Using structfun with wcoherence function

I am attempting to calculate coherence for waveform data. There are a total of 44 channels. The usual way of processing is to go through a for loop (as shown below)

df = rand(8902,44); % create the array

fs = 5; % sample rate is 5Hz

[wcoh1,wcs1,f1,coi1] = wcoherence(df(:,1),df(:,2),fs);

In order to speed up the processing, I am trying to use structfun and gpuArray to run through all possible iterations. I have been able to calculate the mean using structfun but have been unable to use wcoherence. Below is my attempt

df = rand(8902,44); % create the array

df_gpu = gpuArray(df) % convert to GPU array

fs = 5; % sample rate is 5Hz

combinations = nchoosek(uint16([1:44]),uint16(2)); % gives me all the unique combinations of coherence

[wcoh1,wcs1,f1,coi1] = wcoherence(df_gpu(:,1),df_gpu(:,2),fs);

Does anyone have any recommendations on using combinations to index df_gpu and then compute the coherence?

Thanks for any insight.