[Q] Inconsistency in bulk solvent scaling fitting?

[sizmailov]

I have a question about recent change in bulk solvent scaling by @pafonine

In the k_mask_grid_search() we find new k_mask and corresponding k_isotropic in resolution shells via grid search.
Next we try to improve r-factor by linear interpolation of k_mask via populate_bin_to_individual_k_mask_linear_interpolation().
Finally we accept k_mask and k_isotropic if they improve r-factor.

The problem is that populate_bin_to_individual_k_mask_linear_interpolation() always use old k_isotropic to compute r-factor. Is there a rationale behind it? Shouldn't it use new k_isotropic instead?

cctbx_project/mmtbx/bulk_solvent/scaler.py

Lines 363 to 392 in a15fcd9

	def k_mask_grid_search(self, r_start):
	if(self.verbose):
	print(" k_mask_grid_search:", file=self.log)
	print(" r_start: %6.4f (r_low: %6.4f)"%(r_start,self._r_low()))
	#k_mask_trial_range = flex.double([i/1000. for i in range(0,650,50)])
	k_mask_trial_range = flex.double([i/1000. for i in range(0,1010,10)])
	k_mask = flex.double(self.f_obs.size(), 0)
	k_mask_bin = flex.double()
	k_isotropic = flex.double(self.f_obs.size(), 0)
	k_total = self.get_k_total()
	for i_cas, cas in enumerate(self.cores_and_selections):
	selection, core, selection_use, sel_work = cas
	f_obs = self.f_obs.select(selection)
	k_total_ = k_total.select(selection)
	k_mask_bin_, k_isotropic_bin_ = \
	bulk_solvent.k_mask_and_k_overall_grid_search(
	f_obs.data()/k_total_,
	core.f_calc.data(),
	core.f_mask().data(),
	k_mask_trial_range,
	selection_use)
	k_mask_bin.append(k_mask_bin_)
	k_mask.set_selected(selection, k_mask_bin_)
	k_isotropic.set_selected(selection, k_isotropic_bin_)
	k_mask_bin_smooth = self.smooth(k_mask_bin)
	k_mask = self.populate_bin_to_individual_k_mask_linear_interpolation(
	k_mask_bin = k_mask_bin_smooth)
	r_try = self.try_scale(k_mask = k_mask, k_isotropic = k_isotropic)

[sizmailov]

Sorry for confusion, now I see. Actual old k_isotropic values are effectively ignored by populate_bin_to_individual_k_mask_linear_interpolation() since trial r-factor calculation is done in resolution bin with adjusted overall scale factor (inside of try_scale()). So, please scratch initial question.

Let me ask another one here about same piece of code. Next two snippets are final parts of k_mask_grid_search and bulk_solvent_scaling(). Both functions optimize k_bulk and update k_isotropic in accordance.

cctbx_project/mmtbx/bulk_solvent/scaler.py

Lines 388 to 392 in a15fcd9

	k_mask_bin_smooth = self.smooth(k_mask_bin)
	k_mask = self.populate_bin_to_individual_k_mask_linear_interpolation(
	k_mask_bin = k_mask_bin_smooth)
	r_try = self.try_scale(k_mask = k_mask, k_isotropic = k_isotropic)

cctbx_project/mmtbx/bulk_solvent/scaler.py

Lines 572 to 577 in a15fcd9

	k_mask_bin_smooth = self.smooth(k_mask_bin)
	k_mask = self.populate_bin_to_individual_k_mask_linear_interpolation(
	k_mask_bin = k_mask_bin_smooth)
	k_isotropic = self._k_isotropic_as_scale_k1(r_start=r_start,k_mask = k_mask)
	r_try = self.try_scale(k_mask = k_mask, k_isotropic = k_isotropic)

Shouldn't they be exactly the same?