% Analyze_model.m
%
% Originally, Speron Oct. 7 2008
% Modified/extended by pwjones, 2010-2012.
%
function analyze_model(datadir, file_base)
% --- preliminaries
if ~exist('datadir')
datadir = ['simout' filesep];
else
datadir = [datadir filesep];
end
% use a global variable, so need to clear that
clear global totalTrialN;
if (~exist('file_base'))
% these are filenames for variability simulations looking at looming stimuli and response variability
tags = {'loom_excTconst_snr_5_jit_6_exc_0.54_inh_0.0075'};
tags = {'loom_excGconst_snr_0_jit_6_exc_0.54_inh_0.0075'};
tags = {'loom_snr_10_jit_6_exc_0.54_inh_0.0075'};
else
tags = file_base;
end
% main loop
for t=1:length(tags)
% --- analyze the data ...
analyze_tag(datadir, tags{t});
% ---plot some of the data
if strncmp(tags{t}, 'singlefacet', length('singlefacet'))
plotSingleFacetData(datadir, tags{t});
elseif strncmp(tags{t}, 'loom',4)
if (t==1) fh = []; end;
fh = plotLoomData(datadir, tags{t}, [1 1 0 1 1 1], fh);
elseif strcmp(tags{t}, 'iclamp')
plotIClampData(datadir, tags{t});
end
end
% need to compare the variability for the different SNR values - breaking the flow of the code by putting things here, but this is quickest
%compareLoomResps(datadir, tags);
% ----------------------------------------------------
function analyze_tag(datadir, tag)
% ---------------------------------------------------
% This will analyze a set of data - specified as an entry in TAGS, above
outfile = [datadir tag '_analyzed.mat'];
if (exist(outfile) ~= 0) % skip
disp([outfile ' already exists. No analysis.']);
else
stim_str = tag;
% These are the strings from stimuli that I'm using
stim_str = {[tag '_gsnr_2_jit_6'], [tag '_gsnr_4_jit_6'], [tag '_gsnr_6_jit_6'], ... %suffixes for single facet stimuli
[tag '_gsnr_8_jit_6'], [tag '_gsnr_10_jit_6'], [tag '_gsnr_15_jit_6'], ...
[tag '_lv_10'], [tag '_lv_40'], [tag '_lv_80']}; %suffixes for looming
inj_vals = [-3:0 2:10,12,15,20];% set the current injection values used, in nA
for j = 1:length(inj_vals)
stim_str{1+length(stim_str)} = sprintf('%s_na_%d_', tag, inj_vals(j));
end
% -- single facet
for s=1:6
flist = dir([datadir filesep stim_str{s} '*']); for f=1:length(flist) ; fnames{f} = flist(f).name ; end
if (exist('fnames') & ~isempty(fnames))
tstim = analyze_single_dataset(datadir, fnames, [100 125]);
if (isstruct(tstim)); stim(s) = tstim ; end %#ok<AGROW>
end
fnames = {};
end
% -- Looming stimuli
for s=7:9
flist = dir([datadir filesep stim_str{s} '*']); for f=1:length(flist) ; fnames{f} = flist(f).name ; end
if (exist('fnames') & ~isempty(fnames))
tstim = analyze_single_dataset(datadir, fnames, [1900 2000]);
if (isstruct(tstim)); stim(s) = tstim ; end %#ok<AGROW>
end
fnames = {};
end
% -- Current Injections
for s=10:(9+length(inj_vals))
flist = dir([datadir filesep stim_str{s} '*']); for f=1:length(flist) ; fnames{f} = flist(f).name ; end
if (exist('fnames') & ~isempty(fnames))
tstim = analyze_single_dataset(datadir, fnames, [200 350]);
if (isstruct(tstim)); stim(s) = tstim ; end %#ok<AGROW>
end
fnames = {};
end
% --- save ...
save(outfile, 'stim', 'stim_str', '-v7.3');
end
% --------------------------------------------------------------
function stim = analyze_single_dataset(datadir, outfiles, t_fss)
% ---------------------------------------------------------------
% Analyzes a single type of data - all trials here are treated as the same stimulus
%
% flag for the type of firing rate filtering - gaussian smoothing (0) or ISI based (1).
% Looming stimuli data are processed with gaussian smoothed rates, so use 0 for those.
% For current injections, use the ISI based measures for comparability to those in Peron & Gabbiani (Nature Neurosci 2009)
doISIfilt = 0;
if (strncmp(outfiles, 'iclamp', 6)) % if these are current injection trials, set the analysis method to be ISI based
doISIfilt = 1;
end
if (length(outfiles) == 0) ; stim = -1 ; return ; end
% loop thru files
nfiles = length(outfiles);
for o=1:length(outfiles)
disp(['Processing ' datadir '/' outfiles{o}]);
% --- load data
d = load([datadir '/' outfiles{o}]);
% --- process
spike_idx = get_spikes(-30, d(:,4));
stim.tvec = d(:,1);
if (strncmp(outfiles, 'loom', 4)) % if these are looming trials, adjust the time to be relative to collision
stim.tvec = stim.tvec - (1900-50);
end
spike_vec = zeros(size(stim.tvec)); spike_vec(spike_idx) = 1; %spike vector for convolution
sponti = find(d(:,1) > 20 & d(:,1) < 50);
if (o==1) %set up guassian filters
dt = mean(diff(stim.tvec));
stdg = ceil(20/dt); % 20 ms in number of samples
%stdg = ceil(.5/dt); % short gaussian useful for getting spike threshold of single spikes
filtx = -3*stdg:1:3*stdg;
% this is the gaussian to convolve the spike train with.
filty = my_normpdf(filtx,0,stdg); filty = filty/sum(filty);
% Simon's gaussian - if dt = .1ms, this is a 1 ms SD filter.
gauss = dt*normpdf(-100:dt:100,0,2*dt);
end
if (doISIfilt) % the ISI based method of computing the spike rates
% gaussian convolve ... and normalize so that integral equals number of spks
inst_freq = 1000*get_inst_freq(d(:,1), spike_idx); %uses spike timing for instantaneous freq
gauss_if = conv(inst_freq, gauss);
gauss_if = gauss_if((length(gauss)-1)/2:length(inst_freq) + (length(gauss)-1)/2 -1);
denom = sum(gauss_if);
if (denom == 0) %there are no spikes, and dividing by it would give NaNs
denom=1;
end
gauss_if = gauss_if * (sum(inst_freq)/denom);
gauss_if2 = gauss_if;
else % the binary spike vector/gaussian filter method
gauss_if2 = conv2(spike_vec(:),filty(:), 'same');
%normalize to the number of spikes in the trial as in Gabbiani et al. 1999
if (length(spike_idx) ~= 0) %if no spikes, then divide by 0 gives NaNs for data where 0s should be.
gauss_if2 = gauss_if2 ./(nansum2(gauss_if2)*dt/1000);
gauss_if2 = gauss_if2 * length(spike_idx);
end
end
if (1 == 0) %optional plotting showing the difference between the two spike rate analysis methods
figure; hold on;
plot(stim.tvec, gauss_if, 'k', 'linewidth', 2);
plot(stim.tvec, gauss_if2, 'b', 'linewidth', 2);
end
vm_axon = d(:,2);
vm_prox = d(:,3);
vm_siz = d(:,4);
vm_dist = d(:,5);
ca_siz = d(:,6);
% median filter the dendritic Vm to remove spikes
vm_filt = median_filt(d(:,5), 8, mean(diff(stim.tvec))); %8 msec width median filter
vm_filt_prox = median_filt(d(:,3), 8, mean(diff(stim.tvec)));
vm_rest(o) = mean(vm_filt(sponti));
nsamp = length(vm_prox);
% --- store relevant info
stim.trial(o).conv_inst_freq = gauss_if2(:);
stim.trial(o).spike_times = stim.tvec(spike_idx);
stim.trial(o).spike_vec = spike_vec;
stim.trial(o).vm_filt = vm_filt;
stim.trial(o).vm_filt_prox = vm_filt_prox;
stim.trial(o).vm_prox = d(:,3);
stim.trial(o).vm_dist = d(:,5);
stim.trial(o).vm_prox_std = std(vm_prox(sponti));
stim.trial(o).vm_rest = vm_rest(o);
stim.trial(o).ca_siz = ca_siz;
stim.trial(o).gkca = d(:,7);
stim.trial(o).ina = d(:,8);
nspks(o) = length(spike_idx);
if (~isempty(t_fss))
itemp = find(d(:,1) > t_fss(1), 1, 'first');
if ~isempty(itemp)
i_fss(1) = itemp; else
i_fss(1) = 1;
end
itemp = find(d(:,1) >= t_fss(2), 1, 'first');
if ~isempty(itemp)
i_fss(2) = itemp; else
i_fss(2) = length(gauss_if2);
end;
else
i_fss(1) = 1;
i_fss(2) = length(gauss_if2);
end
[fmax(o), maxi] = max(gauss_if2(1:i_fss(1)));
fss(o) = mean(gauss_if2(i_fss(1):i_fss(2)));
vmss(o) = mean(vm_filt(i_fss(1):i_fss(2)));
stim.trial(o).fmax = fmax(o);
stim.trial(o).fmax_t = stim.tvec(maxi);
stim.trial(o).fss = fss(o);
stim.trial(o).vmss = vmss(o);
stim.trial(o).nspks = nspks(o);
[stim.trial(o).vmpeak, peaki] = nanmax2(vm_filt_prox - mean(vm_prox(sponti)));
stim.trial(o).vmpeak_t = stim.tvec(peaki);
%compute the widths of the responses. Useful for some sim conditions.
stim.trial(o).ifr_width = computePeakWidth(stim.tvec, gauss_if2, stim.trial(o).fmax_t, stim.trial(o).fmax, 1, mean(gauss_if2(1:100)));
stim.trial(o).vmpeak_width = computePeakWidth(stim.tvec, vm_filt, stim.trial(o).vmpeak_t, stim.trial(o).vmpeak, 1, mean(vm_filt(1:100)));
stim.trial(o).ca_ss = mean(ca_siz(i_fss(1):i_fss(2)));
if (o==1) % allocate matrices
a_conv_inst_freq = zeros(nsamp, nfiles); a_vm_filt = zeros(nsamp, nfiles);
a_vm_filt_prox = zeros(nsamp, nfiles); a_vm_prox = zeros(nsamp, nfiles);
a_vm_dist = zeros(nsamp, nfiles); a_vm_axon = zeros(nsamp, nfiles);
a_spike_vec = zeros(nsamp, nfiles); a_ca_siz = zeros(nsamp, nfiles);
a_m = zeros(nsamp, nfiles); a_h = zeros(nsamp, nfiles);
a_n = zeros(nsamp, nfiles);
end
a_conv_inst_freq(:,o) = gauss_if2;
a_vm_filt(:,o) = vm_filt;
a_vm_filt_prox(:,o) = vm_filt_prox;
a_vm_prox(:,o) = stim.trial(o).vm_prox;
a_vm_dist(:,o) = stim.trial(o).vm_dist;
a_vm_axon(:,o) = vm_axon;
a_vm_siz(:,o) = vm_siz;
a_spike_vec(:,o) = stim.trial(o).spike_vec;
a_ca_siz(:,o) = ca_siz;
a_m(:,o) = d(:,9); a_h(:,o) = d(:,10); a_n(:,o) = d(:,11);
end
% get 'average' data
nt = length(outfiles);
stim.num_trials = nt;
stim.mu_nspks = mean(nspks);
stim.sd_nspks = std(nspks);
stim.se_nspks = std(nspks)/sqrt(nt);
stim.mu_fss = mean(fss);
stim.sd_fss = std(fss);
stim.se_fss = std(fss)/sqrt(nt);
stim.mu_fmax = mean(fmax);
stim.sd_fmax = std(fmax);
stim.se_fmax = std(fmax)/sqrt(nt);
stim.snr_fmax = stim.mu_fmax/stim.sd_fmax;
stim.sd_fmax_t = std([stim.trial.fmax_t]);
stim.mu_fmax_t = mean([stim.trial.fmax_t]);
stim.mu_ifr_width = mean([stim.trial.ifr_width]);
stim.sd_ifr_width = std([stim.trial.ifr_width]);
stim.mu_conv_inst_freq = nanmean2(a_conv_inst_freq,2);
stim.sd_conv_inst_freq = nanstd2(a_conv_inst_freq');
stim.se_conv_inst_freq = nanstd2(a_conv_inst_freq')./sqrt(nt);
stim.mu_vmrest = mean(vm_rest);
stim.mu_vmfilt = mean(a_vm_filt,2);
stim.sd_vmfilt = std(a_vm_filt');
stim.mu_vmss = mean(vmss);
stim.sd_vmss = std(vmss);
stim.se_vmss = std(vmss)/sqrt(nt);
stim.mu_vmpeak = mean([stim.trial.vmpeak]);
stim.sd_vmpeak = std([stim.trial.vmpeak]);
stim.snr_vmpeak = stim.mu_vmpeak / stim.sd_vmpeak;
stim.sd_vmpeak_t = std([stim.trial.vmpeak_t]);
stim.mu_vmpeak_t = mean([stim.trial.vmpeak_t]);
stim.mu_vmpeak_width = mean([stim.trial.vmpeak_width]);
stim.sd_vmpeak_width = std([stim.trial.vmpeak_width]);
stim.a_vm_filt = a_vm_filt;
stim.mu_vm_prox_std = mean([stim.trial.vm_prox_std]);
stim.a_ca_siz = a_ca_siz;
stim.a_vm_siz = a_vm_siz;
stim.mu_ca_ss = mean([stim.trial.ca_ss]);
stim.mu_gkca = mean([stim.trial.gkca],2);
stim.mu_ina = mean([stim.trial.ina],2);
stim.a_m = a_m; stim.a_h = a_h; stim.a_n = a_n;
%spikes not in axon are at a slight delay, compensate, then measure spike height
tr = [50 200];
prox_off = round(.29/dt);
%plotting to visualize the model responses per trial type
if ( 1 == 1)
simple = 1; %how many subplots you wanna see
if simple
nr = 2; nc = 3;%rows, columns
else
nr = 4; nc = 3;
end
plotn = 5; % number of individual traces to plot
xl = [min(stim.tvec) max(stim.tvec)]; %limits
figure; subplot(nr,3,1); hold on;
title(outfiles{o});
plotSpikeRasters(gca, stim.tvec, a_spike_vec', 0);
ylim([0 20]); xlim(xl);
%look at how the Vm itself looks
subplot(nr,3,2);
if (~simple)
plot(stim.tvec, a_vm_siz(:,1:plotn), 'Color', 'k'); hold on;
plot(stim.tvec, a_vm_axon(:,1:plotn), 'Color', 'c');
end
plot(stim.tvec, a_vm_prox(:,1:plotn), 'Color', 'r'); hold on;
plot(stim.tvec, a_vm_dist(:,1:plotn), 'Color', 'b');
ylabel('Vm (mV)'); xlim(xl);
%look at the filtered Vm and the firing rate
subplot(nr,3,3);
plot(stim.tvec, a_vm_filt(:,1:plotn), 'Color', [.5 .5 .5]); hold on;
plot(stim.tvec, stim.mu_vmfilt, 'linewidth', 2);
xlabel('time (ms)'); ylabel('Vm (mV)'); xlim(xl);
subplot(nr,3,5); plot(stim.tvec, stim.sd_vmfilt);
xlabel('time (ms)'); ylabel('Vm \sigma (mV)'); xlim(xl);
subplot(nr,3,4); % these will be for the firing rates
plot(stim.tvec, a_conv_inst_freq(:, 1:plotn), 'Color', [.5 .5 .5]); hold on;
plot(stim.tvec, stim.mu_conv_inst_freq, 'linewidth', 2);
xlabel('time (ms)'); ylabel('IFR (Hz)'); xlim(xl);
subplot(nr,3,6); plot(stim.tvec, stim.sd_conv_inst_freq);
xlabel('time (ms)'); ylabel('IFR \sigma (Hz)'); xlim(xl);
if (~simple)
subplot(nr,3,7); plot(stim.tvec, stim.a_ca_siz(:,1:plotn), 'Color', 'g');
xlabel('time (ms)'); ylabel('[Ca]'); xlim(xl);
subplot(nr,3,8); plot(stim.tvec, stim.mu_gkca, 'k-'); ylabel('I_{KCa}'); xlim(xl);
subplot(nr,3,9); plot(stim.tvec, stim.mu_ina, 'k-'); ylabel('I_{Na}'); xlim(xl);
subplot(nr,3,10);
plot(stim.tvec, stim.a_m(:,1:plotn), 'b', stim.tvec, stim.a_h(:,1:plotn), 'r', stim.tvec, stim.a_n(:,1:plotn), 'g')
%graph the [Ca] versus firing rate. This should be approximately linear, but better fit by a square root
%relationship, as in Wang (1998)
stimt = [50 200]; stimi = stim.tvec >= stimt(1) & stim.tvec <= stimt(2);
subplot(nr,3,11);
meanCa = mean(stim.a_ca_siz(stimi,:), 2) .* 1e3;
plot(meanCa, stim.mu_conv_inst_freq(stimi), 'k.'); hold on;
end
end
% ------------------------------------------
function plotSingleFacetData(datadir, tag)
% ------------------------------------------
% This function plots the response characteristics of single facet responses as a function of their
% input parameters.
anfile = [datadir tag '_analyzed.mat'];
load(anfile);
bootn = 5000;
for i=1:length(stim_str)
val = textscan(stim_str{i}, 'singlefacet_gain_%d_gsnr_%d_jit_%d');
try
gain(i) = val{1}; gsnr(i) = val{2}; t_jitter(i) = val{3};
end
end
% Calculate and print the "spontaneous" variability of the more distal membrane potential
sponti = find(stim(1).tvec > 50 & stim(1).tvec < 100);
spont_vm = zeros(length(sponti), length(stim(1).trial), length(stim));
snrfun = @(x)mean(x)./std(x);
for i = 1:length(stim)
temp = [stim(i).trial.vm_dist];
spont_vm(:,:,i) = temp(sponti,:);
ci = bootci(bootn, snrfun, [stim(i).trial.vmpeak]);
snr_ci(i) = stim(i).snr_vmpeak - ci(1); %take the lower diff from the mean - assumes symmetry
ci = bootci(bootn, @std, [stim(i).trial.vmpeak_t]);
sd_vmpeak_t_ci(i) = stim(i).sd_vmpeak_t - ci(1);
end
disp(sprintf('Spontaneous membrane potential: mean = %d , sd = %d', mean(spont_vm(:)), std(spont_vm(:))));
figure;
% plot the single facet response height
subplot(2, 2, 1);
plot(gsnr, [stim.mu_vmpeak], 'k-o');
addErrorBars(gca, gsnr, [stim.mu_vmpeak], [stim.sd_vmpeak], 'k');
xlabel('G_{syn} SNR'); ylabel('Vm_{peak}');
% plot the SNR of the peak response
subplot(2,2,2);
plot(gsnr, [stim.snr_vmpeak], 'k-o');
addErrorBars(gca, gsnr, [stim.snr_vmpeak], snr_ci, 'k');
xlabel('G_{syn} SNR'); ylabel('Vm_{peak} SNR');
subplot(2,2,3);
plot(gsnr, [stim.sd_vmpeak_t], 'k-o');
addErrorBars(gca, gsnr, [stim.sd_vmpeak_t], sd_vmpeak_t_ci, 'k');
xlabel('G_{syn} SNR'); ylabel('Vm_{peak} time \sigma (ms)');
subplot(2,2,4);
plot(gsnr, [stim.mu_vmpeak_width], 'k-o');
addErrorBars(gca, gsnr, [stim.mu_vmpeak_width],[stim.sd_vmpeak_width], 'k');
xlabel('G_{syn} SNR'); ylabel('Vm_{peak} width (ms)');
% -----------------------------------
function plotIClampData(datadir, tag)
%-------------------------------------
% Analyzes the current injection data
% This function plots the response characteristics of single facet responses as a function of their
% input parameters.
anfile = [datadir tag '_analyzed.mat'];
load(anfile);
%dt = .025;
for i=1:length(stim) injstims(i) = ~isempty(stim(i).tvec); end %create logical vect for where data exists.
stim = stim(injstims); stim_str = stim_str(injstims); %just select the subset that are current injections
for i=1:length(stim_str)
val = textscan(stim_str{i}, 'iclamp_na_%f');
try
na(i) = val{1}; %get the current injection values
end
end
% plot the Vm responses to current injections and the input resistance
figure;
subplot(2, 2, 1);
plot(na, [stim.mu_vmpeak]+[stim.mu_vmrest], 'k-o'); hold on;
plot(na, [stim.mu_vmss], 'r-o');
xlabel('Current Injection (nA)'); ylabel('Vm');
legend(gca, {'Peak', 'SS'});
addErrorBars(gca, na, [stim.mu_vmpeak]+[stim.mu_vmrest], [stim.sd_vmpeak], 'k',.25);
addErrorBars(gca, na, [stim.mu_vmss], [stim.sd_vmss], 'r',.25);
vmss = [stim.mu_vmss];
negi = na < 0;
if sum(negi)>=2
fitp = polyfit(na(negi), vmss(negi), 1)
text(min(na), stim(1).mu_vmrest-3, ['R_{in} =' num2str(fitp(1)) ' M\Omega']);
disp(['Input resistance is ' num2str(fitp(1)) ' MOhms']);
plot(na,polyval(fitp, na), 'r--');
end
% plot the firing rates in response to the current injections
subplot(2,2,2); hold on;
plot(na, [stim.mu_fss], 'k-o');
plot(na, [stim.mu_fmax], 'r-x');
xlabel('Current Injection (nA)'); ylabel('SS IFR');
legend(gca, {'Steady State', 'Peak'});
subplot(2,2,3);
plot(na, [stim.sd_vmpeak_t], 'k-o');
xlabel('Current Injection (nA)'); ylabel('Vm_{peak} time \sigma (ms)');
subplot(2,2,4);
plot(stim(1).tvec, [stim.mu_conv_inst_freq]');
%plot the steady state [Ca] versus firing rate - this should be roughly linear
figure; axes; hold on;
plot([stim.mu_ca_ss], [stim.mu_fss], 'ko-');
xlabel('[Ca]'); ylabel('F_{ss} (Hz)');
% Calculate and print the "spontaneous" variability of the more distal membrane potential
sponti = find(stim(1).tvec > 30 & stim(1).tvec < 100);
spont_vm = zeros(length(sponti), length(stim(1).trial), length(stim));
for i = 1:length(stim)
temp = [stim(i).trial.vm_dist];
spont_vm(:,:,i) = temp(sponti,:);
end
disp(sprintf('Spontaneous membrane potential: mean = %d , sd = %d', mean(spont_vm(:)), std(spont_vm(:))));
dvm=.1;
vm_range = [-inf, -70:dvm:-58, inf];
bin_centers = [vm_range(2)-dvm*3/2, (vm_range(2:end-1) + dvm/2), vm_range(end-1)+dvm*3/2];
counts = histc(spont_vm(:), vm_range);
figure; bar(bin_centers, counts);
% ------------------------------------------------
function fh = plotLoomData(datadir, tag, pb, fh)
% -----------------------------------------------
% This function plots the response characteristics of single facet responses as a function of their
% input parameters. pb is a vector containing boolean values about which plots to make.
plot_ind = 0;
anfile = [datadir tag '_analyzed.mat'];
load(anfile);
for i=1:length(stim) loomstims(i) = ~isempty(stim(i).tvec); end %create logical vect for where data exists.
stim = stim(loomstims);
stim_str = stim_str(loomstims); %just select the subset that are the looming
%extract the SNR value for the dataset from the file name
%val = textscan(tag, 'loom_jit_%d');
%val = textscan(tag, 'loom_spont_%d');
%val = textscan(tag, 'loom_snr_%d');
%val = textscan(tag, 'loom_constjit_%d');
%val = textscan(tag, 'loom_spontonly_%d');
%val = textscan(tag, 'loom_noinhvar_snr_%d');
%val=textscan(tag, 'loom_snr_%d');
%val=textscan(tag, 'loom_excTconst_snr_%d');
val = {9}; %assigning an snr of 9?
snr = double(val{1}); %and extracting it
color_scale = (10-snr)/10; %since we know it goes to 10
load([datadir '/loomingParameters.mat']);
dt = mean(diff(stim(1).tvec));
pc = {[0 1 0], [1 0 0], [0.3 0.6 1], [0 0 0]};
time_range = [-1500 150];
% Setting up the axes for plotting
if pb(1) % The Firing Rate plots
if (length(fh)>=1 && fh(1) ~= 0)
figure(fh(1));
mu_ifr_ah = findobj(fh(1), 'tag', 'frah');
mu_stim_ah = findobj(fh(1), 'tag', 'frstimah');
mu_spk_ah = findobj(fh(1), 'tag', 'frspkah');
else
fh(1) = figure; mu_ifr_ah = axes('parent', fh(1), 'Position', [.1 .1 .8 .4], 'tag', 'frah'); hold on; xlim(time_range);
mu_stim_ah = axes('parent', gcf, 'Position', [.1 .8 .8 .15], 'Visible', 'off', 'tag', 'frstimah'); hold on; xlim(time_range);
mu_spk_ah = axes('parent', gcf, 'Position', [.1 .51 .8 .29], 'Visible', 'off', 'tag', 'frspkah'); hold on; xlim(time_range);
end
ylabel(mu_ifr_ah, 'IFR (Hz)'); xlabel(mu_ifr_ah, 'Time (ms)');
end
if pb(2) % Vm plots
if (length(fh)>=2 && fh(2) ~= 0)
figure(fh(2));
mu_vm_ah = findobj(fh(2), 'tag', 'vmah');
sd_vm_ah = findobj(fh(2), 'tag', 'vmsdah');
vm_stim_ah = findobj(fh(2), 'tag', 'stimah');
else
fh(2) = figure;
mu_vm_ah = axes('parent', fh(2), 'Position', [.1 .1 .8 .3], 'tag','vmah'); hold on; xlim(time_range);
sd_vm_ah = axes('parent', fh(2), 'Position', [.1 .45 .8 .25], 'tag','vmsdah'); hold on; xlim(time_range);
vm_stim_ah = axes('parent', fh(2), 'Position', [.1 .75 .8 .2], 'Visible', 'off', 'tag', 'stimah'); hold on; xlim(time_range);
end
ylabel(sd_vm_ah, 'Vm SD (mV)'); xlabel(mu_vm_ah, 'Time (ms)'); set(sd_vm_ah, 'TickDir', 'out');
end
if pb(6) % %Figure for the IFR SD over time
if (length(fh)>=6 && fh(6) ~= 0)
figure(fh(6));
ifr_sd_ah = findobj(fh(6), 'tag', 'ifr_sd_ah');
ifrsd_stim_ah = findobj(fh(6), 'tag', 'ifrsd_stim_ah');
else
fh(6) = figure;
ifr_sd_ah = axes('parent', fh(6), 'Position', [.1 .1 .8 .6], 'tag','ifr_sd_ah'); hold on; xlim(time_range);
ifrsd_stim_ah = axes('parent', fh(6), 'Position', [.1 .75 .8 .2], 'tag','ifrsd_stim_ah'); hold on; xlim(time_range);
end
ylabel(ifr_sd_ah, 'IFR SD (Hz)');
end
global totalTrialN;
if isempty(totalTrialN)
ti = 0;
else
ti = totalTrialN;
end
for i=1:length(stim)
% find the l/v value from the string
val = textscan(stim_str{i}, [tag '_lv_%d']);
stim(i).loverv = double(val{1});
if pb(1) % spiking figure
% plot the mean line
plot(mu_ifr_ah, stim(i).tvec, stim(i).mu_conv_inst_freq, 'Color', pc{i}, 'LineWidth', 2);
errc = (pc{i} + [1 1 1])/2; %define the color for error envelope, halfway between the mean line and white
% plot error envelope
plot_err_poly(mu_ifr_ah, stim(i).tvec, stim(i).mu_conv_inst_freq, stim(i).sd_conv_inst_freq, pc{i}, errc, 1, 10, time_range);
line(loomStimParameters(i).mov_t, abs(loomStimParameters(i).theta), 'Color', pc{i}, 'Linewidth', 2,'Parent',mu_stim_ah);
% Now make and plot the rasters
spk_m = zeros(length(stim(i).trial(1).spike_vec), length(stim(i).trial));
for j=1:length(stim(i).trial) % make the spike matrix
spk_m(:, j) = stim(i).trial(j).spike_vec;
end
rh = plotSpikeRasters(mu_spk_ah, stim(i).tvec, spk_m', ti);
set(mu_spk_ah, 'xlim', time_range); set(rh, 'Color', pc{i});
ti = ti + length(stim(i).trial);
end
if pb(6) %Figure for the IFR SD over time
line(loomStimParameters(i).mov_t, abs(loomStimParameters(i).theta), 'Color', pc{i}, 'Linewidth', 2,'Parent', ifrsd_stim_ah);
line(stim(i).tvec, stim(i).mu_conv_inst_freq(:)./stim(i).sd_conv_inst_freq(:), 'Parent',ifr_sd_ah,'Color', pc{i}, 'LineWidth', 1);
end
if pb(2) % Vm figure
figure(fh(2));
% Now plot the filtered Vm for each condition
plot(mu_vm_ah, stim(i).tvec, stim(i).mu_vmfilt, 'Color', pc{i}, 'LineWidth', 2);
set(mu_vm_ah, 'TickDir', 'out');
ylabel(mu_vm_ah, 'V_m (mV)');
plot(vm_stim_ah, loomStimParameters(i).mov_t, abs(loomStimParameters(i).theta), 'Color', pc{i}, 'Linewidth', 2); hold on;
xlabel('Time (ms)'); ylabel('Filtered Vm (mV)'); set(mu_ifr_ah, 'Tickdir', 'out');
% Also, plot the SD of the filtered Vm in another panel.
plot(sd_vm_ah, stim(i).tvec, stim(i).sd_vmfilt, 'Color', pc{i}, 'LineWidth', 2);
end
if pb(3) % If we want to plot individual trials to check them out
fh(3) = figure;
ifr_ah = axes('parent', gcf, 'position', [.1 .1 .8 .6]); hold on; xlim(time_range);
spk_ah = axes('parent', gcf, 'position', [.1 .72 .8 .2], 'visible', 'off'); xlim(time_range);
for j=1:length(stim(i).trial)
line('Parent', ifr_ah, 'xdata', stim(i).tvec, 'ydata', stim(i).trial(j).conv_inst_freq, 'color', [.5 .5 .5]);
end
plotSpikeRasters(spk_ah, stim(i).tvec, spk_m', 0);
set(spk_ah, 'xlim', time_range); set(ifr_ah, 'xlim', time_range);
else
fh(3) = 0;
end
%let's print a few things out
disp(sprintf('l/v = %d', stim(i).loverv));
disp(sprintf('IFR: mean = %g SD = %g SNRpeak=%g SDpeak_time = %g', ...
stim(i).mu_fmax, stim(i).sd_fmax, stim(i).snr_fmax, stim(i).sd_fmax_t));
disp(sprintf('Vm: mean = %g SD = %g SNRpeak=%g SDpeak_time = %g', ...
stim(i).mu_vmpeak, stim(i).sd_vmpeak, stim(i).snr_vmpeak, stim(i).sd_vmpeak_t));
end
totalTrialN = ti; %set a global variable for the total number of trials.
clear totalTrialN;
if pb(4) % plot the SNR values for the responses
if (length(fh)>=4 && fh(4) ~= 0)
figure(fh(4));
else
fh(4) = figure;
end
subplot(1,2,1); hold on;
black = [0 0 0]; red = [1 0 0];
%bc = ([0 0 0] + [1 1 1]*ce)/12; %just getting lighter lines for lower SNR values
snr_max = 21;
bc = black + ([1 1 1] - black)*(snr_max-snr)/snr_max - [.1 .1 .1];
rc = red + ([1 1 1] - red)*(snr_max-snr)/snr_max - [0 .1 .1];
%rc = ([1 0 0] + [1 1 1]*ce)/12;
plot([stim.loverv], [stim.snr_fmax], 'ko-', 'MarkerSize', 10, 'LineWidth', 2, 'Color', bc);
plot([stim.loverv], [stim.snr_vmpeak], 'r-o', 'MarkerSize', 10, 'LineWidth', 2, 'Color', rc);
legend_str = {'Peak IFR', 'Peak Vm'}; legend(legend_str);
xlabel('l/v (ms)'); ylabel('SNR');
set(gca, 'Xtick', [stim.loverv], 'Xticklabel', cellstr(num2str([stim.loverv]')), 'Tickdir', 'out');
subplot(1,2,2); hold on;
plot([stim.loverv], [stim.sd_fmax_t], '-ok', 'MarkerSize', 10, 'LineWidth', 2, 'Color', bc);
plot([stim.loverv], [stim.sd_vmpeak_t], '-or', 'MarkerSize', 10, 'LineWidth', 2, 'Color', rc);
xlabel('l/v (ms)'); ylabel('Peak Timing \sigma (ms)');
set(gca, 'Xtick', [stim.loverv], 'Xticklabel', cellstr(num2str([stim.loverv]')), 'Tickdir', 'out');
legend(legend_str);
% compute bootstrap confidence intervals
%snrfun = @(x)mean(x)./std(x);
for i = 1:length(stim)
fm = [stim(i).trial.fmax];
vmp = [stim(i).trial.vmpeak];
fm_t = [stim(i).trial.fmax_t];
vmp_t = [stim(i).trial.vmpeak_t];
fm_ci = bootci(5000, @snrfun, fm); %bootstrapped 95% confidence intervals for snr
vmp_ci = bootci(5000, @snrfun, vmp);
fm_t_ci = bootci(5000, @std, fm_t);
vmp_t_ci = bootci(5000, @std, vmp_t);
subplot(1,2,1);
addErrorBars(gca, stim(i).loverv, vmp_ci(1) + diff(vmp_ci)/2, diff(vmp_ci)/2, rc);
addErrorBars(gca, stim(i).loverv, fm_ci(1) + diff(fm_ci)/2, diff(fm_ci)/2, bc);
subplot(1,2,2);
addErrorBars(gca, stim(i).loverv, vmp_t_ci(1) + diff(vmp_t_ci)/2, diff(vmp_t_ci)/2, rc);
addErrorBars(gca, stim(i).loverv, fm_t_ci(1) + diff(fm_t_ci)/2, diff(fm_t_ci)/2, bc);
%save these values back to the structure
stim(i).snr_fmax_ci = fm_ci;
stim(i).snr_vmpeak_ci = vmp_ci;
stim(i).sd_fmax_t_ci = fm_t_ci;
stim(i).sd_vmpeak_t_ci = vmp_t_ci;
end
end
if pb(5)
% Also, in order to check that responses to looming are like those that
% we see in vivo, would like to check the timing of the peak, and whether
% it is linear.
fh(5) = figure;
for i=1:2
subplot(1,2,i);
if i==1
plot([stim.loverv], -([stim.mu_vmpeak_t]), 'o-b', 'linewidth', 2, 'MarkerSize', 10); hold on;
addErrorBars(gca, [stim.loverv], -([stim.mu_vmpeak_t]), [stim.sd_vmpeak_t], 'b');
title('Vm Peak');
%plot the individual peak times
x = zeros(max([stim.num_trials]), length(stim)); y = zeros(max([stim.num_trials]), length(stim));
% for j=1:length(stim)
% x(:,j) = stim(j).loverv*ones(stim(j).num_trials,1);
% y(:,j) = -[stim(j).trial.vmpeak_t]';
% end
% plot(x, y, 'k.');
[fit_p, ~, ~, ci] = linear_fit_stats(x(:), y(:))
plot([stim.loverv], polyval(fit_p, [stim.loverv]), 'b--', 'LineWidth', 1);
else
plot([stim.loverv], -([stim.mu_fmax_t]), 'o-b', 'linewidth', 2, 'MarkerSize', 10); hold on;
addErrorBars(gca, [stim.loverv], -([stim.mu_fmax_t]), [stim.sd_fmax_t], 'b');
x = zeros(max([stim.num_trials]), length(stim)); y = zeros(max([stim.num_trials]), length(stim));
for j=1:length(stim)
x(:,j) = stim(j).loverv*ones(stim(j).num_trials,1);
y(:,j) = -[stim(j).trial.fmax_t]';
end
%plot(x, y, 'b.');
[fit_p, ~, ~, ci] = linear_fit_stats(x(:), y(:))
plot([stim.loverv], polyval(fit_p, [stim.loverv]), 'b--', 'LineWidth', 1);
fitstr=sprintf('fit: slope = %.2g, int= %.2f, thresh= %.2f', fit_p(1), fit_p(2), 2*atand(1/fit_p(1)));
title(fitstr);
end
xlabel('l/v (ms)'); ylabel('Peak Time before Collision (ms)');
hold on;
plot(10:10:80, polyval([4.39 -18], 10:10:80, 1), 'r'); %My looming data
%plot(10:10:80, polyval([2.9 26], 10:10:80, 1), 'r'); %haleh DCMD peak (Fotowat and Gabbiani 2007)
%plot(10:10:80, polyval([4.7 -27], 10:10:80, 1), 'r'); %Fabrizio DCMD peak (Gabbiani et al 1999)
end
end
% Let's save a summary that will be quickly loaded.
save(anfile, 'stim', 'stim_str', '-v7.3');
function snr = snrfun(vals)
meanval = mean(vals);
sdval = std(vals);
if (sdval == 0)
snr = 0;
else
snr = meanval./sdval;
end
% --------------------------------------------
function spike_idx = get_spikes(thresh, data)
% --------------------------------------------
% Extracts spike idx, based on threshold + peak method
%
greater = find(data > thresh);
diffs = diff(data);
greater = greater(find(greater < length(diffs)));
spike_idx = [];
greater = greater(find(greater < length(diffs)));
for g=1:length(greater)-1
if (diffs(greater(g)) > 0 & diffs(greater(g)+1) < 0)
spike_idx(length(spike_idx)+1) = greater(g)+1;
end
end
% --------------------------------------------------------------
function [inst_freq_vals] = get_inst_freq(time_vals, spike_idx)
% --------------------------------------------------------------
% Computes instantaneous firing frequency and returns it along with SD
%
% The heart of the matter
inst_freq_vals=zeros(1,length(time_vals));
spike_times = time_vals(spike_idx);
if (length(spike_times) > 1)
for i=1:length(inst_freq_vals)
pre_spike = find(spike_times < time_vals(i));
pre_spike = max(pre_spike);
post_spike = find(spike_times > time_vals(i));
post_spike = min(post_spike);
% Are we AT a spike?
if (length(find(spike_times == time_vals(i))) == 1)
if (length(pre_spike) > 0 & length(post_spike) > 0)
inst_freq_vals(i) = 0.5/(time_vals(i)-spike_times(pre_spike)) ...
+ 0.5/(spike_times(post_spike)-time_vals(i));
% At first spike?
elseif (length(pre_spike) == 0)
inst_freq_vals(i) = 1/(spike_times(post_spike)-time_vals(i));
% At last spike?
elseif (length(post_spike) == 0)
inst_freq_vals(i) = 1/(time_vals(i)-spike_times(pre_spike));
end
% Not at a spike
else
% Before first spike or after last?
if (length(pre_spike) == 0 | length(post_spike) == 0)
inst_freq_vals(i) = 0;
else
inst_freq_vals(i) = 1/(spike_times(post_spike)-spike_times(pre_spike));
end
end
end
end