def __getitem__(self, i):
""" Gets alignment pair.
Parameters
----------
i : int
Index of item
Returns
-------
gene : torch.Tensor
Encoded representation of protein of interest
pos : torch.Tensor
Encoded representation of protein that aligns with `gene`.
states : torch.Tensor
Alignment string
alignment_matrix : torch.Tensor
Ground truth alignment matrix
path_matrix : torch.Tensor
Pairwise path distances, where the smallest distance
to the path is computed for every element in the matrix.
"""
gene = self.pairs.iloc[i]['chain1']
pos = self.pairs.iloc[i]['chain2']
st = self.pairs.iloc[i]['alignment']
states = list(map(tmstate_f, st))
if self.clip_ends:
gene, pos, states, st = clip_boundaries(gene, pos, states, st)
if self.pad_ends:
states = [m] + states + [m]
states = torch.Tensor(states).long()
gene = self.tokenizer(str.encode(gene))
pos = self.tokenizer(str.encode(pos))
gene = torch.Tensor(gene).long()
pos = torch.Tensor(pos).long()
alignment_matrix = torch.from_numpy(states2matrix(states))
path_matrix = torch.empty(*alignment_matrix.shape)
g_mask = torch.ones(*alignment_matrix.shape)
if self.construct_paths:
pi = states2edges(states)
path_matrix = torch.from_numpy(path_distance_matrix(pi))
path_matrix = reshape(path_matrix, len(gene), len(pos))
if self.mask_gaps:
g_mask = torch.from_numpy(gap_mask(st)).bool()
alignment_matrix = reshape(alignment_matrix, len(gene), len(pos))
g_mask = reshape(g_mask, len(gene), len(pos))
if not self.return_names:
return gene, pos, states, alignment_matrix, path_matrix, g_mask
else:
gene_name = self.pairs.iloc[i]['chain1_name']
pos_name = self.pairs.iloc[i]['chain2_name']
return (gene, pos, states, alignment_matrix, path_matrix, g_mask,
gene_name, pos_name)
def test_clip_ends_2(self):
gen = 'YACNHCGATAIRNPNWKNHQREH'
oth = 'FHCKSQRVMSDCGSNGSKPFVTNYYVRHQCRKH'
st = np.array([
1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 1, 1, 1, 2, 2, 2, 2, 1, 1, 1, 0, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1
])
a = ''.join(list(map(revstate_f, list(st))))
rx, ry, rs, _ = clip_boundaries(gen, oth, st, a)
self.assertTrue(1) # just make sure it runs
def test_clip_ends_none(self):
from deepblast.constants import m
s_ = [m, m, m, m]
x_ = 'GSSG'
y_ = 'GEIR'
a_ = '::::'
rx, ry, rs, _ = clip_boundaries(x_, y_, s_, a_)
self.assertEqual(x_, rx)
self.assertEqual(y_, ry)
self.assertEqual(s_, rs)
def test_clip_ends(self):
from deepblast.constants import x, m, y
s = [x, m, m, m, y]
x = 'GSSG'
y = 'GEIR'
a = '1:::2'
rx, ry, rs, _ = clip_boundaries(x, y, s, a)
ex, ey, es = 'SSG', 'GEI', [m, m, m]
self.assertEqual(ex, rx)
self.assertEqual(ey, ry)
self.assertEqual(es, rs)