def setUp(self):
self.a2c = FullAtomModel.Angles2Coords()
self.c2tc = FullAtomModel.Coords2TypedCoords()
self.c2cc = FullAtomModel.CoordsTransform.Coords2CenteredCoords(
rotate=True, translate=True)
self.error = 0.0
self.N = 0
def __init__(self, device='cpu', num_sequences=32, seq_length=350):
self.module = FullAtomModel.Angles2Coords()
self.sequences = [
gen_rand_seq(seq_length) for i in range(num_sequences)
]
# print(self.sequences)
self.angles = torch.randn(len(self.sequences),
7,
len(self.sequences[-1]),
dtype=torch.double,
device='cpu',
requires_grad=True)
def hook(module, input, output):
global g_src
global g_dst
dst = output.grad_fn.Ut_coordinates_dst
src = output.grad_fn.c_coords_input
src = src.resize(int(src.size(1) / 3), 3).cpu().numpy()
dst = dst.resize(int(dst.size(1) / 3), 3).cpu().numpy()
g_src.append(src)
g_dst.append(dst)
if __name__ == '__main__':
#Reading pdb file
p2c = FullAtomModel.PDB2CoordsUnordered()
coords_dst, res_names_dst, atom_names_dst, num_atoms_dst = p2c(
["FullAtomModel/f4TQ1_B.pdb"])
#Making a mask on CA, C, N atoms
is0C = torch.eq(atom_names_dst[:, :, 0], 67).squeeze()
is1A = torch.eq(atom_names_dst[:, :, 1], 65).squeeze()
is20 = torch.eq(atom_names_dst[:, :, 2], 0).squeeze()
is0N = torch.eq(atom_names_dst[:, :, 0], 78).squeeze()
is10 = torch.eq(atom_names_dst[:, :, 1], 0).squeeze()
isCA = is0C * is1A * is20
isC = is0C * is10
isN = is0N * is10
isSelected = torch.ge(isCA + isC + isN, 1)
num_backbone_atoms = int(isSelected.sum())
import torch
from TorchProteinLibrary import ReducedModel, FullAtomModel, RMSD
from torch import optim
import numpy as np
import matplotlib.pylab as plt
import mpl_toolkits.mplot3d.axes3d as p3
from matplotlib import animation
from matplotlib.animation import FuncAnimation
if __name__ == '__main__':
#Reading pdb file
p2c = FullAtomModel.PDB2CoordsUnordered()
coords_dst, chain_names, res_names_dst, res_nums_dst, atom_names_dst, num_atoms_dst = p2c(
["FullAtomModel/f4TQ1_B.pdb"])
#Making a mask on CA, C, N atoms
is0C = torch.eq(atom_names_dst[:, :, 0], 67).squeeze()
is1A = torch.eq(atom_names_dst[:, :, 1], 65).squeeze()
is20 = torch.eq(atom_names_dst[:, :, 2], 0).squeeze()
is0N = torch.eq(atom_names_dst[:, :, 0], 78).squeeze()
is10 = torch.eq(atom_names_dst[:, :, 1], 0).squeeze()
isCA = is0C * is1A * is20
isC = is0C * is10
isN = is0N * is10
isSelected = torch.ge(isCA + isC + isN, 1)
num_backbone_atoms = int(isSelected.sum())
#Resizing coordinates array for convenience
N = int(num_atoms_dst[0].item())
coords_dst.resize_(1, N, 3)
import torch
from TorchProteinLibrary import FullAtomModel
import numpy as np
import matplotlib.pylab as plt
import mpl_toolkits.mplot3d.axes3d as p3
if __name__ == '__main__':
a2c = FullAtomModel.Angles2Coords()
sequences = ['GGMLGWAHFGY']
#Setting conformation to alpha-helix
angles = torch.zeros(len(sequences),
7,
len(sequences[-1]),
dtype=torch.double,
device='cpu')
angles.data[:, 0, :] = -1.047
angles.data[:, 1, :] = -0.698
angles.data[:, 2:, :] = 110.4 * np.pi / 180.0
#Converting angles to coordinates
coords, res_names, atom_names, num_atoms = a2c(angles, sequences)
#Making a mask on CA, C, N atoms
is0C = torch.eq(atom_names[:, :, 0], 67).squeeze()
is1A = torch.eq(atom_names[:, :, 1], 65).squeeze()
is20 = torch.eq(atom_names[:, :, 2], 0).squeeze()
is0N = torch.eq(atom_names[:, :, 0], 78).squeeze()
is10 = torch.eq(atom_names[:, :, 1], 0).squeeze()
isCA = is0C * is1A * is20
isC = is0C * is10
def setUp(self): self.a2c = FullAtomModel.Angles2Coords() self.c2tc = FullAtomModel.Coords2TypedCoords()
def setUp(self):
self.a2c = FullAtomModel.Angles2Coords()
import torch
from TorchProteinLibrary import Volume, FullAtomModel
import _Volume
import numpy as np
import matplotlib.pylab as plt
import mpl_toolkits.mplot3d.axes3d as p3
if __name__ == '__main__':
a2c = FullAtomModel.Angles2Coords()
translate = FullAtomModel.CoordsTranslate()
sequences = ['GGMLGWAHFGY']
#Setting conformation to alpha-helix
angles = torch.zeros(len(sequences),
7,
len(sequences[-1]),
dtype=torch.double,
device='cpu')
angles.data[:, 0, :] = -1.047
angles.data[:, 1, :] = -0.698
angles.data[:, 2:, :] = 110.4 * np.pi / 180.0
#Converting angles to coordinates
coords, res_names, atom_names, num_atoms = a2c(angles, sequences)
#Translating the structure to fit inside the volume
translation = torch.tensor([[60, 60, 60]],
dtype=torch.double,
device='cpu')
coords = translate(coords, translation, num_atoms)
