File size: 10,805 Bytes
e76b79a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
# Copyright (c) Meta Platforms, Inc. and affiliates.

# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.

from abc import ABC, abstractmethod
from typing import List, Optional

import numpy as np
from biotite.structure import annotate_sse, AtomArray, rmsd, sasa, superimpose

from language.folding_callbacks import FoldingResult
from language.utilities import get_atomarray_in_residue_range


class EnergyTerm(ABC):
    def __init__(self) -> None:
        pass

    @abstractmethod
    def compute(self, node, folding_result: FoldingResult) -> float:
        pass


class MaximizePTM(EnergyTerm):
    def __init__(self) -> None:
        super().__init__()

    def compute(self, node, folding_result: FoldingResult) -> float:
        del node
        return 1.0 - folding_result.ptm


class MaximizePLDDT(EnergyTerm):
    def __init__(self) -> None:
        super().__init__()

    def compute(self, node, folding_result: FoldingResult) -> float:
        del node
        return 1.0 - folding_result.plddt


class SymmetryRing(EnergyTerm):
    def __init__(self, all_to_all_protomer_symmetry: bool = False) -> None:
        super().__init__()
        self.all_to_all_protomer_symmetry: bool = all_to_all_protomer_symmetry

    def compute(self, node, folding_result: FoldingResult) -> float:
        protomer_nodes = node.get_children()
        protomer_residue_ranges = [
            protomer_node.get_residue_index_range() for protomer_node in protomer_nodes
        ]

        centers_of_mass = []
        for start, end in protomer_residue_ranges:
            backbone_coordinates = get_backbone_atoms(
                folding_result.atoms[
                    np.logical_and(
                        folding_result.atoms.res_id >= start,
                        folding_result.atoms.res_id < end,
                    )
                ]
            ).coord
            centers_of_mass.append(get_center_of_mass(backbone_coordinates))
        centers_of_mass = np.vstack(centers_of_mass)

        return (
            float(np.std(pairwise_distances(centers_of_mass)))
            if self.all_to_all_protomer_symmetry
            else float(np.std(adjacent_distances(centers_of_mass)))
        )


def get_backbone_atoms(atoms: AtomArray) -> AtomArray:
    return atoms[
        (atoms.atom_name == "CA") | (atoms.atom_name == "N") | (atoms.atom_name == "C")
    ]


def _is_Nx3(array: np.ndarray) -> bool:
    return len(array.shape) == 2 and array.shape[1] == 3


def get_center_of_mass(coordinates: np.ndarray) -> np.ndarray:
    assert _is_Nx3(coordinates), "Coordinates must be Nx3."
    return coordinates.mean(axis=0).reshape(1, 3)


def pairwise_distances(coordinates: np.ndarray) -> np.ndarray:
    assert _is_Nx3(coordinates), "Coordinates must be Nx3."
    m = coordinates[:, np.newaxis, :] - coordinates[np.newaxis, :, :]
    distance_matrix = np.linalg.norm(m, axis=-1)
    return distance_matrix[np.triu_indices(distance_matrix.shape[0], k=1)]


def adjacent_distances(coordinates: np.ndarray) -> np.ndarray:
    assert _is_Nx3(coordinates), "Coordinates must be Nx3."
    m = coordinates - np.roll(coordinates, shift=1, axis=0)
    return np.linalg.norm(m, axis=-1)


class MinimizeSurfaceHydrophobics(EnergyTerm):
    def __init__(self) -> None:
        super().__init__()

    def compute(self, node, folding_result: FoldingResult) -> float:
        start, end = node.get_residue_index_range()

        return hydrophobic_score(folding_result.atoms, start, end)


_HYDROPHOBICS = {"VAL", "ILE", "LEU", "PHE", "MET", "TRP"}


def hydrophobic_score(
    atom_array: AtomArray,
    start_residue_index: Optional[int] = None,
    end_residue_index: Optional[int] = None,
) -> float:
    """
    Computes ratio of hydrophobic atoms in a biotite AtomArray that are also surface
    exposed. Typically, lower is better.
    """

    hydrophobic_mask = np.array([aa in _HYDROPHOBICS for aa in atom_array.res_name])

    if start_residue_index is None and end_residue_index is None:
        selection_mask = np.ones_like(hydrophobic_mask)
    else:
        start_residue_index = 0 if start_residue_index is None else start_residue_index
        end_residue_index = (
            len(hydrophobic_mask) if end_residue_index is None else end_residue_index
        )
        selection_mask = np.array(
            [
                i >= start_residue_index and i < end_residue_index
                for i in range(len(hydrophobic_mask))
            ]
        )

    # TODO(scandido): Resolve the float/bool thing going on here.
    hydrophobic_surf = np.logical_and(
        selection_mask * hydrophobic_mask, sasa(atom_array)
    )
    # TODO(brianhie): Figure out how to handle divide-by-zero.
    return sum(hydrophobic_surf) / sum(selection_mask * hydrophobic_mask)


class MinimizeSurfaceExposure(EnergyTerm):
    def __init__(self) -> None:
        super().__init__()

    def compute(self, node, folding_result: FoldingResult) -> float:
        start, end = node.get_residue_index_range()

        return surface_ratio(folding_result.atoms, list(range(start, end)))


class MaximizeSurfaceExposure(EnergyTerm):
    def __init__(self) -> None:
        super().__init__()

    def compute(self, node, folding_result: FoldingResult) -> float:
        start, end = node.get_residue_index_range()

        return 1.0 - surface_ratio(folding_result.atoms, list(range(start, end)))


def surface_ratio(atom_array: AtomArray, residue_indices: List[int]) -> float:
    """Computes ratio of atoms in specified ratios which are on the protein surface."""

    residue_mask = np.array([res_id in residue_indices for res_id in atom_array.res_id])
    surface = np.logical_and(residue_mask, sasa(atom_array))
    return sum(surface) / sum(residue_mask)


class MinimizeSurfaceExposure(EnergyTerm):
    def __init__(self) -> None:
        super().__init__()

    def compute(self, node, folding_result: FoldingResult) -> float:
        start, end = node.get_residue_index_range()

        return surface_ratio(folding_result.atoms, list(range(start, end)))


class MaximizeSurfaceExposure(EnergyTerm):
    def __init__(self) -> None:
        super().__init__()

    def compute(self, node, folding_result: FoldingResult) -> float:
        start, end = node.get_residue_index_range()

        return 1.0 - surface_ratio(folding_result.atoms, list(range(start, end)))


def surface_ratio(atom_array: AtomArray, residue_indices: List[int]) -> float:
    """Computes ratio of atoms in specified ratios which are on the protein surface."""

    residue_mask = np.array([res_id in residue_indices for res_id in atom_array.res_id])
    surface = np.logical_and(residue_mask, sasa(atom_array))
    return sum(surface) / sum(residue_mask)


class MaximizeGlobularity(EnergyTerm):
    def __init__(self) -> None:
        super().__init__()

    def compute(self, node, folding_result: FoldingResult) -> float:
        start, end = node.get_residue_index_range()

        backbone = get_backbone_atoms(
            folding_result.atoms[
                np.logical_and(
                    folding_result.atoms.res_id >= start,
                    folding_result.atoms.res_id < end,
                )
            ]
        ).coord

        return float(np.std(distances_to_centroid(backbone)))


def distances_to_centroid(coordinates: np.ndarray) -> np.ndarray:
    """
    Computes the distances from each of the coordinates to the
    centroid of all coordinates.
    """
    assert _is_Nx3(coordinates), "Coordinates must be Nx3."
    center_of_mass = get_center_of_mass(coordinates)
    m = coordinates - center_of_mass
    return np.linalg.norm(m, axis=-1)


class MinimizeCRmsd(EnergyTerm):
    def __init__(self, template: AtomArray, backbone_only: bool = False) -> None:
        super().__init__()

        self.template: AtomArray = template
        self.backbone_only: bool = backbone_only
        if self.backbone_only:
            self.template = get_backbone_atoms(template)

    def compute(self, node, folding_result: FoldingResult) -> float:
        start, end = node.get_residue_index_range()

        atoms = get_atomarray_in_residue_range(folding_result.atoms, start, end)

        if self.backbone_only:
            atoms = get_backbone_atoms(atoms)

        return crmsd(self.template, atoms)


def crmsd(atom_array_a: AtomArray, atom_array_b: AtomArray) -> float:
    # TODO(scandido): Add this back.
    # atom_array_a = canonicalize_within_residue_atom_order(atom_array_a)
    # atom_array_b = canonicalize_within_residue_atom_order(atom_array_b)
    superimposed_atom_array_b_onto_a, _ = superimpose(atom_array_a, atom_array_b)
    return float(rmsd(atom_array_a, superimposed_atom_array_b_onto_a).mean())


class MinimizeDRmsd(EnergyTerm):
    def __init__(self, template: AtomArray, backbone_only: bool = False) -> None:
        super().__init__()

        self.template: AtomArray = template
        self.backbone_only: bool = backbone_only
        if self.backbone_only:
            self.template = get_backbone_atoms(template)

    def compute(self, node, folding_result: FoldingResult) -> float:
        start, end = node.get_residue_index_range()

        atoms = get_atomarray_in_residue_range(folding_result.atoms, start, end)

        if self.backbone_only:
            atoms = get_backbone_atoms(atoms)

        return drmsd(self.template, atoms)


def drmsd(atom_array_a: AtomArray, atom_array_b: AtomArray) -> float:
    # TODO(scandido): Add this back.
    # atom_array_a = canonicalize_within_residue_atom_order(atom_array_a)
    # atom_array_b = canonicalize_within_residue_atom_order(atom_array_b)

    dp = pairwise_distances(atom_array_a.coord)
    dq = pairwise_distances(atom_array_b.coord)

    return float(np.sqrt(((dp - dq) ** 2).mean()))


def pairwise_distances(coordinates: np.ndarray) -> np.ndarray:
    assert _is_Nx3(coordinates), "Coordinates must be Nx3."
    m = coordinates[:, np.newaxis, :] - coordinates[np.newaxis, :, :]
    distance_matrix = np.linalg.norm(m, axis=-1)
    return distance_matrix[np.triu_indices(distance_matrix.shape[0], k=1)]


class MatchSecondaryStructure(EnergyTerm):
    def __init__(self, secondary_structure_element: str) -> None:
        super().__init__()
        self.secondary_structure_element = secondary_structure_element

    def compute(self, node, folding_result: FoldingResult) -> float:
        start, end = node.get_residue_index_range()

        subprotein = folding_result.atoms[
            np.logical_and(
                folding_result.atoms.res_id >= start,
                folding_result.atoms.res_id < end,
            )
        ]
        sse = annotate_sse(subprotein)

        return np.mean(sse != self.secondary_structure_element)