File size: 113,379 Bytes
e611d1f |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 |
from __future__ import annotations
import json
import h5py
import traceback
import numpy as np
import traceback
from tqdm import tqdm
from pathlib import Path
from typing import Optional, Union, BinaryIO, TextIO
from dataclasses import dataclass
from scipy.spatial.distance import cdist
import torch
# ESM
from esm.utils import residue_constants as RC
from esm.utils.structure.protein_chain import ProteinChain
# Biotite
import biotite.structure as bs
from biotite.database import rcsb
from biotite.structure.io.pdb import PDBFile
from biotite.structure import annotate_sse
from cloudpathlib import CloudPath
from Bio.Data import PDBData # Ensure BioPython is imported.
import py3Dmol
# ReCEP Packages
from ..utils.constants import BASE_DIR
from ..utils.loading import load_epitopes_csv, load_epitopes_csv_single, load_species
from .pc import AMINO_ACID_1TO3, AMINO_ACID_3TO1, MAX_ASA
from ..model.ReCEP import ReCEP
from ..data.utils import create_graph_data
PathOrBuffer = Union[str, Path, BinaryIO, TextIO]
@dataclass
class AntigenChain(ProteinChain):
"""
Extended ProteinChain class that adds additional functionalities,
such as computing surface residues based on SASA and maxASA constants.
"""
def __post_init__(self, token: Optional[str] = "1mzAo8l1uxaU8UfVcGgV7B"):
super().__post_init__() # Ensure parent class initialization
# Map residue number to index
self.resnum_to_index = {int(rnum): i for i, rnum in enumerate(self.residue_index)}
# Get epitopes as boolean array
self.epitopes = self.get_epitopes() # Automatically get epitopes on initialization
# Set token from parameter or environment variable
self.token = token
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
@staticmethod
def convert_letter_1to3(letter: str) -> str:
"""
Convert a one-letter amino acid code to its corresponding three-letter code.
Args:
letter (str): A single-character amino acid code (e.g., "A").
Returns:
str: The corresponding three-letter code (e.g., "ALA").
Returns "UNK" if the code is not recognized.
"""
return AMINO_ACID_1TO3.get(letter.upper(), "UNK")
@staticmethod
def convert_letter_3to1(three_letter: str) -> str:
"""
Convert a three-letter amino acid code to its corresponding one-letter code.
Args:
three_letter (str): A three-letter amino acid code (e.g., "ALA").
Returns:
str: The corresponding one-letter code (e.g., "A").
Returns "X" if the code is not recognized.
"""
return AMINO_ACID_3TO1.get(three_letter.upper(), "X")
def get_species(self) -> str:
"""
Get the species of the antigen.
"""
from ..utils.tools import get_chain_organism
species_dict = load_species()
if self.id in species_dict:
species = species_dict[self.id]['classification']
else:
try:
species = get_chain_organism(self.id, self.chain_id)
species_dict[self.id] = {'classification': species}
# Create directory if it doesn't exist
species_file_path = Path(f"{BASE_DIR}/data/species.json")
species_file_path.parent.mkdir(parents=True, exist_ok=True)
with open(species_file_path, "w") as f:
json.dump(species_dict, f, indent=2)
except Exception as e:
print(f"[ERROR] Failed to get species for {self.id}_{self.chain_id}: {str(e)}")
species = "Unknown"
return species
def get_backbone_atoms(self) -> np.ndarray:
"""
Get backbone atom coordinates in the order: CA, C, N.
Returns:
np.ndarray: [L, 3, 3] array where [:, 0] is CA, [:, 1] is C, [:, 2] is N.
"""
file = Path(f"{BASE_DIR}/data/coords/{self.id}_{self.chain_id}.npy")
if file.exists():
return np.load(file)
else:
idx_CA = RC.atom_order["CA"]
idx_C = RC.atom_order["C"]
idx_N = RC.atom_order["N"]
backbone_atoms = self.atom37_positions[:, [idx_N, idx_CA, idx_C], :] # shape: [L, 3, 3]
# Create directory if it doesn't exist
file.parent.mkdir(parents=True, exist_ok=True)
np.save(file, backbone_atoms)
return backbone_atoms
def get_secondary_structure(self) -> np.ndarray:
"""
Get secondary structure information using numpy operations.
"""
try:
ss3_arr = annotate_sse(self.atom_array)
biotite_ss3_str = "".join(ss3_arr)
if len(biotite_ss3_str) != len(self.sequence):
print(f"[WARNING] Secondary structure prediction length ({len(biotite_ss3_str)}) "
f"doesn't match sequence length ({len(self.sequence)}) "
f"for protein {self.id}_{self.chain_id}")
return None
translation_table = str.maketrans({
"a": "H", # alpha helix
"b": "E", # beta sheet
"c": "C", # coil
})
return biotite_ss3_str.translate(translation_table)
except Exception as e:
print(f"[ERROR] Failed to predict secondary structure for "
f"{self.id}_{self.chain_id}: {str(e)}")
return None
def get_ss_onehot(self) -> np.ndarray:
"""
Get one-hot encoded secondary structure information using numpy operations.
Only encode H (helix) and E (sheet), as C (coil) can be inferred.
Returns:
np.ndarray: One-hot encoded secondary structure array of shape (seq_len, 2)
where 2 represents [H, E] (Helix, Sheet)
"""
self.secondary_structure = self.get_secondary_structure()
seq_len = len(self.secondary_structure)
ss_onehot = np.zeros((seq_len, 2), dtype=np.float32)
# Use boolean indexing for helix and sheet only
ss_array = np.array(list(self.secondary_structure))
ss_onehot[:, 0] = (ss_array == 'H')
ss_onehot[:, 1] = (ss_array == 'E')
return ss_onehot
def get_rsa(self) -> np.ndarray:
"""
Calculate relative solvent accessibility (RSA) for all residues.
RSA is the ratio of SASA to maximum ASA for each residue.
Returns:
np.ndarray: An array of RSA values for each residue in the sequence.
"""
cache_file = Path(BASE_DIR) / "data" / "rsa" / f"{self.id}_{self.chain_id}.npy"
if cache_file.exists():
return np.load(cache_file)
sasa_values = self.sasa() # Get SASA values for all residues
rsa_values = np.zeros(len(self.sequence), dtype=np.float32)
# Calculate RSA for each residue
for i, (letter, sasa) in enumerate(zip(self.sequence, sasa_values)):
three_letter = self.convert_letter_1to3(letter)
max_asa = MAX_ASA.get(three_letter)
if max_asa is not None and max_asa != 0:
rsa_values[i] = sasa / max_asa
# Create directory if it doesn't exist
cache_file.parent.mkdir(parents=True, exist_ok=True)
np.save(cache_file, rsa_values)
return rsa_values
def get_surface_residues(self, threshold: float = 0.25) -> list:
"""
Identify surface-exposed residues using RSA values.
A residue is considered surface-exposed if its RSA value
is at least `threshold`.
Args:
threshold (float): The minimum RSA value required to consider
the residue as surface-exposed.
Returns:
tuple: A tuple of two lists, where the first list contains residue numbers (from the PDB) that are surface-exposed,
and the second list contains the indices of the surface residues in the sequence.
"""
rsa_values = self.get_rsa()
surface_residue_numbers = []
surface_residue_indices = []
# Identify surface residues based on RSA threshold
for idx, rsa in enumerate(rsa_values):
if rsa >= threshold:
surface_residue_numbers.append(int(self.residue_index[idx]))
surface_residue_indices.append(idx)
return surface_residue_numbers, surface_residue_indices
def get_epitopes(self, threshold: float = 0.25) -> np.ndarray:
"""
Retrieve epitopes for this chain as a boolean array.
Args:
threshold (float): SASA threshold for determining surface residues.
Returns:
np.ndarray: A boolean array of length L (sequence length) where True indicates
epitope positions and False indicates non-epitope positions.
Only surface-exposed residues can be True.
"""
_, _, epitopes = load_epitopes_csv()
if f'{self.id}_{self.chain_id}' in epitopes:
binary_labels = epitopes.get(f'{self.id}_{self.chain_id}', [0] * len(self.sequence)) # default to 0 if not found
else:
print(f"[WARNING] Epitopes not found for {self.id}_{self.chain_id}. Use single epitopes.")
binary_labels = self.get_epitopes_single()
# Initialize epitope array with False values
epitope_array = np.zeros(len(self.sequence), dtype=bool)
# Check if we have binary labels - handle both list and numpy array cases
if binary_labels is not None and len(binary_labels) > 0:
# Ensure the binary labels match the sequence length
if len(binary_labels) == len(self.sequence):
epitope_array = np.array(binary_labels, dtype=bool)
else:
print(f"[WARNING] Binary labels length ({len(binary_labels)}) doesn't match "
f"sequence length ({len(self.sequence)}) for {self.id}_{self.chain_id}")
return epitope_array
if threshold == 0.0:
return epitope_array
# Filter to ensure only surface residues can be epitopes
_, surface_indices = self.get_surface_residues(threshold=threshold)
# Create surface mask: True for surface residues, False for buried residues
surface_mask = np.zeros(len(self.sequence), dtype=bool)
for res_idx in surface_indices:
if 0 <= res_idx < len(self.sequence):
surface_mask[res_idx] = True
# Apply surface filter: epitopes can only be surface residues
epitope_array = epitope_array & surface_mask
return epitope_array
def get_epitopes_single(self) -> np.ndarray:
"""
Retrieve epitopes for this chain as a boolean array.
"""
_, _, epitopes = load_epitopes_csv_single()
# Try different key formats to find epitopes
possible_keys = [
f'{self.id.upper()}_{self.chain_id}',
f'{self.id}_{self.chain_id}',
f'{self.id.lower()}_{self.chain_id}'
]
epitopes_resnums = None
for key in possible_keys:
if key in epitopes:
epitopes_resnums = epitopes.get(key)
break
if epitopes_resnums is not None:
epitope_array = np.zeros(len(self.sequence), dtype=int)
for resnum in epitopes_resnums:
if resnum in self.resnum_to_index:
epitope_array[self.resnum_to_index[resnum]] = 1
return epitope_array
else:
print(f"[WARNING] Single Epitopes not found for {self.id}_{self.chain_id}. Use no epitopes.")
epitope_array = np.zeros(len(self.sequence), dtype=int)
return epitope_array
def get_epitope_residue_numbers(self) -> list:
"""
Get epitope residue numbers from the boolean epitope array.
Returns:
list: List of residue numbers that are epitopes.
"""
epitope_indices = np.where(self.epitopes)[0]
epitope_residue_numbers = [int(self.residue_index[idx]) for idx in epitope_indices]
return epitope_residue_numbers
def get_embeddings(self, override: bool = False, encoder: str = "esmc") -> np.ndarray:
"""
Retrieve or compute per-residue (full) ESM-C embeddings.
Returns:
np.ndarray: Array of shape (seq_len, embed_dim), dtype float32.
"""
full_file = Path(BASE_DIR) / "data" / "embeddings" / f"{encoder}" / f"{self.id}_{self.chain_id}.h5"
if full_file.exists() and not override:
with h5py.File(full_file, "r") as h5f:
full_embedding = h5f["embedding"][:]
else:
if encoder == "esmc":
if self.token is None:
raise ValueError("ESM token is not set. Please go to https://forge.evolutionaryscale.ai/ to get a token.")
else:
print(f"[INFO] Generating with ESM-C...")
from esm.sdk.api import ESMProtein, LogitsConfig
from esm.sdk.forge import ESM3ForgeInferenceClient
token = self.token
model = ESM3ForgeInferenceClient(
model="esmc-6b-2024-12",
url="https://forge.evolutionaryscale.ai",
token=token
)
config = LogitsConfig(sequence=True, return_embeddings=True)
sequence = self.sequence[:2046] # truncate if too long
protein = ESMProtein(sequence)
protein_tensor = model.encode(protein)
output = model.logits(protein_tensor, config)
full_embedding = output.embeddings.squeeze(0)[1:-1, :].to(torch.float32).cpu().numpy()
full_file.parent.mkdir(parents=True, exist_ok=True)
with h5py.File(full_file, "w") as h5f:
h5f.create_dataset("embedding", data=full_embedding, compression="gzip")
elif encoder == "esm2":
model, alphabet = torch.hub.load("facebookresearch/esm:main", "esm2_t33_650M_UR50D")
batch_converter = alphabet.get_batch_converter()
model.eval()
data = [
("antigen", self.sequence[:2046])
]
batch_labels, batch_strs, batch_tokens = batch_converter(data)
batch_lens = (batch_tokens != alphabet.padding_idx).sum(1)
model.to(self.device)
batch_tokens = batch_tokens.to(self.device)
with torch.no_grad():
results = model(batch_tokens, repr_layers=[33], return_contacts=True)
token_representations = results["representations"][33]
full_embedding = token_representations.squeeze(0)[1:-1, :].to(torch.float32).cpu().numpy()
full_file.parent.mkdir(parents=True, exist_ok=True)
with h5py.File(full_file, "w") as h5f:
h5f.create_dataset("embedding", data=full_embedding, compression="gzip")
return full_embedding
def _scan_surface_residues(self, radius: float, threshold: float = 0.25) -> tuple:
"""
Helper function to compute the surface coverage for each surface residue.
For each surface residue, using its C_alpha coordinate as the center of a sphere with
radius `radius`, determine which surface residues are covered.
Args:
radius (float): The radius of the sphere (in Ångstroms)
threshold (float): Fraction of maximum ASA to define a residue as surface-exposed
Returns:
tuple:
- coverage (dict): Mapping from center residue index to:
(list[int]): List of covered residue indices
(list[int]): List of covered epitope residue indices
(float): Precision
(float): Recall
- max_recall_res (int): Center residue index with highest recall
- max_precision_res (int): Center residue index with highest precision
"""
# Input validation
if radius <= 0:
raise ValueError("Radius must be positive")
if threshold < 0 or threshold > 1:
raise ValueError("Threshold must be between 0 and 1")
# Get surface residues number and indices
surface_res_nums, surface_indices = self.get_surface_residues(threshold=threshold)
# Ensure indices are valid
valid_surface_indices = [
idx for idx in surface_indices
if 0 <= idx < len(self.sequence)
]
valid_surface_res_nums = [
surface_res_nums[surface_indices.index(idx)]
for idx in valid_surface_indices
]
if not valid_surface_indices:
return {}, None, None
# Collect all atoms and their residue indices from surface residues
all_atoms = []
all_res_indices = []
for idx in valid_surface_indices:
mask = self.atom37_mask[idx]
coords = self.atom37_positions[idx][mask]
if len(coords) > 0: # Ensure there are atoms
all_atoms.append(coords)
all_res_indices.extend([idx] * len(coords))
if not all_atoms: # No atoms to process
return {idx: ([], [], 0.0, 0.0) for idx in valid_surface_indices}, None, None
all_atoms = np.vstack(all_atoms).astype(np.float32) # shape: (total_atoms, 3)
all_res_indices = np.array(all_res_indices)
# Collect C-alpha coordinates of surface residues
surface_ca = []
valid_center_indices = []
ca_idx = RC.atom_order["CA"] # Get CA atom index from atom order
for idx in valid_surface_indices:
# Get CA coordinates from atom37_positions
ca_coord = self.atom37_positions[idx, ca_idx, :]
if not np.any(np.isnan(ca_coord)) and self.atom37_mask[idx, ca_idx]: # Ensure CA atom coordinates are valid and atom exists
surface_ca.append(ca_coord)
valid_center_indices.append(idx)
if not surface_ca: # No valid CA atoms
return {}, None, None
surface_ca = np.array(surface_ca, dtype=np.float32)
surface_ca = surface_ca.reshape(-1, 3) # Ensure shape is (n_residues, 3)
# Compute distance matrix between each C-alpha and all atoms
try:
dist_matrix = cdist(surface_ca, all_atoms)
except ValueError as e:
print(f"Error in distance calculation: {e}")
print(f"surface_ca shape: {surface_ca.shape}")
print(f"all_atoms shape: {all_atoms.shape}")
return {}, None, None
max_recall = -1
max_recall_res = None
max_precision = -1
max_precision_res = None
coverage = {}
epitope_indices = np.where(self.epitopes)[0] # Get epitope indices directly
if len(epitope_indices) == 0:
print(f"No epitopes records for protein {self.id}_{self.chain_id}")
for i, center_idx in enumerate(valid_center_indices):
within_radius = dist_matrix[i] < radius
covered_indices = np.unique(all_res_indices[within_radius])
covered_indices_list = covered_indices.tolist()
# Find intersection with epitopes (using indices)
covered_epitope_indices = list(set(covered_indices_list).intersection(set(epitope_indices)))
# Calculate precision and recall
precision = len(covered_epitope_indices) / len(covered_indices_list) if covered_indices_list else 0.0
recall = len(covered_epitope_indices) / len(epitope_indices) if len(epitope_indices) > 0 else 0.0
if recall > max_recall:
max_recall = recall
max_recall_res = center_idx
if precision > max_precision:
max_precision = precision
max_precision_res = center_idx
# Convert to native Python types for JSON compatibility
coverage[int(center_idx)] = (
[int(idx) for idx in covered_indices_list],
[int(idx) for idx in covered_epitope_indices],
float(precision),
float(recall)
)
return coverage, max_recall_res, max_precision_res
def get_surface_coverage(self, radius: float = 18,
threshold: float = 0.25,
index: bool = True,
override: bool = False) -> tuple:
"""
Retrieve (or compute and cache) the coverage mapping for surface residues.
For each surface residue, using its C_alpha as the sphere center (with radius `radius`),
determine which surface residues are covered (i.e. if any atom falls within that sphere).
The result is cached to an HDF5 file for faster subsequent retrieval.
The cache file is saved in BASE_DIR / "data/antigen_sphere", with the file name
"{self.id}_{self.chain_id}.h5", and radius as the first-level key.
Args:
radius (float): The radius of the sphere (in Ångstroms).
threshold (float): Fraction of maximum ASA to define a residue as surface-exposed.
index (bool): If True, return indices instead of residue numbers for easier embeddings/coords access.
override (bool): If True, recompute even if cache exists.
Returns:
tuple:
- coverage (dict): A dictionary mapping each surface residue to a tuple of:
If index=True: center_index -> (list[int]): List of covered residue indices
(list[int]): List of covered epitope residue indices
(float): Precision
(float): Recall
If index=False: center_residue_num -> (list[int]): List of covered residue numbers
(list[int]): List of covered epitope residue numbers
(float): Precision
(float): Recall
- max_recall_res (int): The surface residue number with the highest recall.
- max_precision_res (int): The surface residue number with the highest precision.
"""
# Define the cache directory and file
cache_dir = BASE_DIR / "data" / "antigen_sphere"
cache_dir.mkdir(parents=True, exist_ok=True)
cache_filename = f"{self.id}_{self.chain_id}.h5"
cache_path = cache_dir / cache_filename
radius_key = f"r{radius}"
# If the cache file exists and the radius key exists, load and return the cached result.
if cache_path.exists() and not override:
try:
with h5py.File(cache_path, "r") as h5f:
if radius_key in h5f:
# Load cached data for this radius
radius_group = h5f[radius_key]
if index:
# Cache stores indices, so directly use them
coverage = {}
for center_idx_str in radius_group.keys():
center_idx = int(center_idx_str)
center_group = radius_group[center_idx_str]
covered_indices = center_group['covered_indices'][:].tolist()
covered_epitope_indices = center_group['covered_epitope_indices'][:].tolist()
precision = float(center_group.attrs['precision'])
recall = float(center_group.attrs['recall'])
coverage[center_idx] = (covered_indices, covered_epitope_indices, precision, recall)
return coverage, None, None
else:
# Convert indices to residue numbers
coverage = {}
max_recall = -1
max_recall_res = None
max_precision = -1
max_precision_res = None
for center_idx_str in radius_group.keys():
center_idx = int(center_idx_str)
center_res_num = int(self.residue_index[center_idx])
center_group = radius_group[center_idx_str]
covered_indices = center_group['covered_indices'][:].tolist()
covered_epitope_indices = center_group['covered_epitope_indices'][:].tolist()
precision = float(center_group.attrs['precision'])
recall = float(center_group.attrs['recall'])
# Convert covered indices to residue numbers
covered_res_nums = [int(self.residue_index[idx]) for idx in covered_indices if 0 <= idx < len(self.residue_index)]
covered_epitope_res_nums = [int(self.residue_index[idx]) for idx in covered_epitope_indices if 0 <= idx < len(self.residue_index)]
coverage[center_res_num] = (covered_res_nums, covered_epitope_res_nums, precision, recall)
if recall > max_recall:
max_recall = recall
max_recall_res = center_res_num
if precision > max_precision:
max_precision = precision
max_precision_res = center_res_num
return coverage, max_recall_res, max_precision_res
except (OSError, KeyError, ValueError) as e:
print(f"[WARNING] Error reading cache file {cache_path}: {e}")
print(f"[INFO] Recomputing surface coverage...")
# Otherwise, compute the coverage mapping (returns index-based results)
coverage, max_recall_res, max_precision_res = self._scan_surface_residues(radius, threshold)
# Save the result to HDF5 file
# Create or open the HDF5 file and save data under the radius key
with h5py.File(cache_path, "a") as h5f: # "a" mode: read/write if exists, create otherwise
# Create or overwrite the radius group
if radius_key in h5f:
del h5f[radius_key] # Remove existing group if override or recompute
radius_group = h5f.create_group(radius_key)
# Save each center residue's data
for center_idx, (covered_indices, covered_epitope_indices, precision, recall) in coverage.items():
center_group = radius_group.create_group(str(center_idx))
center_group.create_dataset('covered_indices', data=np.array(covered_indices, dtype=np.int32), compression='gzip')
center_group.create_dataset('covered_epitope_indices', data=np.array(covered_epitope_indices, dtype=np.int32), compression='gzip')
center_group.attrs['precision'] = precision
center_group.attrs['recall'] = recall
# Convert to residue numbers if index=False is requested
if not index:
coverage_resnums = {}
max_recall_res_num = None
max_precision_res_num = None
if max_recall_res is not None:
max_recall_res_num = int(self.residue_index[max_recall_res])
if max_precision_res is not None:
max_precision_res_num = int(self.residue_index[max_precision_res])
for center_idx, (covered_indices, covered_epitope_indices, precision, recall) in coverage.items():
center_res_num = int(self.residue_index[center_idx])
# Convert covered indices to residue numbers
covered_res_nums = [int(self.residue_index[idx]) for idx in covered_indices if 0 <= idx < len(self.residue_index)]
covered_epitope_res_nums = [int(self.residue_index[idx]) for idx in covered_epitope_indices if 0 <= idx < len(self.residue_index)]
coverage_resnums[center_res_num] = (covered_res_nums, covered_epitope_res_nums, precision, recall)
return coverage_resnums, max_recall_res_num, max_precision_res_num
return coverage, max_recall_res, max_precision_res
def data_preparation(self, radius: float = None, encoder: str = "esmc", override: bool = False):
"""
Retrieve or compute region embeddings for surface residues using spherical regions.
Args:
radius (float): Radius to define the neighborhood of each center residue.
threshold (float): Threshold to determine surface residues.
cover (bool): Whether to recompute and overwrite cached data.
verbose (bool): Whether to print progress information.
Returns:
tuple:
- embeddings (np.ndarray): Array of embeddings mean of the region. (num_regions, embedding_dim)
- center_residues (np.ndarray): Array of center residue numbers. (num_regions,)
- precisions (np.ndarray): Array of precision values for each center residue. (num_regions,)
- recalls (np.ndarray): Array of recall values for each center residue. (num_regions,)
"""
embeddings = self.get_embeddings(encoder=encoder)
backbone_atoms = self.get_backbone_atoms()
rsa = self.get_rsa()
if radius is None:
# Used for creating data
for i in range(16,21,2):
_, _, _ = self.get_surface_coverage(radius=i, override=override)
return embeddings, backbone_atoms, rsa, None
else:
coverage_dict, _, _ = self.get_surface_coverage(radius=radius, override=override)
return embeddings, backbone_atoms, rsa, coverage_dict
def evaluate(self, model_path: str = None, device_id: int = 1, radius: float = 19.0, k: int = 7,
threshold: float = None, verbose: bool = True, encoder: str = "esmc", use_gpu: bool = True):
"""
Evaluate epitopes using ReCEP model with spherical regions.
Args:
model_path (str): Path to the trained ReCEP model
device_id (int): GPU device ID to use
radius (float): Radius for spherical regions
k (int): Number of top regions to select
threshold (float): Threshold for node-level epitope prediction
verbose (bool): Whether to print progress information
Returns:
dict: Dictionary containing:
- 'predicted_epitopes': List of predicted epitope residue numbers
- 'true_epitopes': Set of true epitope residue numbers
- 'precision': Final prediction precision
- 'recall': Final prediction recall
- 'top_k_regions': Information about selected regions
"""
# Set device
if use_gpu and torch.cuda.is_available() and device_id >= 0:
device = torch.device(f"cuda:{device_id}")
else:
device = torch.device("cpu")
if verbose:
print(f"[INFO] Using device: {device}")
# Load ReCEP model
try:
if model_path is None:
model_path = f"{BASE_DIR}/models/ReCEP/20250626_110438/best_mcc_model.bin"
if threshold is None:
model, threshold = ReCEP.load(model_path, device=device, strict=False, verbose=False)
else:
model, _ = ReCEP.load(model_path, device=device, strict=False, verbose=False)
model.eval()
if verbose:
print(f"[INFO] Loaded ReCEP model from {model_path}")
except Exception as e:
if verbose:
print(f"[ERROR] Failed to load model: {str(e)}")
return {}
# Get protein data using data_preparation
try:
embeddings, backbone_atoms, rsa, coverage_dict = self.data_preparation(radius=radius, encoder=encoder)
if verbose:
print(f"[INFO] Retrieved protein data for {len(coverage_dict)} surface regions")
except Exception as e:
if verbose:
print(f"[ERROR] Failed to prepare data: {str(e)}")
traceback.print_exc()
return {}
if not coverage_dict:
if verbose:
print("[WARNING] No surface regions found")
return {}
# Get epitope indices
epitope_indices = np.where(self.epitopes)[0].tolist()
# Phase 1: Predict graph-level values for all regions
region_predictions = []
with torch.no_grad():
for center_idx, (covered_indices, covered_epitope_indices, precision, recall) in tqdm(
coverage_dict.items(), desc="Predicting region values", disable=not verbose):
if len(covered_indices) < 2: # Skip regions with too few residues
continue
try:
# Create graph data for this region
graph_data = create_graph_data(
center_idx=center_idx,
covered_indices=covered_indices,
covered_epitope_indices=covered_epitope_indices,
embeddings=embeddings,
backbone_atoms=backbone_atoms,
rsa_values=rsa,
epitope_indices=epitope_indices,
recall=recall,
precision=precision,
pdb_id=self.id,
chain_id=self.chain_id,
verbose=True # Enable verbose to see errors
)
if graph_data is None:
if verbose:
print(f"[WARNING] Failed to create graph data for region {center_idx}")
continue
# Move data to device
graph_data = graph_data.to(device)
# Create batch tensor for single graph - this is crucial!
graph_data.batch = torch.zeros(graph_data.num_nodes, dtype=torch.long, device=device)
# Predict using ReCEP model (following trainer.py pattern)
outputs = model(graph_data)
# Get graph-level prediction
if 'global_pred' in outputs:
graph_pred = torch.sigmoid(outputs['global_pred']).cpu().item()
else:
# Fallback: use mean of node predictions as graph prediction
node_preds = torch.sigmoid(outputs['node_preds']).cpu().numpy()
graph_pred = float(np.mean(node_preds))
region_predictions.append({
'center_idx': center_idx,
'covered_indices': covered_indices,
'covered_epitope_indices': covered_epitope_indices,
'graph_pred': graph_pred,
'true_recall': recall,
'graph_data': graph_data
})
except Exception as e:
if verbose:
print(f"[WARNING] Error processing region {center_idx}: {str(e)}")
traceback.print_exc()
continue
if not region_predictions:
if verbose:
print("[WARNING] No valid region predictions")
return {}
# Phase 2: Select top-k regions based on graph predictions
region_predictions.sort(key=lambda x: x['graph_pred'], reverse=True)
top_k_regions = region_predictions[:k]
if verbose:
print(f"[INFO] Selected top {len(top_k_regions)} regions:")
for i, region in enumerate(top_k_regions):
print(f" Region {i+1}: center={region['center_idx']}, "
f"predicted_value={region['graph_pred']:.3f}, "
f"true_recall={region['true_recall']:.3f}")
# Phase 3: Predict node-level epitopes for selected regions
residue_votes = {} # residue_idx -> [list of binary predictions]
residue_probs = {} # residue_idx -> [list of probabilities]
with torch.no_grad():
for region in tqdm(top_k_regions, desc="Predicting node values", disable=not verbose):
try:
graph_data = region['graph_data']
# Ensure graph data has batch information - this is crucial!
if not hasattr(graph_data, 'batch') or graph_data.batch is None:
graph_data.batch = torch.zeros(graph_data.num_nodes, dtype=torch.long, device=device)
# Predict using ReCEP model (following trainer.py pattern)
outputs = model(graph_data)
# Get node-level predictions
node_preds = torch.sigmoid(outputs['node_preds']).cpu().numpy()
# Store votes and probabilities for each residue
for local_idx, residue_idx in enumerate(region['covered_indices']):
if residue_idx not in residue_votes:
residue_votes[residue_idx] = []
residue_probs[residue_idx] = []
# Store probability and binary vote
prob = float(node_preds[local_idx])
residue_probs[residue_idx].append(prob)
# Binary vote based on threshold
vote = 1 if prob >= threshold else 0
residue_votes[residue_idx].append(vote)
except Exception as e:
if verbose:
print(f"[WARNING] Error in node prediction for region {region['center_idx']}: {str(e)}")
traceback.print_exc()
continue
# Create predictions dictionary for all residues
all_residue_predictions = {}
for idx in range(len(self.residue_index)):
residue_num = int(self.residue_index[idx])
if idx in residue_probs:
# Calculate mean probability for residues in top-k regions
all_residue_predictions[residue_num] = float(np.mean(residue_probs[idx]))
else:
# Set probability to 1e-5 for residues not in any top-k region
all_residue_predictions[residue_num] = 1e-2
# Phase 4a: Apply voting mechanism for voted_epitopes
voted_epitope_indices = []
for residue_idx, votes in residue_votes.items():
# If >= half of the votes are positive, predict as epitope
if sum(votes) >= len(votes) / 2:
voted_epitope_indices.append(residue_idx)
# Convert indices to residue numbers for voted epitopes
voted_epitope_resnums = [int(self.residue_index[idx]) for idx in voted_epitope_indices
if 0 <= idx < len(self.residue_index)]
# Phase 4b: Apply probability threshold for predicted_epitopes
predicted_epitope_resnums = []
for residue_num, prob in all_residue_predictions.items():
if prob >= threshold:
predicted_epitope_resnums.append(residue_num)
# Get true epitopes
true_epitope_resnums = set(self.get_epitope_residue_numbers())
# Calculate metrics for both prediction methods
# Metrics for voted epitopes
voted_tp = len(set(voted_epitope_resnums) & true_epitope_resnums)
voted_precision = voted_tp / len(voted_epitope_resnums) if voted_epitope_resnums else 0
voted_recall = voted_tp / len(true_epitope_resnums) if true_epitope_resnums else 0
# Metrics for probability-based epitopes
predicted_tp = len(set(predicted_epitope_resnums) & true_epitope_resnums)
predicted_precision = predicted_tp / len(predicted_epitope_resnums) if predicted_epitope_resnums else 0
predicted_recall = predicted_tp / len(true_epitope_resnums) if true_epitope_resnums else 0
if verbose:
print(f"\n[INFO] Final Results:")
print(f" True epitopes: {len(true_epitope_resnums)}")
print(f" Residues in top-k regions: {len(residue_probs)}/{len(self.residue_index)}")
print(f"\n Voting-based prediction:")
print(f" Voted epitopes: {len(voted_epitope_resnums)}")
print(f" Voted precision: {voted_precision:.3f}")
print(f" Voted recall: {voted_recall:.3f}")
print(f"\n Probability-based prediction (threshold={threshold}):")
print(f" Predicted epitopes: {len(predicted_epitope_resnums)}")
print(f" Predicted precision: {predicted_precision:.3f}")
print(f" Predicted recall: {predicted_recall:.3f}")
return {
'predicted_epitopes': predicted_epitope_resnums, # Based on probability threshold
'voted_epitopes': voted_epitope_resnums, # Based on voting mechanism
'true_epitopes': true_epitope_resnums,
'predicted_precision': predicted_precision, # Precision for probability-based
'predicted_recall': predicted_recall, # Recall for probability-based
'voted_precision': voted_precision, # Precision for voting-based
'voted_recall': voted_recall, # Recall for voting-based
'predictions': all_residue_predictions, # All residue probabilities
'top_k_regions': [
{
'center_residue': int(self.residue_index[region['center_idx']]),
'center_idx': region['center_idx'],
'predicted_value': region['graph_pred'],
'true_recall': region['true_recall'],
'covered_residues': [int(self.residue_index[idx]) for idx in region['covered_indices']]
}
for region in top_k_regions
],
'residue_votes': {
int(self.residue_index[idx]): votes
for idx, votes in residue_votes.items()
if 0 <= idx < len(self.residue_index)
}
}
def predict(self, model_path: str = None, device_id: int = 1, radius: float = 19.0, k: int = 7,
threshold: float = None, verbose: bool = True, encoder: str = "esmc", use_gpu: bool = True,
auto_cleanup: bool = False):
"""
Predict epitopes using ReCEP model with spherical regions (for unknown true epitopes).
Args:
model_path (str): Path to the trained ReCEP model
device_id (int): GPU device ID to use
radius (float): Radius for spherical regions
k (int): Number of top regions to select
threshold (float): Threshold for node-level epitope prediction
verbose (bool): Whether to print progress information
encoder (str): Encoder type for embeddings
use_gpu (bool): Whether to use GPU for computation
auto_cleanup (bool): Whether to automatically delete generated data files after prediction
Returns:
dict: Dictionary containing:
- 'predicted_epitopes': List of predicted epitope residue numbers
- 'predictions': Dictionary of all residue probabilities {resnum: probability}
- 'top_k_centers': List of top-k center residue numbers
- 'top_k_region_residues': List of all residues covered by top-k regions (union)
- 'top_k_regions': Detailed information about selected regions
"""
# Set device
if use_gpu and torch.cuda.is_available() and device_id >= 0:
device = torch.device(f"cuda:{device_id}")
else:
device = torch.device("cpu")
if verbose:
print(f"[INFO] Using device: {device}")
# Load ReCEP model
try:
if model_path is None:
model_path = f"{BASE_DIR}/models/ReCEP/20250626_110438/best_mcc_model.bin"
if threshold is None:
model, threshold = ReCEP.load(model_path, device=device, strict=False, verbose=False)
else:
model, _ = ReCEP.load(model_path, device=device, strict=False, verbose=False)
model.eval()
if verbose:
print(f"[INFO] Loaded ReCEP model from {model_path}")
except Exception as e:
if verbose:
print(f"[ERROR] Failed to load model: {str(e)}")
return {}
# Get protein data using data_preparation
try:
embeddings, backbone_atoms, rsa, coverage_dict = self.data_preparation(radius=radius, encoder=encoder)
if verbose:
print(f"[INFO] Retrieved protein data for {len(coverage_dict)} surface regions")
except Exception as e:
if verbose:
print(f"[ERROR] Failed to prepare data: {str(e)}")
traceback.print_exc()
return {}
if not coverage_dict:
if verbose:
print("[WARNING] No surface regions found")
return {}
# Phase 1: Predict graph-level values for all regions
region_predictions = []
with torch.no_grad():
for center_idx, (covered_indices, covered_epitope_indices, precision, recall) in tqdm(
coverage_dict.items(), desc="Predicting region values", disable=not verbose):
if len(covered_indices) < 2: # Skip regions with too few residues
continue
try:
# Create graph data for this region (without epitope information)
graph_data = create_graph_data(
center_idx=center_idx,
covered_indices=covered_indices,
covered_epitope_indices=[], # No epitope information for prediction
embeddings=embeddings,
backbone_atoms=backbone_atoms,
rsa_values=rsa,
epitope_indices=[], # No epitope information for prediction
recall=0.0, # No recall information
precision=0.0, # No precision information
pdb_id=self.id,
chain_id=self.chain_id,
verbose=False
)
if graph_data is None:
if verbose:
print(f"[WARNING] Failed to create graph data for region {center_idx}")
continue
# Move data to device
graph_data = graph_data.to(device)
# Create batch tensor for single graph
graph_data.batch = torch.zeros(graph_data.num_nodes, dtype=torch.long, device=device)
# Predict using ReCEP model
outputs = model(graph_data)
# Get graph-level prediction
if 'global_pred' in outputs:
graph_pred = torch.sigmoid(outputs['global_pred']).cpu().item()
else:
# Fallback: use mean of node predictions as graph prediction
node_preds = torch.sigmoid(outputs['node_preds']).cpu().numpy()
graph_pred = float(np.mean(node_preds))
region_predictions.append({
'center_idx': center_idx,
'covered_indices': covered_indices,
'graph_pred': graph_pred,
'graph_data': graph_data
})
except Exception as e:
if verbose:
print(f"[WARNING] Error processing region {center_idx}: {str(e)}")
traceback.print_exc()
continue
if not region_predictions:
if verbose:
print("[WARNING] No valid region predictions")
return {}
# Phase 2: Select top-k regions based on graph predictions
region_predictions.sort(key=lambda x: x['graph_pred'], reverse=True)
top_k_regions = region_predictions[:k]
if verbose:
print(f"[INFO] Selected top {len(top_k_regions)} regions:")
for i, region in enumerate(top_k_regions):
print(f" Region {i+1}: center={region['center_idx']}, "
f"predicted_value={region['graph_pred']:.3f}")
# Phase 3: Predict node-level epitopes for selected regions
residue_probs = {} # residue_idx -> [list of probabilities]
with torch.no_grad():
for region in tqdm(top_k_regions, desc="Predicting node values", disable=not verbose):
try:
graph_data = region['graph_data']
# Ensure graph data has batch information
if not hasattr(graph_data, 'batch') or graph_data.batch is None:
graph_data.batch = torch.zeros(graph_data.num_nodes, dtype=torch.long, device=device)
# Predict using ReCEP model
outputs = model(graph_data)
# Get node-level predictions
node_preds = torch.sigmoid(outputs['node_preds']).cpu().numpy()
# Store probabilities for each residue
for local_idx, residue_idx in enumerate(region['covered_indices']):
if residue_idx not in residue_probs:
residue_probs[residue_idx] = []
# Store probability
prob = float(node_preds[local_idx])
residue_probs[residue_idx].append(prob)
except Exception as e:
if verbose:
print(f"[WARNING] Error in node prediction for region {region['center_idx']}: {str(e)}")
traceback.print_exc()
continue
# Create predictions dictionary for all residues
all_residue_predictions = {}
for idx in range(len(self.residue_index)):
residue_num = int(self.residue_index[idx])
if idx in residue_probs:
# Calculate mean probability for residues in top-k regions
all_residue_predictions[residue_num] = float(np.mean(residue_probs[idx]))
else:
# Set probability to 0 for residues not in any top-k region
all_residue_predictions[residue_num] = 0.0
# Apply probability threshold for predicted epitopes
predicted_epitope_resnums = []
node_mean = 0.0
for residue_num, prob in all_residue_predictions.items():
node_mean += prob
if prob >= threshold:
predicted_epitope_resnums.append(residue_num)
node_mean /= len(all_residue_predictions) if all_residue_predictions else 1
# Get top-k center residue numbers
top_k_centers = [int(self.residue_index[region['center_idx']]) for region in top_k_regions]
# Get union of all residues covered by top-k regions and mean graph predicted value
graph_mean = 0.0
all_covered_indices = set()
for region in top_k_regions:
all_covered_indices.update(region['covered_indices'])
graph_mean += region['graph_pred']
graph_mean /= len(top_k_regions)
top_k_region_residues = [int(self.residue_index[idx]) for idx in all_covered_indices
if 0 <= idx < len(self.residue_index)]
if verbose:
print(f"\n[INFO] Prediction Results:")
print(f" Predicted epitopes: {len(predicted_epitope_resnums)}")
print(f" Top-k centers: {top_k_centers}")
print(f" Total residues in top-k regions: {len(top_k_region_residues)}")
# Prepare return results
results = {
'predicted_epitopes': predicted_epitope_resnums,
'predictions': all_residue_predictions,
'top_k_centers': top_k_centers,
'top_k_region_residues': top_k_region_residues,
'top_k_regions': [
{
'center_residue': int(self.residue_index[region['center_idx']]),
'center_idx': region['center_idx'],
'predicted_value': region['graph_pred'],
'covered_residues': [int(self.residue_index[idx]) for idx in region['covered_indices']]
}
for region in top_k_regions
],
'antigen_rate': graph_mean,
'epitope_rate': node_mean
}
# Auto-cleanup generated data files if requested
if auto_cleanup:
self._cleanup_generated_data(encoder=encoder, verbose=verbose)
return results
def _cleanup_generated_data(self, encoder: str = "esmc", verbose: bool = True):
"""
Clean up generated data files for this antigen chain.
Args:
encoder (str): Encoder type used for embeddings
verbose (bool): Whether to print cleanup information
"""
import os
# List of files to delete
files_to_delete = [
# Embeddings file
Path(BASE_DIR) / "data" / "embeddings" / encoder / f"{self.id}_{self.chain_id}.h5",
# Backbone atoms file
Path(BASE_DIR) / "data" / "coords" / f"{self.id}_{self.chain_id}.npy",
# RSA file
Path(BASE_DIR) / "data" / "rsa" / f"{self.id}_{self.chain_id}.npy",
# Surface coverage file
Path(BASE_DIR) / "data" / "antigen_sphere" / f"{self.id}_{self.chain_id}.h5"
]
deleted_files = []
failed_deletions = []
total_size = 0
for file_path in files_to_delete:
if file_path.exists():
try:
# Get file size before deletion
file_size = file_path.stat().st_size
os.remove(file_path)
deleted_files.append(file_path)
total_size += file_size
if verbose:
print(f"[INFO] Deleted: {file_path}")
except Exception as e:
failed_deletions.append((file_path, str(e)))
if verbose:
print(f"[WARNING] Failed to delete {file_path}: {str(e)}")
else:
if verbose:
print(f"[INFO] File not found (already deleted or not generated): {file_path}")
if verbose:
print(f"[INFO] Cleanup completed for {self.id}_{self.chain_id}")
print(f" - Files deleted: {len(deleted_files)}")
print(f" - Failed deletions: {len(failed_deletions)}")
if total_size > 0:
print(f" - Total space freed: {total_size / (1024**2):.2f} MB")
def visualize(self,
mode: str = 'normal',
style: str = 'cartoon',
predicted_epitopes: list = None,
predict_results: dict = None,
prediction_mode: str = 'residue', # 'residue' or 'region'
center_res: int = None,
radius: float = None,
region_index: int = None, # Index of specific region to show (0-based)
width: int = 800,
height: int = 600,
base_color: str = '#e6e6f7',
true_epitope_color: str = '#f1b54c', # True epitopes (deeper blue)
false_positive_color: str = '#ef5331', # False positives (deeper red)
true_positive_color: str = '#a0d293', # True positives (deeper green)
coverage_color: str = '#9C6ADE', # Coverage regions (purple)
prediction_color: str = '#9C6ADE', # Prediction color (purple)
center_color: str = '#2C3E50', # Center residue (dark gray)
probability_colormap: str = 'RdYlBu_r', # Colormap for probability visualization
show_surface: bool = True,
show_shape: bool = True,
show_center: bool = True,
center_radius: float = 0.7,
n_points: int = 50,
shape_opacity: float = 0.3,
surface_opacity: float = 1.0,
wireframe: bool = True,
show_epitope: bool = True,
show_coverage: bool = True,
show_top_regions: bool = True,
max_spheres: int = None, # Maximum number of spheres to show
prob_threshold: float = 0.5):
"""
Visualize the protein chain with various modes and integration with predict results.
Args:
mode (str): Visualization mode. Options:
- 'normal': Basic protein structure
- 'epitope': Show predicted epitopes vs true epitopes
- 'coverage': Show spherical coverage region
- 'evaluation': Show evaluation results from evaluate() function
- 'prediction': Show prediction results from predict() function
- 'probability': Show residue probabilities as color gradient
- 'top_regions': Show top-k regions from prediction
- 'comparison': Compare voted vs predicted epitopes
prediction_mode (str): Sub-mode for prediction visualization ('residue' or 'region')
- 'residue': Color predicted epitopes by probability (gradient purple)
- 'region': Color all residues in top-k regions uniformly
style (str): Protein representation style ('cartoon', 'stick', 'sphere', 'surface')
predicted_epitopes (list): List of predicted epitope residue numbers
predict_results (dict): Results dictionary from predict() function
center_res (int): Center residue number for coverage visualization
radius (float): Radius for spherical coverage
region_index (int): Index of specific region to show in probability mode (0-based)
If None, shows all regions
Each region uses a distinct color for shape visualization
probability_colormap (str): Colormap name for probability visualization
prob_threshold (float): Threshold for probability-based coloring
... (other parameters as before)
Returns:
py3Dmol.view: The molecular visualization view object
"""
# Create view object and add basic structure
view = self._create_base_view(width, height)
# Set basic style
style_dict = {
'cartoon': {'cartoon': {}},
'stick': {'stick': {}},
'sphere': {'sphere': {}},
'surface': {'surface': {}}
}
base_style = style_dict.get(style, {'cartoon': {}})
# Visualization based on mode
if mode == 'epitope' and predicted_epitopes is not None:
self._add_epitope_visualization(
view, style, predicted_epitopes,
base_color, true_epitope_color, false_positive_color,
true_positive_color, coverage_color,
show_surface, surface_opacity, show_coverage,
center_res, radius
)
# Add shape visualization if needed
if show_shape and center_res is not None and radius is not None:
self._add_shape_visualization(
view, center_res, radius,
coverage_color, center_color,
show_center, center_radius,
shape_opacity, wireframe
)
elif mode == 'coverage' and center_res is not None and radius is not None:
self._add_coverage_visualization(
view, style, center_res, radius,
base_color, coverage_color, true_positive_color, true_epitope_color,
show_surface, show_shape, show_center,
surface_opacity, shape_opacity, center_radius,
n_points, center_color, wireframe, show_epitope
)
elif mode == 'evaluation' and predict_results is not None:
self._add_evaluation_visualization(
view, style, predict_results,
base_color, true_epitope_color, false_positive_color,
true_positive_color, coverage_color,
show_surface, surface_opacity, show_shape, radius, max_spheres
)
elif mode == 'prediction' and predict_results is not None:
self._add_prediction_visualization(
view, style, predict_results, prediction_mode,
base_color, prediction_color, show_surface, surface_opacity,
show_shape, shape_opacity, show_center, center_radius,
wireframe, radius, max_spheres
)
elif mode == 'probability' and predict_results is not None:
self._add_probability_visualization(
view, style, predict_results,
base_color, probability_colormap, show_surface, surface_opacity,
prob_threshold, region_index, radius, show_shape, shape_opacity,
show_center, center_radius, wireframe, coverage_color, center_color
)
elif mode == 'top_regions' and predict_results is not None:
self._add_top_regions_visualization(
view, style, predict_results,
base_color, coverage_color, center_color,
show_surface, show_shape, show_center,
surface_opacity, shape_opacity, center_radius,
wireframe, radius, max_spheres
)
elif mode == 'comparison' and predict_results is not None:
self._add_comparison_visualization(
view, style, predict_results,
base_color, true_epitope_color, false_positive_color,
true_positive_color, coverage_color, show_surface, surface_opacity
)
else:
# Default mode: just show the basic structure
view.setStyle({'chain': self.chain_id}, base_style)
# Adjust view
view.zoomTo()
return view
def _add_prediction_visualization(self, view, style, predict_results, prediction_mode,
base_color, prediction_color, show_surface, surface_opacity,
show_shape, shape_opacity, show_center, center_radius,
wireframe, radius, max_spheres):
"""Add visualization for prediction results"""
if prediction_mode == 'residue':
self._add_prediction_residue_mode(
view, style, predict_results, base_color, prediction_color,
show_surface, surface_opacity
)
elif prediction_mode == 'region':
self._add_prediction_region_mode(
view, style, predict_results, base_color, prediction_color,
show_surface, surface_opacity, show_shape, shape_opacity,
show_center, center_radius, wireframe, radius, max_spheres
)
def _add_prediction_residue_mode(self, view, style, predict_results, base_color, prediction_color,
show_surface, surface_opacity):
"""Add visualization for prediction results in residue mode"""
import matplotlib.pyplot as plt
import matplotlib.colors as mcolors
# Get predictions dictionary
predictions = predict_results.get('predictions', {})
predicted_epitopes = predict_results.get('predicted_epitopes', [])
# Get style configuration
style_dict = {
'cartoon': {'cartoon': {}},
'stick': {'stick': {}},
'sphere': {'sphere': {}},
'surface': {'surface': {}}
}
base_style = style_dict.get(style, {'cartoon': {}})
if not predictions:
# Fallback to basic visualization
view.setStyle({'chain': self.chain_id}, {**base_style,
list(base_style.keys())[0]: {**list(base_style.values())[0], 'color': base_color}})
if show_surface:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Softer opacity for fallback
'color': base_color
}, {'chain': self.chain_id})
return
# Filter predictions to only include predicted epitopes
epitope_predictions = {res: prob for res, prob in predictions.items()
if res in predicted_epitopes}
if not epitope_predictions:
# No predicted epitopes, show base style
view.setStyle({'chain': self.chain_id}, {**base_style, list(base_style.keys())[0]: {**list(base_style.values())[0], 'color': base_color}})
if show_surface:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Softer opacity for fallback
'color': base_color
}, {'chain': self.chain_id})
return
# Get probability range for predicted epitopes only
probs = list(epitope_predictions.values())
min_prob, max_prob = min(probs), max(probs)
# Improved color scheme - use orange to red gradient for better contrast with gray
# This avoids confusion with gray background when probability is low
epitope_colors = [
'#FFE4B5', # Light orange (moccasin)
'#FFD700', # Gold
'#FFA500', # Orange
'#FF8C00', # Dark orange
'#FF6347', # Tomato
'#FF4500', # Orange red
'#DC143C' # Crimson
]
n_colors = len(epitope_colors)
# Set base style for entire protein
view.setStyle({'chain': self.chain_id}, {**base_style, list(base_style.keys())[0]: {**list(base_style.values())[0], 'color': base_color}})
# Color predicted epitopes based on probability with orange-red gradient
for residue_num, prob in epitope_predictions.items():
# Normalize probability to [0, 1] within the epitope range
if max_prob > min_prob:
norm_prob = (prob - min_prob) / (max_prob - min_prob)
else:
norm_prob = 0.5
# Map to color index
color_idx = int(norm_prob * (n_colors - 1))
color_idx = max(0, min(color_idx, n_colors - 1))
color = epitope_colors[color_idx]
# Add style for this residue with vivid color
style_name = list(base_style.keys())[0]
colored_style = {style_name: {'color': color}}
view.addStyle(
{'chain': self.chain_id, 'resi': residue_num},
colored_style
)
# Add surface overlay if requested
if show_surface:
# Add base surface for non-epitope regions
all_residues = set(int(res) for res in self.residue_index)
non_epitope_residues = all_residues - set(predicted_epitopes)
if non_epitope_residues:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Softer opacity for fallback
'color': base_color
}, {'chain': self.chain_id, 'resi': list(non_epitope_residues)})
# Add colored surfaces for predicted epitopes
for residue_num, prob in epitope_predictions.items():
# Normalize probability to [0, 1] within the epitope range
if max_prob > min_prob:
norm_prob = (prob - min_prob) / (max_prob - min_prob)
else:
norm_prob = 0.5
# Map to color index
color_idx = int(norm_prob * (n_colors - 1))
color_idx = max(0, min(color_idx, n_colors - 1))
color = epitope_colors[color_idx]
# Add surface for this residue
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity,
'color': color
}, {'chain': self.chain_id, 'resi': residue_num})
def _add_prediction_region_mode(self, view, style, predict_results, base_color, prediction_color,
show_surface, surface_opacity, show_shape, shape_opacity,
show_center, center_radius, wireframe, radius, max_spheres):
"""Add visualization for prediction results in region mode"""
# Get top-k regions
top_k_regions = predict_results.get('top_k_regions', [])
top_k_region_residues = predict_results.get('top_k_region_residues', [])
# Get style configuration
style_dict = {
'cartoon': {'cartoon': {}},
'stick': {'stick': {}},
'sphere': {'sphere': {}},
'surface': {'surface': {}}
}
base_style = style_dict.get(style, {'cartoon': {}})
if not top_k_region_residues:
# No regions, show base style
view.setStyle({'chain': self.chain_id}, {**base_style, list(base_style.keys())[0]: {**list(base_style.values())[0], 'color': base_color}})
if show_surface:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Softer opacity for fallback
'color': base_color
}, {'chain': self.chain_id})
return
# Set base style for entire protein
view.setStyle({'chain': self.chain_id}, {**base_style, list(base_style.keys())[0]: {**list(base_style.values())[0], 'color': base_color}})
# Color all residues in top-k regions with uniform purple
if top_k_region_residues:
style_name = list(base_style.keys())[0]
colored_style = {style_name: {'color': prediction_color}}
view.addStyle(
{'chain': self.chain_id, 'resi': top_k_region_residues},
colored_style
)
# Add surface overlay if requested
if show_surface:
# Add base surface for non-region residues
all_residues = set(int(res) for res in self.residue_index)
non_region_residues = all_residues - set(top_k_region_residues)
if non_region_residues:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Softer opacity for fallback
'color': base_color
}, {'chain': self.chain_id, 'resi': list(non_region_residues)})
# Color all residues in top-k regions with uniform purple surface
if top_k_region_residues:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity,
'color': prediction_color
}, {'chain': self.chain_id, 'resi': top_k_region_residues})
# Add spherical regions if requested
if show_shape and top_k_regions:
self._add_multi_shape_visualization(
view, top_k_regions, radius, max_spheres,
show_center, center_radius, shape_opacity, wireframe
)
def _add_evaluation_visualization(self, view, style, predict_results,
base_color, true_epitope_color, false_positive_color,
true_positive_color, coverage_color,
show_surface, surface_opacity, show_shape, radius, max_spheres):
"""Add visualization for evaluation results"""
# Get prediction results
predicted_epitopes = set(predict_results.get('predicted_epitopes', []))
true_epitopes = set(predict_results.get('true_epitopes', []))
# Calculate different categories
true_positives = predicted_epitopes & true_epitopes
false_positives = predicted_epitopes - true_epitopes
false_negatives = true_epitopes - predicted_epitopes
# Get style configuration
style_dict = {
'cartoon': {'cartoon': {}},
'stick': {'stick': {}},
'sphere': {'sphere': {}},
'surface': {'surface': {}}
}
base_style = style_dict.get(style, {'cartoon': {}})
# Set base style for entire protein
view.setStyle({'chain': self.chain_id}, {**base_style, list(base_style.keys())[0]: {**list(base_style.values())[0], 'color': base_color}})
# Add colored styles for specific categories with vivid colors
for residues, color in [
(true_positives, true_positive_color),
(false_positives, false_positive_color),
(false_negatives, true_epitope_color)
]:
if residues:
# Create style with the specified color
style_name = list(base_style.keys())[0]
colored_style = {style_name: {'color': color}}
view.addStyle(
{'chain': self.chain_id, 'resi': list(residues)},
colored_style
)
# Add surface overlay if requested (works with any base style)
if show_surface:
# Get all colored residues
all_colored_residues = true_positives | false_positives | false_negatives
# Only add base surface for non-colored regions to avoid covering colored surfaces
if all_colored_residues:
all_residues = set(int(res) for res in self.residue_index)
non_colored_residues = all_residues - all_colored_residues
if non_colored_residues:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Softer opacity for fallback
'color': base_color
}, {'chain': self.chain_id, 'resi': list(non_colored_residues)})
else:
# If no colored residues, show entire surface in base color
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Softer opacity for fallback
'color': base_color
}, {'chain': self.chain_id})
# Add colored surfaces for specific categories
for residues, color in [
(true_positives, true_positive_color),
(false_positives, false_positive_color),
(false_negatives, true_epitope_color)
]:
if residues:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity, # Full opacity for clear colors
'color': color
}, {'chain': self.chain_id, 'resi': list(residues)})
# Show top regions with different colors if requested
if show_shape and 'top_k_regions' in predict_results:
top_regions = predict_results['top_k_regions']
self._add_multi_shape_visualization(
view, top_regions, radius, max_spheres,
True, 0.5, 0.2, True
)
def _add_probability_visualization(self, view, style, predict_results,
base_color, colormap, show_surface, surface_opacity, threshold,
region_index, radius, show_shape, shape_opacity,
show_center, center_radius, wireframe, coverage_color, center_color):
"""
Add visualization based on prediction probabilities with enhanced support for
specific region selection and surface rendering.
Args:
view: py3Dmol view object
style (str): Protein representation style
predict_results (dict): Results from predict() function
base_color (str): Base color for non-highlighted residues
colormap (str): Colormap name for probability visualization
show_surface (bool): Whether to show surface
surface_opacity (float): Surface opacity
threshold (float): Probability threshold for coloring
region_index (int): Index of specific region to show (0-based), None for all
Each region_index uses a distinct color for shape visualization
radius (float): Radius for spherical regions
show_shape (bool): Whether to show spherical shapes
shape_opacity (float): Shape opacity
show_center (bool): Whether to show center points
center_radius (float): Center point radius
wireframe (bool): Whether to show wireframe spheres
coverage_color (str): Color for coverage regions (not used when region_index is specified)
center_color (str): Color for center points
"""
import matplotlib.pyplot as plt
import matplotlib.colors as mcolors
# Get probability predictions and top-k regions
predictions = predict_results.get('predictions', {})
top_k_regions = predict_results.get('top_k_regions', [])
# Get style configuration
style_dict = {
'cartoon': {'cartoon': {}},
'stick': {'stick': {}},
'sphere': {'sphere': {}},
'surface': {'surface': {}}
}
base_style = style_dict.get(style, {'cartoon': {}})
if not predictions:
# Fallback to basic visualization
view.setStyle({'chain': self.chain_id}, {**base_style,
list(base_style.keys())[0]: {**list(base_style.values())[0], 'color': base_color}})
if show_surface:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Softer opacity for fallback
'color': base_color
}, {'chain': self.chain_id})
return
# Set base style for entire protein
view.setStyle({'chain': self.chain_id}, {**base_style,
list(base_style.keys())[0]: {**list(base_style.values())[0], 'color': base_color}})
# Determine which residues to color based on region_index
target_residues = {} # residue_num -> probability
selected_region = None
if region_index is not None and 0 <= region_index < len(top_k_regions):
# Show only the selected region
selected_region = top_k_regions[region_index]
covered_residues = selected_region.get('covered_residues', [])
# Get probabilities for residues in the selected region
for res_num in covered_residues:
if res_num in predictions:
target_residues[res_num] = predictions[res_num]
else:
# Show all residues with probabilities above threshold
target_residues = {res: prob for res, prob in predictions.items()
if prob >= threshold}
if not target_residues:
# No residues to color, show base style with surface
if show_surface:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Softer opacity for fallback
'color': base_color
}, {'chain': self.chain_id})
return
# Normalize probabilities for selected residues
probs = list(target_residues.values())
min_prob, max_prob = min(probs), max(probs)
# Enhanced color scheme for better visibility on surface
if colormap in ['RdYlBu_r', 'coolwarm', 'RdBu_r']:
# Use predefined soft color scheme for better visual comfort
probability_colors = [
'#c6dbef', # Very light blue (low probability)
'#9ecae1', # Light blue
'#6baed6', # Medium light blue
'#4292c6', # Medium blue
'#2171b5', # Medium blue
'#fcbba1', # Light orange
'#fc9272', # Medium light orange
'#fb6a4a', # Medium orange
'#ef3b2c', # Medium red
'#cb181d' # Medium red (high probability)
]
n_colors = len(probability_colors)
else:
# Use matplotlib colormap with reduced intensity
cmap = plt.cm.get_cmap(colormap)
probability_colors = []
n_colors = 10
for i in range(n_colors):
color_rgba = cmap(i / (n_colors - 1))
# Soften the colors by blending with white (0.3 factor)
softened_rgba = [
color_rgba[0] * 0.7 + 0.3, # Red channel
color_rgba[1] * 0.7 + 0.3, # Green channel
color_rgba[2] * 0.7 + 0.3, # Blue channel
]
# Ensure values don't exceed 1.0
softened_rgba = [min(1.0, val) for val in softened_rgba]
probability_colors.append(mcolors.rgb2hex(softened_rgba))
# Color residues based on normalized probability
colored_residues = []
for residue_num, prob in target_residues.items():
# Normalize probability to [0, 1] within the selected range
if max_prob > min_prob:
norm_prob = (prob - min_prob) / (max_prob - min_prob)
else:
norm_prob = 0.5
# Map to color index
color_idx = int(norm_prob * (n_colors - 1))
color_idx = max(0, min(color_idx, n_colors - 1))
color = probability_colors[color_idx]
# Add style for this residue with vivid color
style_name = list(base_style.keys())[0]
colored_style = {style_name: {'color': color}}
view.addStyle(
{'chain': self.chain_id, 'resi': residue_num},
colored_style
)
colored_residues.append(residue_num)
# Add surface rendering with improved visibility
if show_surface:
# Add base surface for non-colored regions
all_residues = set(int(res) for res in self.residue_index)
non_colored_residues = all_residues - set(colored_residues)
if non_colored_residues:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Further reduced opacity for softer appearance
'color': base_color
}, {'chain': self.chain_id, 'resi': list(non_colored_residues)})
# Add colored surfaces for probability residues with enhanced visibility
for residue_num, prob in target_residues.items():
# Normalize probability to [0, 1] within the selected range
if max_prob > min_prob:
norm_prob = (prob - min_prob) / (max_prob - min_prob)
else:
norm_prob = 0.5
# Map to color index
color_idx = int(norm_prob * (n_colors - 1))
color_idx = max(0, min(color_idx, n_colors - 1))
color = probability_colors[color_idx]
# Add surface for this residue with softer opacity for gentler visualization
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Slightly reduced opacity for softer colors
'color': color
}, {'chain': self.chain_id, 'resi': residue_num})
# Add spherical region visualization if specific region is selected
if selected_region is not None and show_shape:
center_res = selected_region['center_residue']
# Use radius from prediction results or provided radius
sphere_radius = radius or 19.0
# Define distinct colors for different regions
region_colors = [
'#FF6B6B', # Soft red
'#4ECDC4', # Soft teal
'#45B7D1', # Soft blue
'#96CEB4', # Soft green
'#FFEAA7', # Soft yellow
'#DDA0DD', # Soft plum
'#87CEEB', # Sky blue
'#F0E68C', # Soft khaki
'#FFB6C1', # Light pink
'#98FB98', # Pale green
'#9C6ADE', # Soft purple
'#FF9A8B' # Soft coral
]
# Select color based on region_index
shape_color = region_colors[region_index % len(region_colors)]
# Add sphere for the selected region with softer appearance and region-specific color
self._add_shape_visualization(
view, center_res, sphere_radius,
shape_color, center_color, # Use region-specific color for shape
show_center, center_radius,
shape_opacity * 0.6, # Reduced shape opacity for softer appearance
wireframe
)
# Highlight center residue with softer color matching the region
view.addStyle(
{'chain': self.chain_id, 'resi': center_res},
{list(base_style.keys())[0]: {'color': shape_color}} # Use region-specific color
)
def _add_top_regions_visualization(self, view, style, predict_results,
base_color, coverage_color, center_color,
show_surface, show_shape, show_center,
surface_opacity, shape_opacity, center_radius,
wireframe, radius, max_spheres):
"""Add visualization for top-k regions"""
# Set base style
view.setStyle({'chain': self.chain_id}, {style: {'color': base_color}})
# Get top regions
top_regions = predict_results.get('top_k_regions', [])
# Limit number of regions if max_spheres is specified
if max_spheres is not None:
top_regions = top_regions[:max_spheres]
# Enhanced color scheme for different regions
region_colors = [
'#FF6B6B', # Red
'#96CEB4', # Green
'#4ECDC4', # Teal
'#45B7D1', # Blue
'#FFEAA7', # Yellow
'#DDA0DD', # Plum
'#87CEEB', # Sky blue
'#F0E68C', # Khaki
'#FFB6C1', # Light pink
'#98FB98' # Pale green
]
for i, region in enumerate(top_regions):
center_res = region['center_residue']
covered_residues = region.get('covered_residues', [])
region_color = region_colors[i % len(region_colors)]
# Color covered residues
if covered_residues:
view.addStyle(
{'chain': self.chain_id, 'resi': covered_residues},
{style: {'color': region_color}}
)
# Add spherical region
if show_shape:
self._add_shape_visualization(
view, center_res, radius or 18.0,
region_color, center_color,
show_center, center_radius * 0.8,
shape_opacity, wireframe
)
# Add surface with balanced visibility
if show_surface:
# Base surface with good visibility
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Keep base surface visible
'color': base_color
})
# Colored surface for covered residues
for i, region in enumerate(top_regions):
covered_residues = region.get('covered_residues', [])
region_color = region_colors[i % len(region_colors)]
if covered_residues:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity, # Full opacity for clear coloring
'color': region_color
}, {'resi': covered_residues})
def _add_comparison_visualization(self, view, style, predict_results,
base_color, true_epitope_color, false_positive_color,
true_positive_color, coverage_color, show_surface, surface_opacity):
"""Add visualization comparing voted vs predicted epitopes"""
# Set base style
view.setStyle({'chain': self.chain_id}, {style: {'color': base_color}})
# Get different prediction sets
predicted_epitopes = set(predict_results.get('predicted_epitopes', []))
voted_epitopes = set(predict_results.get('voted_epitopes', []))
true_epitopes = set(predict_results.get('true_epitopes', []))
# Calculate overlaps
both_methods = predicted_epitopes & voted_epitopes # Agreed by both methods
only_predicted = predicted_epitopes - voted_epitopes # Only by probability
only_voted = voted_epitopes - predicted_epitopes # Only by voting
# Further categorize by true epitopes
both_correct = both_methods & true_epitopes
both_incorrect = both_methods - true_epitopes
only_pred_correct = only_predicted & true_epitopes
only_pred_incorrect = only_predicted - true_epitopes
only_vote_correct = only_voted & true_epitopes
only_vote_incorrect = only_voted - true_epitopes
# Assign colors and styles
color_assignments = [
(both_correct, '#00FF00'), # Bright green: both correct
(both_incorrect, '#FF0000'), # Red: both wrong
(only_pred_correct, '#90EE90'), # Light green: only predicted correct
(only_pred_incorrect, '#FFB6C1'), # Light red: only predicted wrong
(only_vote_correct, '#87CEEB'), # Sky blue: only voted correct
(only_vote_incorrect, '#DDA0DD') # Plum: only voted wrong
]
for residues, color in color_assignments:
if residues:
view.addStyle(
{'chain': self.chain_id, 'resi': list(residues)},
{style: {'color': color}}
)
# Add surface
if show_surface:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity,
'color': base_color
})
for residues, color in color_assignments:
if residues:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity,
'color': color
}, {'resi': list(residues)})
def _create_base_view(self, width: int, height: int) -> py3Dmol.view:
"""创建基本的py3Dmol视图并添加蛋白质结构"""
view = py3Dmol.view(width=width, height=height)
# 构建PDB字符串
pdb_str = "MODEL 1\n"
atom_num = 1
for res_idx in range(len(self.sequence)):
one_letter = self.sequence[res_idx]
resname = self.convert_letter_1to3(one_letter)
resnum = self.residue_index[res_idx]
mask = self.atom37_mask[res_idx]
coords = self.atom37_positions[res_idx][mask]
atoms = [name for name, exists in zip(RC.atom_types, mask) if exists]
for atom_name, coord in zip(atoms, coords):
x, y, z = coord
pdb_str += (f"ATOM {atom_num:5d} {atom_name:<3s} {resname:>3s} {self.chain_id:1s}{resnum:4d}"
f" {x:8.3f}{y:8.3f}{z:8.3f} 1.00 0.00\n")
atom_num += 1
pdb_str += "ENDMDL\n"
view.addModel(pdb_str, "pdb")
return view
def _add_epitope_visualization(self, view, style, predicted_epitopes,
base_color, true_epitope_color, false_positive_color, true_positive_color, coverage_color,
show_surface, surface_opacity, show_coverage,
center_res=None, radius=None):
"""添加表位可视化"""
# 设置基础颜色
view.setStyle({'chain': self.chain_id}, {style: {'color': base_color}})
true_epitopes = set(self.get_epitope_residue_numbers())
true_positives = set(predicted_epitopes) & true_epitopes
false_positives = set(predicted_epitopes) - true_epitopes
false_negatives = true_epitopes - set(predicted_epitopes)
# 如果提供了center_res和radius,获取覆盖区域
covered_residues = []
if center_res is not None and radius is not None:
coverage_dict, _, _ = self.get_surface_coverage(
radius=radius, threshold=0.25, index=False # Use residue numbers for visualization
)
covered_res_list = coverage_dict.get(center_res, [[], [], 0, 0])[0]
covered_residues = covered_res_list
# 计算覆盖区域内的True Negative (不是表位也没被预测为表位)
if covered_residues:
true_negatives = [res for res in covered_residues
if res not in true_epitopes and res not in predicted_epitopes]
# 为True Negative添加特殊颜色 (使用更明显的灰色)
true_negative_color = '#888888' # 更深的灰色
if true_negatives:
view.addStyle(
{'chain': self.chain_id, 'resi': true_negatives},
{style: {'color': true_negative_color}}
)
# 添加样式 - 增加颜色的饱和度
for residues, color in [
(true_positives, true_positive_color),
(false_positives, false_positive_color),
(false_negatives, true_epitope_color)
]:
if residues:
view.addStyle(
{'chain': self.chain_id, 'resi': list(residues)},
{style: {'color': color}}
)
# 先添加基础表面
if show_surface:
# Base surface with good visibility for overall structure
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Keep base surface visible
'color': base_color
})
# Colored surfaces for specific categories overlay on base surface
for residues, color in [
(true_positives, true_positive_color),
(false_positives, false_positive_color),
(false_negatives, true_epitope_color)
]:
if residues:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity, # Full opacity for clear coloring
'color': color
}, {'resi': list(residues)})
# 为覆盖区域内的True Negative添加表面
if center_res is not None and radius is not None and covered_residues and show_coverage:
true_negatives = [res for res in covered_residues
if res not in true_epitopes and res not in predicted_epitopes]
if true_negatives:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity,
'color': coverage_color
}, {'resi': true_negatives})
def _add_shape_visualization(self, view, center_res, radius,
coverage_color, center_color,
show_center, center_radius,
shape_opacity, wireframe):
"""添加球形可视化"""
center_idx = self.resnum_to_index.get(center_res)
if center_idx is None:
return
ca_idx = RC.atom_order["CA"] # Get CA atom index from atom order
center_coord = self.atom37_positions[center_idx, ca_idx, :]
# 添加球形
sphere_params = {
'center': {'x': float(center_coord[0]),
'y': float(center_coord[1]),
'z': float(center_coord[2])},
'radius': float(radius),
'color': coverage_color
}
if wireframe:
sphere_params.update({'wireframe': True, 'linewidth': 1.5}) # 增加线宽
else:
sphere_params.update({'opacity': shape_opacity})
view.addSphere(sphere_params)
# 添加中心点标记
if show_center:
view.addSphere({
'center': {'x': float(center_coord[0]),
'y': float(center_coord[1]),
'z': float(center_coord[2])},
'radius': float(center_radius),
'color': center_color,
'opacity': 1.0
})
def _add_coverage_visualization(self, view, style, center_res, radius,
base_color, coverage_color, true_positive_color, true_epitope_color,
show_surface, show_shape, show_center,
surface_opacity, shape_opacity, center_radius,
n_points, center_color, wireframe, show_epitope):
"""添加覆盖区域可视化"""
# 首先设置基础样式和颜色
view.setStyle({'chain': self.chain_id}, {style: {'color': base_color}})
# 获取覆盖区域
coverage_dict, _, _ = self.get_surface_coverage(
radius=radius, threshold=0.25, index=False # Use residue numbers for visualization
)
covered_res_list = coverage_dict.get(center_res, [[], [], 0, 0])[0]
covered_residues = covered_res_list
if show_epitope:
true_epitopes = set(self.get_epitope_residue_numbers())
else:
true_epitopes = set()
# 计算不同类别的残基
true_positives = set(covered_residues) & true_epitopes # 被覆盖的表位
false_negatives = true_epitopes - set(covered_residues) # 未被覆盖的表位
covered_non_epitopes = set(covered_residues) - true_epitopes # 被覆盖的非表位
# 添加表面渲染
if show_surface:
# 1. 添加基础表面,保持可见性
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 1.0, # Keep base surface visible
'color': base_color
})
# 2. 添加未被覆盖的表位表面
if false_negatives:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity, # Full opacity for clear coloring
'color': true_epitope_color
}, {'resi': list(false_negatives)})
# 3. 添加被覆盖的表位表面
if true_positives:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity, # Full opacity for clear coloring
'color': true_positive_color
}, {'resi': list(true_positives)})
# 4. 添加被覆盖的非表位表面
if covered_non_epitopes:
view.addSurface(py3Dmol.VDW, {
'opacity': surface_opacity * 0.9, # Slightly reduced for distinction
'color': coverage_color
}, {'resi': list(covered_non_epitopes)})
# 添加样式
if false_negatives:
view.addStyle(
{'chain': self.chain_id, 'resi': list(false_negatives)},
{style: {'color': true_epitope_color}}
)
if true_positives:
view.addStyle(
{'chain': self.chain_id, 'resi': list(true_positives)},
{style: {'color': true_positive_color}}
)
if covered_non_epitopes:
view.addStyle(
{'chain': self.chain_id, 'resi': list(covered_non_epitopes)},
{style: {'color': coverage_color}}
)
# 为中心残基添加黄色样式
view.addStyle(
{'chain': self.chain_id, 'resi': center_res},
{style: {'color': '#FFD700'}} # 使用更鲜艳的黄色
)
# 添加形状
if show_shape:
self._add_shape_visualization(
view, center_res, radius,
coverage_color,
center_color,
show_center, center_radius,
shape_opacity, wireframe
)
def _add_multi_shape_visualization(self, view, regions_data, radius, max_spheres,
show_center, center_radius, shape_opacity, wireframe):
"""Add multiple spherical regions with different colors"""
if not regions_data:
return
# Limit number of spheres if max_spheres is specified
regions_to_show = regions_data[:max_spheres] if max_spheres else regions_data
# Enhanced color scheme for different regions
sphere_colors = [
'#d671f1', # Plum
'#7190f1',
'#FF6B6B', # Red
'#96CEB4', # Green
'#FFEAA7', # Yellow
'#FFB6C1', # Light pink
'#4ECDC4', # Teal
'#87CEEB', # Sky blue
'#F0E68C', # Khaki
'#98FB98', # Pale green,
'#45B7D1' # Blue
]
for i, region_data in enumerate(regions_to_show):
if isinstance(region_data, dict):
# For prediction/evaluation results format
center_res = region_data['center_residue']
else:
# For simple center residue format
center_res = region_data
sphere_color = sphere_colors[i % len(sphere_colors)]
self._add_shape_visualization(
view, center_res, radius or 18.0,
sphere_color, '#FFD700', # Gold for center
show_center, center_radius, shape_opacity, wireframe
)
@classmethod
def from_pdb(
cls,
path: Optional[PathOrBuffer] = None,
chain_id: str = "detect",
id: Optional[str] = None,
is_predicted: bool = False,
) -> "AntigenChain":
"""
Return a AntigenChain object from a pdb file.
If `path` is not provided, the function will try multiple possible paths:
1. {id}_{chain_id}.pdb
2. {id}.pdb
3. {id.lower()}_{chain_id}.pdb
4. {id.upper()}_{chain_id}.pdb
If none of these paths exist, it will download the structure from RCSB PDB
and save it to the antigen_structs directory.
Args:
path (Optional[PathOrBuffer]): Path or buffer to read pdb file from. If None,
the default path is constructed from DATA_DIR.
chain_id (str, optional): Select a chain corresponding to (author) chain id.
"detect" uses the first detected chain.
id (Optional[str], optional): Protein identifier (pdb_id). If not provided and `path`
is given, the id will be inferred from the file name.
is_predicted (bool, optional): If True, reads b factor as the confidence readout.
Returns:
AntigenChain: The constructed antigen chain.
"""
# If no path is provided, try multiple possible paths
id = id.lower()
if path is None:
if id is None:
raise ValueError("Either 'path' or 'id' must be provided to locate the pdb file.")
# Try multiple possible paths
possible_paths = [
Path(BASE_DIR) / "data" / "antigen_structs" / f"{id}_{chain_id}.pdb",
Path(BASE_DIR) / "data" / "antigen_structs" / f"{id}.pdb",
# Path(BASE_DIR) / "data" / "antigen_structs" / f"{id.lower()}_{chain_id}.pdb",
# Path(BASE_DIR) / "data" / "antigen_structs" / f"{id.upper()}_{chain_id}.pdb",
# Path(BASE_DIR) / "data" / "pdb" / f"{id.lower()}.pdb",
# Path(BASE_DIR) / "data" / "pdb" / f"{id.upper()}.pdb",
]
# Try each path
path = None
for p in possible_paths:
if p.exists():
path = p
print(f"Found pdb file at {path}")
break
# If no path exists, download from RCSB
if path is None:
try:
# Create directory if it doesn't exist
save_dir = Path(BASE_DIR) / "data" / "pdb"
save_dir.mkdir(parents=True, exist_ok=True)
# Download from RCSB
rcsb.fetch(id, "pdb", target_path=save_dir)
path = save_dir / f"{id}.pdb"
print(f"No existing pdb file for {id}_{chain_id}, downloaded {id} complex pdb file to {path}")
except Exception as e:
print(f"[ERROR] Failed to download pdb file for {id}: {str(e)}")
return None
else:
path = Path(path) # Ensure path is a Path object
# Determine the file_id from the provided id or from the path.
if id is not None:
file_id = id
else:
# Infer file_id from the file name if id is not provided.
file_id = path.with_suffix("").name
# Read the pdb file.
try:
atom_array = PDBFile.read(path).get_structure(model=1, extra_fields=["b_factor"])
except Exception as e:
print(f"[ERROR] Failed to read pdb file {path}: {str(e)}")
return None
# If chain_id is "detect", use the first detected chain.
if chain_id == "detect":
chain_id = atom_array.chain_id[0]
print(f"[WARNING] No chain_id provided, using the first detected chain: {chain_id}")
# Filter the AtomArray: amino acids, non-hetero atoms, and matching chain.
atom_array = atom_array[
bs.filter_amino_acids(atom_array)
& ~atom_array.hetero
& (atom_array.chain_id == chain_id)
]
# Set entity_id as 1 (not supplied in PDB files)
entity_id = 1
# Build the sequence by converting three-letter codes to one-letter codes.
sequence = "".join(
(
r if len((r := PDBData.protein_letters_3to1.get(monomer[0].res_name, "X"))) == 1 else "X"
)
for monomer in bs.residue_iter(atom_array)
)
num_res = len(sequence)
# Prepare arrays for atom coordinates, mask, residue indices, etc.
atom_positions = np.full([num_res, RC.atom_type_num, 3], np.nan, dtype=np.float32)
atom_mask = np.full([num_res, RC.atom_type_num], False, dtype=bool)
residue_index = np.full([num_res], -1, dtype=np.int64)
insertion_code = np.full([num_res], "", dtype="<U4")
confidence = np.ones([num_res], dtype=np.float32)
# Populate arrays from the pdb data.
for i, res in enumerate(bs.residue_iter(atom_array)):
for atom in res:
atom_name = atom.atom_name
if atom_name == "SE" and atom.res_name == "MSE":
atom_name = "SD"
if atom_name in RC.atom_order:
atom_positions[i, RC.atom_order[atom_name]] = atom.coord
atom_mask[i, RC.atom_order[atom_name]] = True
if is_predicted and atom_name == "CA":
confidence[i] = atom.b_factor
residue_index[i] = res[0].res_id
insertion_code[i] = res[0].ins_code
# Ensure that sequence is valid.
assert all(sequence), "Some residue name was not specified correctly"
return cls(
id=file_id,
sequence=sequence,
chain_id=chain_id,
entity_id=entity_id,
atom37_positions=atom_positions,
atom37_mask=atom_mask,
residue_index=residue_index,
insertion_code=insertion_code,
confidence=confidence,
) |