| 1 |
""" |
| 2 |
Rosetta fragment libraries. |
| 3 |
|
| 4 |
This module defines the L{RosettaFragmentMap} objects, which describes a |
| 5 |
fragment library in Rosetta NNmake format. L{RosettaFragmentMap} has a |
| 6 |
static factory method for building a library from a fragment file: |
| 7 |
|
| 8 |
>>> RosettaFragmentMap.read('fragments.txt') |
| 9 |
<RosettaFragmentMap> |
| 10 |
|
| 11 |
@note: Consider extracting L{RosettaFragmentMap.read} as a Rosetta |
| 12 |
fragment parser which naturally belongs to csb.bio.io. |
| 13 |
""" |
| 14 |
|
| 15 |
from csb.bio.structure import TorsionAnglesCollection, TorsionAngles |
| 16 |
from csb.core import AbstractContainer |
| 17 |
|
| 18 |
|
| 19 |
class ResidueInfo(object): |
| 20 |
""" |
| 21 |
Container struct for a single rosetta fragment residue. |
| 22 |
|
| 23 |
@param rank: residue position (in the source chain, 1-based) |
| 24 |
@type rank: int |
| 25 |
@param aa: amino acid |
| 26 |
@type aa: str |
| 27 |
@param ss: secondary structure |
| 28 |
@type ss: str |
| 29 |
@param torsion: torsion angles |
| 30 |
@type torsion: L{csb.bio.structure.TorsionAngles} |
| 31 |
""" |
| 32 |
|
| 33 |
def __init__(self, rank, aa, ss, torsion, calpha=[]): |
| 34 |
|
| 35 |
self.rank = rank |
| 36 |
self.aa = aa |
| 37 |
self.ss = ss |
| 38 |
self.torsion = torsion |
| 39 |
self.calpha = tuple(calpha) |
| 40 |
|
| 41 |
@property |
| 42 |
def phi(self): |
| 43 |
return self.torsion.phi or 0. |
| 44 |
|
| 45 |
@property |
| 46 |
def psi(self): |
| 47 |
return self.torsion.psi or 0. |
| 48 |
|
| 49 |
@property |
| 50 |
def omega(self): |
| 51 |
return self.torsion.omega or 0. |
| 52 |
|
| 53 |
def copy(self): |
| 54 |
""" |
| 55 |
@return: a deep copy of the struct |
| 56 |
@rtype: L{ResidueInfo} |
| 57 |
""" |
| 58 |
return ResidueInfo(self.rank, self.aa, self.ss, self.torsion.copy(), self.calpha) |
| 59 |
|
| 60 |
class RosettaFragment(object): |
| 61 |
""" |
| 62 |
Represents a single Rosetta fragment match. |
| 63 |
|
| 64 |
@param source_id: entry ID of the source PDB chain (in accnC format) |
| 65 |
@type source_id: str |
| 66 |
@param qstart: start position in target (rank) |
| 67 |
@type qstart: int |
| 68 |
@param qend: end position in target (rank) |
| 69 |
@type qend: int |
| 70 |
@param start: start position in C{source} (rank) |
| 71 |
@type start: int |
| 72 |
@param end: end position in C{source} (rank) |
| 73 |
@type end: int |
| 74 |
@param score: score of the fragment |
| 75 |
@type score: float |
| 76 |
@param residues: fragment residue structs |
| 77 |
@type residues: iterable of L{ResidueInfo} |
| 78 |
""" |
| 79 |
|
| 80 |
def __init__(self, source_id, qstart, qend, start, end, score, residues): |
| 81 |
|
| 82 |
if not (qend - qstart + 1) == (end - start + 1) == len(residues): |
| 83 |
raise ValueError() |
| 84 |
if not len(source_id) == 5: |
| 85 |
raise ValueError(source_id) |
| 86 |
|
| 87 |
self._source_id = str(source_id) |
| 88 |
self._qstart = int(qstart) |
| 89 |
self._qend = int(qend) |
| 90 |
self._start = int(start) |
| 91 |
self._end = int(end) |
| 92 |
self._score = float(score) |
| 93 |
self._residues = list(residues) |
| 94 |
|
| 95 |
def subregion(self, qstart, qend): |
| 96 |
""" |
| 97 |
Extract a subregion from the fragment. |
| 98 |
|
| 99 |
@param qstart: start position in target |
| 100 |
@type qstart: int |
| 101 |
@param qend: end position in target |
| 102 |
@type qend: int |
| 103 |
|
| 104 |
@return: a new fragment (deep copy) |
| 105 |
@rtype: L{RosettaFragment} |
| 106 |
""" |
| 107 |
|
| 108 |
if not self.qstart <= qstart <= qend <= self.qend: |
| 109 |
raise ValueError('Invalid subregion') |
| 110 |
|
| 111 |
start = qstart - self.qstart + self.start |
| 112 |
end = qend - self.qend + self.end |
| 113 |
|
| 114 |
diff = qstart - self.qstart |
| 115 |
size = qend - qstart + 1 |
| 116 |
assert 0 <= diff |
| 117 |
|
| 118 |
residues = [ r.copy() for r in self.residues[diff : diff + size] ] |
| 119 |
assert len(residues) == size |
| 120 |
|
| 121 |
return RosettaFragment(self.source_id, qstart, qend, start, end, self.score, residues) |
| 122 |
|
| 123 |
def __lt__(self, other): |
| 124 |
# lower score means a better fragment |
| 125 |
return self.score < other.score |
| 126 |
|
| 127 |
def __iter__(self): |
| 128 |
return iter(self._residues) |
| 129 |
|
| 130 |
def __len__(self): |
| 131 |
return len(self._residues) |
| 132 |
|
| 133 |
def __str__(self): |
| 134 |
|
| 135 |
out = [] |
| 136 |
|
| 137 |
for residue in self.residues: |
| 138 |
line = ' {0.accession:4} {0.chain:1} {1.rank:>5} {1.aa:1} {1.ss:1} {1.phi:>8.3f} {1.psi:>8.3f} {1.omega:>8.3f} {0.score:>8.3f}' |
| 139 |
out.append(line.format(self, residue)) |
| 140 |
|
| 141 |
return '\n'.join(out) |
| 142 |
|
| 143 |
@staticmethod |
| 144 |
def from_object(assignment): |
| 145 |
""" |
| 146 |
Factory method: build a rosetta fragment from an assignment object. |
| 147 |
|
| 148 |
@param assignment: source assignment |
| 149 |
@type assignment: L{Assignment} |
| 150 |
|
| 151 |
@rtype: L{RosettaFragment} |
| 152 |
""" |
| 153 |
residues = [] |
| 154 |
a = assignment |
| 155 |
|
| 156 |
for rank, aa, torsion, calpha in zip(range(a.start, a.end + 1), a.sequence, a.torsion, a.backbone): |
| 157 |
residues.append(ResidueInfo(rank, aa, 'L', torsion, calpha)) |
| 158 |
|
| 159 |
return RosettaFragment(a.source_id, a.qstart, a.qend, a.start, a.end, 1 - (a.probability or 0.0), residues) |
| 160 |
|
| 161 |
@property |
| 162 |
def length(self): |
| 163 |
return len(self) |
| 164 |
|
| 165 |
@property |
| 166 |
def source_id(self): |
| 167 |
return self._source_id |
| 168 |
|
| 169 |
@property |
| 170 |
def accession(self): |
| 171 |
return self.source_id[:4] |
| 172 |
|
| 173 |
@property |
| 174 |
def chain(self): |
| 175 |
return self.source_id[4:] |
| 176 |
|
| 177 |
@property |
| 178 |
def id(self): |
| 179 |
return '{0.source_id}:{0.start}-{0.end}'.format(self) |
| 180 |
|
| 181 |
@property |
| 182 |
def qstart(self): |
| 183 |
return self._qstart |
| 184 |
|
| 185 |
@property |
| 186 |
def qend(self): |
| 187 |
return self._qend |
| 188 |
|
| 189 |
@property |
| 190 |
def start(self): |
| 191 |
return self._start |
| 192 |
|
| 193 |
@property |
| 194 |
def end(self): |
| 195 |
return self._end |
| 196 |
|
| 197 |
@property |
| 198 |
def score(self): |
| 199 |
return self._score |
| 200 |
|
| 201 |
@property |
| 202 |
def residues(self): |
| 203 |
return tuple(self._residues) |
| 204 |
|
| 205 |
@property |
| 206 |
def torsion(self): |
| 207 |
return TorsionAnglesCollection([r.torsion for r in self._residues], start=0) |
| 208 |
|
| 209 |
class OutputBuilder(object): |
| 210 |
""" |
| 211 |
Rosetta fragment file formatter. |
| 212 |
|
| 213 |
@param output: destination stream |
| 214 |
@type output: file |
| 215 |
""" |
| 216 |
|
| 217 |
def __init__(self, output): |
| 218 |
self._out = output |
| 219 |
|
| 220 |
@property |
| 221 |
def output(self): |
| 222 |
return self._out |
| 223 |
|
| 224 |
def add_position(self, qstart, frags): |
| 225 |
""" |
| 226 |
Write a new assignment origin. |
| 227 |
|
| 228 |
@param qstart: target position |
| 229 |
@type qstart: float |
| 230 |
@param frags: number of fragments, starting at that position |
| 231 |
@type frags: int |
| 232 |
""" |
| 233 |
self.output.write(' position: {0:>12} neighbors: {1:>12}\n\n'.format(qstart, len(frags))) |
| 234 |
|
| 235 |
def add_fragment(self, fragment): |
| 236 |
""" |
| 237 |
Write a new fragment. |
| 238 |
@type fragment: L{RosettaFragment} |
| 239 |
""" |
| 240 |
for residue in fragment.residues: |
| 241 |
self.add_residue(fragment, residue) |
| 242 |
self.output.write('\n') |
| 243 |
|
| 244 |
self.output.write('\n') |
| 245 |
|
| 246 |
def add_residue(self, fragment, residue): |
| 247 |
""" |
| 248 |
Write a new fragment residue. |
| 249 |
@type fragment: L{RosettaFragment} |
| 250 |
@type residue: L{ResidueInfo} |
| 251 |
""" |
| 252 |
line = ' {0.accession:4} {0.chain:1} {1.rank:>5} {1.aa:1} {1.ss:1} {1.phi:>8.3f} {1.psi:>8.3f} {1.omega:>8.3f} {0.score:>8.3f}' |
| 253 |
self.output.write(line.format(fragment, residue)) |
| 254 |
|
| 255 |
class ExtendedOutputBuilder(OutputBuilder): |
| 256 |
""" |
| 257 |
Builds non-standard fragment files, which contain the CA coordinates of |
| 258 |
each residue at the end of each line. |
| 259 |
""" |
| 260 |
|
| 261 |
def add_residue(self, fragment, residue): |
| 262 |
|
| 263 |
super(ExtendedOutputBuilder, self).add_residue(fragment, residue) |
| 264 |
|
| 265 |
if residue.calpha: |
| 266 |
calpha = residue.calpha |
| 267 |
else: |
| 268 |
calpha = [0, 0, 0] |
| 269 |
|
| 270 |
self.output.write(' {0:>7.3f} {1:>7.3f} {2:>7.3f}'.format(*calpha)) |
| 271 |
|
| 272 |
class RosettaFragmentMap(AbstractContainer): |
| 273 |
""" |
| 274 |
Represents a Rosetta fragment library. |
| 275 |
|
| 276 |
@param fragments: library fragments |
| 277 |
@type fragments: iterable of L{RosettaFragment} |
| 278 |
@param length: target sequence's length. If not defined, the qend of the |
| 279 |
last fragment will be used instead. |
| 280 |
@type length: int |
| 281 |
""" |
| 282 |
|
| 283 |
def __init__(self, fragments, length=None): |
| 284 |
|
| 285 |
self._fragments = [] |
| 286 |
|
| 287 |
self._unconf = set() |
| 288 |
self._sources = set() |
| 289 |
self._starts = set() |
| 290 |
self._ends = set() |
| 291 |
self._length = None |
| 292 |
|
| 293 |
for f in fragments: |
| 294 |
self.append(f) |
| 295 |
|
| 296 |
if length is not None: |
| 297 |
assert length >= self._maxend |
| 298 |
self._length = int(length) |
| 299 |
else: |
| 300 |
self._length = self._maxend |
| 301 |
|
| 302 |
@property |
| 303 |
def _maxend(self): |
| 304 |
return max(self._ends or [0]) |
| 305 |
|
| 306 |
def append(self, fragment): |
| 307 |
""" |
| 308 |
Append a new L{RosettaFragment} |
| 309 |
""" |
| 310 |
|
| 311 |
if self._length and fragment.qend > self._length: |
| 312 |
raise ValueError('fragment out of range') |
| 313 |
|
| 314 |
self._fragments.append(fragment) |
| 315 |
self._sources.add(fragment.accession) |
| 316 |
self._starts.add(fragment.qstart) |
| 317 |
self._ends.add(fragment.qend) |
| 318 |
|
| 319 |
def __len__(self): |
| 320 |
return len(self._fragments) |
| 321 |
|
| 322 |
@property |
| 323 |
def _children(self): |
| 324 |
return self._fragments |
| 325 |
|
| 326 |
@property |
| 327 |
def unconfident_positions(self): |
| 328 |
return tuple(sorted(self._unconf)) |
| 329 |
|
| 330 |
@property |
| 331 |
def size(self): |
| 332 |
return len(self) |
| 333 |
|
| 334 |
@property |
| 335 |
def sources(self): |
| 336 |
return tuple(self._sources) |
| 337 |
|
| 338 |
@property |
| 339 |
def start_positions(self): |
| 340 |
return tuple(sorted(self._starts)) |
| 341 |
|
| 342 |
def fromsource(self, accession): |
| 343 |
""" |
| 344 |
@return: a tuple of all fragments, extracted from the specified C{source}. |
| 345 |
|
| 346 |
@param accession: source entry ID |
| 347 |
@type accession: str |
| 348 |
""" |
| 349 |
return tuple(f for f in self._fragments if f.accession == accession) |
| 350 |
|
| 351 |
def starting_at(self, qrank): |
| 352 |
""" |
| 353 |
@return: a tuple of all fragments, starting at the specified target position. |
| 354 |
|
| 355 |
@param qrank: fragment origin (in target, rank) |
| 356 |
@type qrank: int |
| 357 |
""" |
| 358 |
return tuple(f for f in self._fragments if f.qstart == qrank) |
| 359 |
|
| 360 |
def at(self, qrank): |
| 361 |
""" |
| 362 |
@return: a tuple of all fragments, covering the specified position. |
| 363 |
|
| 364 |
@param qrank: position in target, rank |
| 365 |
@type qrank: int |
| 366 |
""" |
| 367 |
return tuple(f for f in self._fragments if f.qstart <= qrank <= f.qend) |
| 368 |
|
| 369 |
def mark_unconfident(self, rank): |
| 370 |
""" |
| 371 |
Mark the specified position in the target as a low-confidence one. |
| 372 |
|
| 373 |
@param rank: position in target |
| 374 |
@type rank: int |
| 375 |
""" |
| 376 |
if not 1 <= rank <= self._length: |
| 377 |
raise ValueError(rank) |
| 378 |
|
| 379 |
self._unconf.add(rank) |
| 380 |
|
| 381 |
def complement(self, fragment): |
| 382 |
""" |
| 383 |
Append C{fragment} to the library, if the fragment is anchored |
| 384 |
around a low-confidence position. |
| 385 |
|
| 386 |
@type fragment: L{RosettaFragment} |
| 387 |
""" |
| 388 |
if not self._unconf: |
| 389 |
raise ValueError('no unconfident regions to complement') |
| 390 |
|
| 391 |
f = fragment |
| 392 |
for rank in self._unconf: |
| 393 |
if f.qstart < rank < f.qend: |
| 394 |
if (rank - f.qstart + 1) > 0.4 * (f.qend - f.qstart + 1): |
| 395 |
self.append(f) |
| 396 |
break |
| 397 |
|
| 398 |
def sort(self, field='score', reverse=False): |
| 399 |
""" |
| 400 |
Sort the fragments in the library. |
| 401 |
""" |
| 402 |
|
| 403 |
self._fragments.sort(key=lambda i:getattr(i, field), reverse=reverse) |
| 404 |
|
| 405 |
def dump(self, file, builder=OutputBuilder): |
| 406 |
""" |
| 407 |
Write the library to a Rosetta fragment file. |
| 408 |
|
| 409 |
@param file: destination file name |
| 410 |
@type file: str |
| 411 |
""" |
| 412 |
|
| 413 |
with open(file, 'w') as out: |
| 414 |
builder = builder(out) |
| 415 |
|
| 416 |
for qstart in self.start_positions: |
| 417 |
|
| 418 |
frags = self.starting_at(qstart) |
| 419 |
builder.add_position(qstart, frags) |
| 420 |
|
| 421 |
for fragment in frags: |
| 422 |
builder.add_fragment(fragment) |
| 423 |
|
| 424 |
@staticmethod |
| 425 |
def read(file, top=None): |
| 426 |
""" |
| 427 |
Read a standard fragment file. |
| 428 |
|
| 429 |
@param file: file name |
| 430 |
@type file: str |
| 431 |
@param top: if defined, read only C{top} fragments per start position |
| 432 |
(default=all) |
| 433 |
@type top: int or None |
| 434 |
|
| 435 |
@return: parsed fragment library |
| 436 |
@rtype: L{RosettaFragmentMap} |
| 437 |
""" |
| 438 |
# This is the format (rosetta_fragments/nnmake/makeoutput.f): |
| 439 |
# source chain rank residue ss phi psi omega score dme dme_f best_nn_ss_type dipolar+noe 'P' position 'F' fragment# |
| 440 |
|
| 441 |
def ang(a): |
| 442 |
a = float(a) |
| 443 |
if a < -180: return 360 + a |
| 444 |
elif a > 180: return -360 + a |
| 445 |
else: return a |
| 446 |
|
| 447 |
frags = [] |
| 448 |
|
| 449 |
qstart, qend, start, end = 0, 0, 0, 0 |
| 450 |
id = '' |
| 451 |
score = None |
| 452 |
count = 0 |
| 453 |
residues = [] |
| 454 |
|
| 455 |
in_frag = False |
| 456 |
|
| 457 |
for line in open(file): |
| 458 |
|
| 459 |
if line.startswith(' position:'): |
| 460 |
qstart = int(line.split()[1]) |
| 461 |
count = 0 |
| 462 |
|
| 463 |
elif not line.strip(): |
| 464 |
if in_frag: |
| 465 |
count += 1 |
| 466 |
if not top or count <= top: |
| 467 |
frags.append(RosettaFragment(id, qstart, qend, start, end, score, residues)) |
| 468 |
in_frag = False |
| 469 |
id = '' |
| 470 |
start = 0 |
| 471 |
end = 0 |
| 472 |
score = None |
| 473 |
residues = [] |
| 474 |
|
| 475 |
else: |
| 476 |
fields = line.split() |
| 477 |
if not in_frag: |
| 478 |
start = int(fields[2]) |
| 479 |
qend = qstart |
| 480 |
in_frag = True |
| 481 |
else: |
| 482 |
qend += 1 |
| 483 |
|
| 484 |
end = int(fields[2]) |
| 485 |
id = fields[0].lower() + fields[1] |
| 486 |
score = float(fields[8]) |
| 487 |
rank = int(fields[2]) |
| 488 |
aa, ss = fields[3:5] |
| 489 |
phi, psi, omega = map(ang, fields[5:8]) |
| 490 |
residues.append(ResidueInfo(rank, aa, ss, TorsionAngles(phi, psi, omega))) |
| 491 |
|
| 492 |
return RosettaFragmentMap(frags) |