Spaces:
Running
Running
File size: 13,373 Bytes
cb6f1ba | 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 | """
Smoke test for ProteinRelocalizer using multiple source->target pairs.
The script tries candidate shifts:
- Cytoplasm -> Extracellular
- Cytoplasm -> Membrane
- Cytoplasm -> Nucleus
For each pair, it picks a protein in a configurable source-probability band and
low target probability, then runs a short probe optimization to estimate movement.
It chooses the pair with the clearest target-probability gain and runs the full
demo optimization.
"""
from __future__ import annotations
import argparse
import sys
from pathlib import Path
from typing import Any, Dict, List, Tuple
import pandas as pd
ROOT = Path(__file__).resolve().parent.parent
if str(ROOT) not in sys.path:
sys.path.insert(0, str(ROOT))
class _QuietTqdm:
"""Disable progress bar noise for smoke tests."""
def __init__(self, iterable, *args, **kwargs):
self._it = iterable
def __iter__(self):
return iter(self._it)
def set_postfix(self, *args, **kwargs) -> None:
pass
def close(self) -> None:
pass
import tqdm as _tqdm_mod # noqa: E402
_tqdm_mod.tqdm = _QuietTqdm # noqa: E402
from src.design.relocalizer import ProteinRelocalizer # noqa: E402
def _pick_test_protein_for_pair(
df: pd.DataFrame,
relocalizer: ProteinRelocalizer,
*,
source: str,
target: str,
min_seq_len: int,
max_seq_len: int,
source_prob_min: float,
source_prob_max: float,
target_prob_max: float,
max_scan: int,
) -> Tuple[str, str, Dict[str, Any]]:
"""
From CSV rows labeled by `source`, pick a sequence in [min_seq_len, max_seq_len]
whose classifier probabilities satisfy:
source_prob_min <= P(source) <= source_prob_max
P(target) < target_prob_max
Rows are ordered **longest first** (within the length band).
"""
required = {"ACC", "Sequence", source, target}
missing = required.difference(df.columns)
if missing:
raise ValueError(f"Dataset missing columns: {sorted(missing)}")
pool = df[df[source] == 1].copy()
if pool.empty:
raise ValueError(f"No rows with {source} label == 1 in CSV.")
pool["seq_len"] = pool["Sequence"].astype(str).str.len()
pool = pool[(pool["seq_len"] >= min_seq_len) & (pool["seq_len"] <= max_seq_len)]
pool = pool.sort_values("seq_len", ascending=False)
if pool.empty:
raise ValueError(
f"No Cytoplasm-positive rows with {min_seq_len} <= length <= {max_seq_len}."
)
tried = 0
for _, row in pool.iterrows():
if tried >= max_scan:
break
tried += 1
seq = str(row["Sequence"]).upper().strip()
if len(seq) < min_seq_len or len(seq) > max_seq_len:
continue
scored = relocalizer.score_variant(seq)
probs = scored["localization_probs"]
p_src = float(probs.get(source, 0.0))
p_tgt = float(probs.get(target, 0.0))
if source_prob_min <= p_src <= source_prob_max and p_tgt < target_prob_max:
return str(row["ACC"]), seq, scored
raise RuntimeError(
f"No sequence found after scoring {tried} {source}-labeled proteins "
f"(len in [{min_seq_len}, {max_seq_len}], need {source_prob_min}<=P({source})<={source_prob_max} "
f"and P({target})<{target_prob_max}). "
"Try increasing --max-scan or relaxing probability/length bounds."
)
def _print_trajectory(traj: list, source: str, target: str) -> None:
print("\n--- Optimization trajectory ---")
if not traj:
print("(empty)")
return
for step in traj:
it = step.get("iteration", "?")
p_tgt = step.get("best_target_prob")
p_src = step.get("best_source_prob")
nmut = step.get("num_mutations")
neval = step.get("num_candidates_evaluated")
err = step.get("error")
parts = [f"iter={it}"]
if p_tgt is not None:
parts.append(f"P({target})={p_tgt:.4f}")
if p_src is not None:
parts.append(f"P({source})={p_src:.4f}")
if nmut is not None:
parts.append(f"muts={nmut}")
if neval is not None:
parts.append(f"evaluated={neval}")
if err:
parts.append(f"error={err}")
print(" " + " | ".join(parts))
def _print_top_candidate(
results: Dict[str, Any],
source: str,
target: str,
) -> None:
tops = results.get("top_candidates") or []
print("\n--- Top candidate ---")
if not tops:
print("(none)")
return
best = tops[0]
orig = results["original_scores"]["localization_probs"]
newp = best["localization_probs"]
print(f"Composite score: {best['composite_score']:.4f}")
print(f"Plausibility: {best['plausibility_score']:.4f}")
print("")
print("Probability changes (original -> candidate):")
for loc in (source, target):
o = float(orig.get(loc, 0.0))
n = float(newp.get(loc, 0.0))
print(f" P({loc}): {o:.4f} -> {n:.4f} (delta {n - o:+.4f})")
muts = best.get("mutations") or []
print("")
print(f"Mutations vs original ({len(muts)}):")
if muts:
shown = muts[:40]
line = ", ".join(f"{p}{a}>{b}" for p, a, b in shown)
if len(muts) > 40:
line += ", ..."
print(f" {line}")
else:
print(" (none; same as original)")
def _print_all_location_changes(results: Dict[str, Any]) -> None:
print("\n--- Full localization profile shift (all labels) ---")
tops = results.get("top_candidates") or []
if not tops:
print("(no top candidate)")
return
orig = results.get("original_scores", {}).get("localization_probs", {})
best = tops[0].get("localization_probs", {})
labels = sorted(set(orig.keys()) | set(best.keys()))
for label in labels:
o = float(orig.get(label, 0.0))
n = float(best.get(label, 0.0))
print(f" {label:24s} {o:.4f} -> {n:.4f} (delta {n - o:+.4f})")
def _movement_score(results: Dict[str, Any], source: str, target: str) -> float:
"""Positive score means shift toward target and away from source."""
orig = results.get("original_scores", {}).get("localization_probs", {})
tops = results.get("top_candidates") or []
if not tops:
return float("-inf")
best = tops[0].get("localization_probs", {})
delta_t = float(best.get(target, 0.0)) - float(orig.get(target, 0.0))
delta_s = float(orig.get(source, 0.0)) - float(best.get(source, 0.0))
return delta_t + 0.5 * delta_s
def main() -> None:
try:
import captum # noqa: F401
except ImportError:
print(
"ERROR: captum is required for relocalize() (integrated gradients). "
"Install with: pip install captum",
file=sys.stderr,
)
sys.exit(2)
p = argparse.ArgumentParser(description="Smoke test ProteinRelocalizer (lightweight, fast).")
p.add_argument(
"--classifier-path",
type=Path,
default=ROOT / "models" / "best_model.pt",
help="Trained classifier checkpoint.",
)
p.add_argument(
"--csv-path",
type=Path,
default=ROOT / "data" / "processed" / "deeploc_multilabel.csv",
help="Multilabel CSV with ACC, Sequence, Cytoplasm, ...",
)
p.add_argument("--device", default=None, help="cuda | cpu (default: auto)")
p.add_argument(
"--min-seq-len",
type=int,
default=150,
help="Minimum sequence length for the test protein (default: 150).",
)
p.add_argument(
"--max-seq-len",
type=int,
default=350,
help="Maximum sequence length for the test protein (default: 350).",
)
p.add_argument(
"--source-prob-min",
type=float,
default=0.5,
help="Minimum P(source) for selected test protein (default: 0.5).",
)
p.add_argument(
"--source-prob-max",
type=float,
default=0.9,
help="Maximum P(source) for selected test protein (default: 0.9).",
)
p.add_argument(
"--target-prob-max",
type=float,
default=0.25,
help="Require P(target) below this for selected protein (default: 0.25).",
)
p.add_argument(
"--max-scan",
type=int,
default=500,
help="Max CSV rows (after length filter) to score when searching for a match.",
)
p.add_argument(
"--probe-iterations",
type=int,
default=3,
help="Short probe iterations for pair selection (default: 3).",
)
p.add_argument(
"--probe-candidates",
type=int,
default=10,
help="Short probe candidates/iter for pair selection (default: 10).",
)
args = p.parse_args()
classifier_path = args.classifier_path
if not classifier_path.is_absolute():
classifier_path = (ROOT / classifier_path).resolve()
csv_path = args.csv_path
if not csv_path.is_absolute():
csv_path = (ROOT / csv_path).resolve()
if not classifier_path.is_file():
print(f"ERROR: missing classifier: {classifier_path}", file=sys.stderr)
sys.exit(1)
if not csv_path.is_file():
print(f"ERROR: missing dataset CSV: {csv_path}", file=sys.stderr)
sys.exit(1)
print("Loading ProteinRelocalizer.from_lightweight() (t33 encoder + t12 MLM)...", flush=True)
device = args.device
relocalizer = ProteinRelocalizer.from_lightweight(
classifier_path=classifier_path,
device=device,
)
if args.min_seq_len > args.max_seq_len:
print("ERROR: --min-seq-len must be <= --max-seq-len", file=sys.stderr)
sys.exit(1)
if args.source_prob_min > args.source_prob_max:
print("ERROR: --source-prob-min must be <= --source-prob-max", file=sys.stderr)
sys.exit(1)
df = pd.read_csv(csv_path)
pair_candidates: List[Tuple[str, str]] = [
("Cytoplasm", "Extracellular"),
("Cytoplasm", "Membrane"),
("Cytoplasm", "Nucleus"),
]
pair_trials: List[Dict[str, Any]] = []
print("\nTrying source->target pairs to find a clear demo shift...", flush=True)
for source, target in pair_candidates:
print(f"\n[Pair probe] {source} -> {target}", flush=True)
try:
acc, sequence, initial = _pick_test_protein_for_pair(
df,
relocalizer,
source=source,
target=target,
min_seq_len=args.min_seq_len,
max_seq_len=args.max_seq_len,
source_prob_min=args.source_prob_min,
source_prob_max=args.source_prob_max,
target_prob_max=args.target_prob_max,
max_scan=args.max_scan,
)
probe = relocalizer.relocalize(
sequence,
source_location=source,
target_location=target,
n_iterations=args.probe_iterations,
candidates_per_iteration=args.probe_candidates,
)
move = _movement_score(probe, source, target)
print(
f" ACC={acc} len={len(sequence)} "
f"P({source})={initial['localization_probs'].get(source, 0):.4f} "
f"P({target})={initial['localization_probs'].get(target, 0):.4f} "
f"probe_movement={move:+.4f}",
flush=True,
)
pair_trials.append(
{
"source": source,
"target": target,
"acc": acc,
"sequence": sequence,
"initial": initial,
"probe_results": probe,
"movement": move,
}
)
except Exception as ex:
print(f" skipped: {ex}", flush=True)
if not pair_trials:
raise RuntimeError("No valid pair/test-protein found for the configured constraints.")
pair_trials.sort(key=lambda x: float(x["movement"]), reverse=True)
chosen = pair_trials[0]
source = str(chosen["source"])
target = str(chosen["target"])
acc = str(chosen["acc"])
sequence = str(chosen["sequence"])
initial = chosen["initial"]
print(
f"\nSelected demo pair: {source} -> {target} "
f"(probe movement {float(chosen['movement']):+.4f})",
flush=True,
)
print(
f"Selected ACC={acc}, length={len(sequence)} "
f"P({source})={initial['localization_probs'].get(source, 0):.4f} "
f"P({target})={initial['localization_probs'].get(target, 0):.4f}",
flush=True,
)
print("\nRunning full relocalize (10 iterations, 20 candidates/iter)...", flush=True)
results = relocalizer.relocalize(
sequence,
source_location=source,
target_location=target,
n_iterations=10,
candidates_per_iteration=20,
)
_print_trajectory(results.get("optimization_trajectory") or [], source, target)
_print_top_candidate(results, source, target)
_print_all_location_changes(results)
print("\n--- Summary ---")
print(relocalizer.get_summary(results))
print(f"\nTotal time: {results.get('total_time_seconds', 0):.2f}s "
f"variants scored: {results.get('total_variants_evaluated', 0)}")
if __name__ == "__main__":
main()
|