Skip to content

fix(mask): normalize Fmask FFT by cell volume to match Fprotein scale #36

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
minhuanli merged 1 commit into from
Jun 30, 2026
Merged

fix(mask): normalize Fmask FFT by cell volume to match Fprotein scale #36

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 5 additions & 3 deletions SFC_Torch/Fmodel.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -754,15 +754,15 @@ def calc_fsolvent(
rs_grid, solvent_percent=solventpct, exponent=exponent,
) # type: ignore
if not self.HKL_array is None:
self.Fmask_HKL = realmask2Fmask(self.real_grid_mask, self.HKL_array)
self.Fmask_HKL = realmask2Fmask(self.real_grid_mask, self.HKL_array, cell_volume=self.unit_cell.volume)
zero_hkl_bool = torch.tensor(self.dHKL <= dmin_nonzero, device=self.device)
self.Fmask_HKL[zero_hkl_bool] = torch.tensor(
0.0, device=self.device, dtype=torch.complex64
)
if Return:
return self.Fmask_HKL
else:
self.Fmask_asu = realmask2Fmask(self.real_grid_mask, self.Hasu_array)
self.Fmask_asu = realmask2Fmask(self.real_grid_mask, self.Hasu_array, cell_volume=self.unit_cell.volume)
zero_hkl_bool = torch.tensor(self.dHasu <= dmin_nonzero, device=self.device)
self.Fmask_asu[zero_hkl_bool] = torch.tensor(
0.0, device=self.device, dtype=torch.complex64
Expand Down Expand Up @@ -1373,7 +1373,9 @@ def calc_fsolvent_batch(
real_grid_mask = rsgrid2realmask(
rs_grid, solvent_percent=solventpct, exponent=exponent, Batch=True
) # type: ignore
Fmask_batch_j = realmask2Fmask(real_grid_mask, HKL_array, end - start)
Fmask_batch_j = realmask2Fmask(
real_grid_mask, HKL_array, cell_volume=self.unit_cell.volume, batchsize=end - start
)
if j == 0:
Fmask_batch = Fmask_batch_j
else:
Expand Down
9 changes: 7 additions & 2 deletions SFC_Torch/mask.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,7 +80,7 @@ def rsgrid2realmask(rs_grid, solvent_percent=0.50, exponent=50.0, Batch=False):
return real_grid_mask


def realmask2Fmask(real_grid_mask, H_array, batchsize=None):
def realmask2Fmask(real_grid_mask, H_array, cell_volume=1.0, batchsize=None):
"""
Convert real space solvent mask grid to mask structural factor, in a fully differentiable
manner
Expand All @@ -93,12 +93,17 @@ def realmask2Fmask(real_grid_mask, H_array, batchsize=None):
H_array: array-like, int
The HKL list we are interested in to assign structural factors

cell_volume: float
Unit cell volume, used to normalize the FFT so that Fmask is on the same
absolute (electron) scale as the atomic structure factors Fprotein.
Fmask(h) = (V/N) * sum_j rho_j * exp(2*pi*i*h.r_j)

Return:
-------
torch.complex64 tensor
Solvent mask structural factor corresponding to the HKL list in H_array
"""
Fmask_grid = torch.fft.ifftn(real_grid_mask, dim=(-3, -2, -1), norm="forward")
Fmask_grid = torch.fft.ifftn(real_grid_mask, dim=(-3, -2, -1)) * cell_volume
tuple_index = tuple(torch.tensor(H_array.T, device=Fmask_grid.device, dtype=int)) # type: ignore
if batchsize is not None:
Fmask = Fmask_grid[(slice(None), *tuple_index)]
Expand Down
Loading