
% saves in matroot/stimtype_proc.mat with variables respAll and istimAll
% respAll{k} is stims x neurons x 2 repeats
% istimAll{k} are the stimulus identities for images that were repeated
% twice => imgs(:,:,istimAll{k}) are presented images
compileResps(dataroot, matroot, useGPU);

% compute cross-validated PCs and signal variance and SNR and responsiveness
% saves in matroot/eigs_and_stats_all.mat
statsShuffledPCA(dataroot, matroot);

% treat responses to 32 natural images repeated 96 times as 32*48 images
% repeated twice, and compute the cross-validated PCs
% saves in matroot/natimg32_reps.mat
natimg32PCA(dataroot, matroot, useGPU);

% decode responses to 2800 natural images from one repeat
% decoder correlates responses on first half to second half
% stimulus that is most correlated is the decoded stimulus
% saves in matroot/decoder2800.mat
decode2800(matroot);

% fits natural image responses using low-rank model of RFs
% saves in matroot/lowrank_fits.mat
fitLowRankRFs(dataroot,matroot,useGPU);

% fits gabors to natural image responses
% each cell is a simple + complex cell with fit scaling constants
% saves in matroot/gabor_fits.mat
fitGaborRFs(dataroot,matroot,useGPU);

% computes responses of gabor model to image stimuli fit to mouse that was
% shown all 6 different image sets 
% saves in matroot/gabor_spectrum.mat
simGaborFits(dataroot, matroot, useGPU);

% compute cross-validated PCs for varying numbers of neurons and stimuli
% saves in matroot/eigs_incneurstim_X.mat where X is stimset
incNeurStimPowerLaw(dataroot,matroot);

% simulations of powerlaw with various tuning curves
% saves in matroot/scalefree.mat
fractalvmanifold(matroot)

%% SCRIPTS FOR CONTROLS

% compute cross-validated PCs of z-scored data
% saves in matroot/zscored_spectrum.mat
zscoredControl(matroot,useGPU);

% subtract varying numbers of spontaneous PCs and compute cross-validated
% stimulus PCs
% saves in matroot/spontPC_spectrum.mat
spontPCPowerLaw(dataroot, matroot, useGPU);

% empirical noise spectrum
noiseSpectrum(matroot);

%%% simulations to validate cross-validated PCA method
% simulations with additive, gain, or both noise
% also computes normal PCA to compare cvPCA vs PCA
simPowerLaw_with_gain(matroot);
% simulations with 2P noise + deconvolution
% (will need to download OASIS_matlab repository from github and addpath)
simPowerLaw_deconv(matroot);

% computes power laws of concatenated recordings
% (recordings with similar receptive fields have been paired to increase
% the total number of neurons in a single recording)
concatenateRecordings(matroot);

% compute power law with only 700 stimuli (# of stims in ephys recordings)
% and with 946 neurons (# of neurons in ephys recordings)
incNeurEphysComparison(matroot);