If I would like to use the response function to examine the connectivity between subcortical regions and cortical regions …
Slightly concerned that there is a more fundamental misunderstanding happening here. The response function(s) is/are utilised for deconvolution, but they are assumed to be fixed throughout the entire image (the so-called canonical response function assumption); so there is no way to use the response function(s) to “estimate connectivity”.
If that’s however just a mis-type:
I would like to examine the connectivity between subcortical regions and cortical regions, will the white-matter-only processing be valid?
Depends on what you are “only processing in the white matter”. As far as streamlines tractography reconstruction, especially with ACT, yes one expects the reconstruction to at least be dominated by the empirical signal in the white matter. Alternatively, if you are wanting to define a mask within which to sample quantitative values, then it is entirely up to you as to how to define such a mask, but one would typically expect to make some effort to constrain such to the white matter.
Beyond this, unfortunately your current description is simply too vague to provide more concrete feedback. Perhaps if you could give more details of the specific experiment you are trying to perform?