Datasets:
File size: 3,598 Bytes
7bf35b0 1c12c1a 7bf35b0 1c12c1a 7bf35b0 1c12c1a 7bf35b0 1c12c1a 7bf35b0 1c12c1a | 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 | """Does our label pipeline reproduce NetSurfP-2.0's TRAINING labels (TAPE's secondary_structure_train.json)?
1. sample 200 training chains (seed 0), download their current PDB entries from RCSB
2. mkdssp on each chain alone, polyproline (P) counted as coil, TAPE's 8->3 mapping
3. compare with the training labels on residues NetSurfP marks as resolved (disorder == 1)
"""
import os, sys, json, random, collections, urllib.request, urllib.error
from concurrent.futures import ThreadPoolExecutor
sys.path.insert(0, "scripts")
import ijson
import ss_assign as A
from compare_ss_tools_helpers import project
S8, S3 = "GHIBESTC", "HEC"
TRAIN = "data/original/tape/secondary_structure/secondary_structure/secondary_structure_train.json" # TAPE's download, extracted
PDB_DIR = "data/original/netsurfp_train_pdb"
os.makedirs(PDB_DIR, exist_ok=True)
N = 8678
pick = set(random.Random(0).sample(range(N), 200))
recs = []
with open(TRAIN, "rb") as f:
for i, r in enumerate(ijson.items(f, "item", use_float=True)):
if i in pick:
recs.append({k: r[k] for k in ("id", "primary", "ss8", "ss3", "disorder", "valid_mask")})
print(f"sampled {len(recs)} training chains", flush=True)
def fetch(pdb):
path = f"{PDB_DIR}/{pdb}.cif.gz"
if os.path.exists(path):
return pdb, "ok"
try:
urllib.request.urlretrieve(f"https://files.rcsb.org/download/{pdb}.cif.gz", path)
return pdb, "ok"
except urllib.error.HTTPError as e:
return pdb, f"HTTP {e.code}"
with ThreadPoolExecutor(8) as ex:
status = dict(ex.map(fetch, sorted({r["id"].split("-")[0] for r in recs})))
print("downloads:", collections.Counter(status.values()), flush=True)
tot = collections.Counter()
per = []
for r in recs:
pdb, ch = r["id"].split("-", 1)
if status[pdb] != "ok":
per.append((r["id"], None, f"entry not downloadable ({status[pdb]})")); continue
try:
st = A.load(f"{PDB_DIR}/{pdb}.cif.gz")
seqres, lab = A.mkdssp_assign(st, ch, "mkdssp", "chain")
except Exception as e:
per.append((r["id"], None, f"failed: {type(e).__name__}: {str(e)[:60]}")); continue
lab, how = project(r["primary"], seqres, lab)
c = collections.Counter()
for i, d in enumerate(r["disorder"]):
mine = lab[i]
if d == 1 and mine is None:
c["resolved_in_netsurfp_missing_now"] += 1
if d == 1 and mine is not None:
l8 = "C" if mine == "P" else mine
c["n"] += 1; c["s8"] += l8 == S8[r["ss8"][i]]; c["s3"] += A.SS8_TO_SS3[l8] == S3[r["ss3"][i]]
tot.update(c)
per.append((r["id"], c, how))
ok = [p for p in per if p[1] is not None and p[1]["n"] > 0]
print(f"\nchains compared: {len(ok)} of {len(recs)}")
print(f"OVERALL on {tot['n']:,} residues: ss8 {100 * tot['s8'] / tot['n']:.2f}% ss3 {100 * tot['s3'] / tot['n']:.2f}%"
f" | residues resolved in NetSurfP but missing in today's entry: {tot['resolved_in_netsurfp_missing_now']}")
bins = collections.Counter()
for _, c, _ in ok:
a8 = c["s8"] / c["n"]
bins["100%" if a8 == 1 else ">=99%" if a8 >= .99 else ">=95%" if a8 >= .95 else "<95%"] += 1
print("chains by ss8 agreement:", dict(bins))
print("\nlowest-agreement chains:")
for pid, c, how in sorted(ok, key=lambda p: p[1]["s8"] / p[1]["n"])[:10]:
print(f" {pid:9s} ss8 {100 * c['s8'] / c['n']:6.2f}% on {c['n']} residues | sequence vs NetSurfP: {how}")
for pid, c, why in per:
if c is None:
print(f" {pid:9s} {why}")
json.dump({"ids": [r["id"] for r in recs], "total": tot}, open("reports/netsurfp_train_check.json", "w"), indent=1)
|