Deploy AILIXIR generation API Docker Space

.dockerignore

@@ -0,0 +1,18 @@
+outputs
+.venv
+.git
+__pycache__
+.pytest_cache
+.mypy_cache
+.cache
+bin
+
+*.tar
+*.tar.gz
+*.zip
+*.bak
+*.pyc
+test_*.csv
+demo_*.csv
+release_test_*.csv
+generated_results.csv

.gitattributes

@@ -0,0 +1,4 @@
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.chkpt filter=lfs diff=lfs merge=lfs -text
+*.prior filter=lfs diff=lfs merge=lfs -text
+*.map filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,17 @@
+outputs/
+.venv/
+__pycache__/
+.pytest_cache/
+.mypy_cache/
+.cache/
+bin/
+
+*.tar
+*.tar.gz
+*.zip
+*.bak
+*.pyc
+test_*.csv
+demo_*.csv
+release_test_*.csv
+generated_results.csv

Dockerfile

@@ -0,0 +1,142 @@
+FROM condaforge/miniforge3:24.11.3-0
+
+WORKDIR /app
+
+ENV DEBIAN_FRONTEND=noninteractive
+ENV PATH=/opt/conda/bin:$PATH
+ENV LD_LIBRARY_PATH=/opt/conda/envs/ailixir/lib:/opt/conda/envs/dp/lib:/opt/conda/lib:/usr/local/lib:/usr/lib/x86_64-linux-gnu
+
+SHELL ["/bin/bash", "-lc"]
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    git \
+    wget \
+    curl \
+    ca-certificates \
+    build-essential \
+    make \
+    procps \
+    ocl-icd-opencl-dev \
+    pocl-opencl-icd \
+    clinfo \
+    && rm -rf /var/lib/apt/lists/*
+
+# -----------------------------
+# Generation / REINVENT env
+# -----------------------------
+RUN conda create -n ailixir python=3.10 -y && conda clean -afy
+
+SHELL ["conda", "run", "--no-capture-output", "-n", "ailixir", "/bin/bash", "-lc"]
+
+RUN set -eux; \
+    for i in 1 2 3 4 5; do \
+      if conda install -y -c conda-forge \
+        pandas \
+        numpy \
+        scikit-learn \
+        rdkit \
+        fastapi \
+        uvicorn \
+        requests \
+        tqdm \
+        pip; then \
+        conda clean -afy; \
+        exit 0; \
+      fi; \
+      echo "conda install failed on attempt ${i}, retrying..."; \
+      conda clean -i -y || true; \
+      sleep $((i * 20)); \
+    done; \
+    echo "conda install failed after retries"; \
+    exit 1
+
+RUN git clone --depth 1 https://github.com/MolecularAI/REINVENT4.git /opt/REINVENT4 \
+    && cd /opt/REINVENT4 \
+    && python install.py cpu
+
+RUN python -c "from rdkit import Chem; print('RDKit import OK')"
+RUN python -c "import importlib.metadata as m; print('REINVENT version', m.version('reinvent'))"
+
+# -----------------------------
+# DeepPurpose affinity env
+# -----------------------------
+SHELL ["/bin/bash", "-lc"]
+
+RUN conda create -n dp python=3.8 -y && conda clean -afy
+
+SHELL ["conda", "run", "--no-capture-output", "-n", "dp", "/bin/bash", "-lc"]
+
+RUN set -eux; \
+    for i in 1 2 3 4 5; do \
+      if conda install -y -c conda-forge \
+        pandas \
+        numpy \
+        scikit-learn \
+        rdkit \
+        fastapi \
+        uvicorn \
+        requests \
+        tqdm \
+        pip; then \
+        conda clean -afy; \
+        exit 0; \
+      fi; \
+      echo "conda install failed on attempt ${i}, retrying..."; \
+      conda clean -i -y || true; \
+      sleep $((i * 20)); \
+    done; \
+    echo "conda install failed after retries"; \
+    exit 1
+
+ENV PIP_DEFAULT_TIMEOUT=1000
+ENV PIP_RETRIES=10
+ENV PIP_DISABLE_PIP_VERSION_CHECK=1
+
+RUN pip install --no-cache-dir --default-timeout=1000 --retries=10 \
+    git+https://github.com/bp-kelley/descriptastorus
+
+RUN pip install --no-cache-dir --default-timeout=1000 --retries=10 \
+    torch==2.1.0 \
+    --index-url https://download.pytorch.org/whl/cpu
+
+RUN python -c "import torch; print('Torch version:', torch.__version__); print('CUDA available:', torch.cuda.is_available())"
+
+RUN pip install --no-cache-dir --default-timeout=1000 --retries=10 \
+    DeepPurpose
+
+# -----------------------------
+# AutoDock-GPU CPU/OpenCL build
+# -----------------------------
+SHELL ["/bin/bash", "-lc"]
+
+RUN set -eux; \
+    clinfo || true; \
+    rm -rf /opt/AutoDock-GPU; \
+    git clone --depth 1 https://github.com/ccsb-scripps/AutoDock-GPU.git /opt/AutoDock-GPU; \
+    cd /opt/AutoDock-GPU; \
+    make DEVICE=CPU NUMWI=1 GPU_INCLUDE_PATH=/usr/include GPU_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu; \
+    find /opt/AutoDock-GPU -maxdepth 3 -type f -name "autodock_*" -print; \
+    test -f /opt/AutoDock-GPU/bin/autodock_cpu_1wi; \
+    chmod +x /opt/AutoDock-GPU/bin/autodock_cpu_1wi
+
+COPY api.py /app/api.py
+COPY bundle /app/bundle
+COPY start_hf.sh /app/start_hf.sh
+
+RUN sed -i 's/\r$//' /app/start_hf.sh \
+    && chmod +x /app/start_hf.sh \
+    && mkdir -p /app/outputs/jobs /app/outputs/deeppurpose \
+    && mkdir -p /home/abdullah/projects/egfr_drug_discovery/runs/deeppurpose
+
+
+ENV PUBLIC_BASE_URL=http://localhost:7860
+ENV REINVENT_DEVICE=cpu
+ENV DEEPPURPOSE_URL=http://127.0.0.1:8001/reinvent_predict
+ENV ADGPU_BIN=/opt/AutoDock-GPU/bin/autodock_cpu_1wi
+ENV PYTHON=python
+ENV LD_PRELOAD=/opt/conda/envs/ailixir/lib/libstdc++.so.6
+ENV PYTHONPATH=/app/bundle
+
+EXPOSE 7860
+
+CMD ["/app/start_hf.sh"]

README.md

@@ -0,0 +1,19 @@
+---
+title: AILIXIR EGFR Generation API
+emoji: 🧬
+colorFrom: blue
+colorTo: green
+sdk: docker
+app_port: 7860
+pinned: false
+---
+
+# AILIXIR EGFR Generation API
+
+FastAPI service for EGFR molecule generation, DeepPurpose affinity scoring, and optional CPU docking.
+
+Recommended defaults for Hugging Face free CPU:
+- docking_mode: off
+- dock_top_k: 1
+
+CPU docking is supported but slow.

api.py

@@ -0,0 +1,675 @@
+import os
+import json
+import time
+import uuid
+import shutil
+import subprocess
+from datetime import datetime
+from pathlib import Path
+from typing import Literal, Optional, List, Dict, Any
+
+import pandas as pd
+from fastapi import FastAPI, HTTPException, BackgroundTasks, Request
+from fastapi.responses import FileResponse, JSONResponse
+from fastapi.middleware.cors import CORSMiddleware
+from pydantic import BaseModel, Field
+
+from rdkit import Chem
+from rdkit.Chem import AllChem
+
+
+APP_ROOT = Path(__file__).resolve().parent
+BUNDLE_DIR = APP_ROOT / "bundle"
+OUTPUTS_DIR = APP_ROOT / "outputs"
+JOBS_DIR = OUTPUTS_DIR / "jobs"
+
+PUBLIC_BASE_URL = os.getenv("PUBLIC_BASE_URL", "http://localhost:8000").rstrip("/")
+DEEPPURPOSE_URL = os.getenv("DEEPPURPOSE_URL", "http://127.0.0.1:7860/reinvent_predict")
+REINVENT_DEVICE = os.getenv("REINVENT_DEVICE", "cpu")
+ADGPU_BIN = os.getenv("ADGPU_BIN", "")
+
+JOBS_DIR.mkdir(parents=True, exist_ok=True)
+
+# Attempt to load DeepPurpose affinity service (optional)
+try:
+    from bundle.services.deeppurpose.serve_affinity import predict_smiles_list
+    HAS_AFFINITY = True
+except Exception:
+    HAS_AFFINITY = False
+
+
+app = FastAPI(
+    title="Ailixir EGFR Pipeline API",
+    version="1.0.0",
+    description="Production API for EGFR generation, scoring, docking, and ligand export."
+)
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=os.getenv("CORS_ALLOW_ORIGINS", "*").split(","),
+    allow_credentials=False,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+
+# -----------------------------
+# Request models
+# -----------------------------
+
+class GenerateRequest(BaseModel):
+    preset: str = Field(default="egfr_generator")
+    num_molecules: int = Field(default=100, ge=1, le=5000)
+    return_top_k: int = Field(default=20, ge=1, le=1000)
+    docking_mode: Literal["off", "top_k", "all"] = "off"
+    dock_top_k: int = Field(default=10, ge=1, le=1000)
+
+
+class LigandExportRequest(BaseModel):
+    smiles: str
+    format: Literal["pdb", "pdbqt", "mol2"]
+
+
+# -----------------------------
+# Helpers
+# -----------------------------
+
+def new_job_id(prefix: str) -> str:
+    stamp = datetime.utcnow().strftime("%Y%m%d_%H%M%S")
+    short = uuid.uuid4().hex[:6]
+    return f"{prefix}_{stamp}_{short}"
+
+
+def make_file_meta(job_id: str, file_path: Path) -> Dict[str, str]:
+    rel_path = file_path.relative_to(JOBS_DIR / job_id).as_posix()
+    relative_url = f"/files/jobs/{job_id}/{rel_path}"
+    return {
+        "filename": file_path.name,
+        "relative_url": relative_url,
+        "download_url": f"{PUBLIC_BASE_URL}{relative_url}",
+    }
+
+
+def clean_record(record: Dict[str, Any]) -> Dict[str, Any]:
+    """
+    Public response cleaner.
+    Removes internal paths and confusing technical fields.
+    Keeps molecule scores and user-facing values only.
+    """
+    drop_keys = {
+        "docking_pose_file",
+        "file_path",
+        "internal_path",
+        "pose_file",
+    }
+
+    out = {}
+    for k, v in record.items():
+        if k in drop_keys:
+            continue
+
+        # Convert NaN to None for valid JSON
+        try:
+            if pd.isna(v):
+                out[k] = None
+            else:
+                out[k] = v
+        except Exception:
+            out[k] = v
+
+    return out
+
+
+def dataframe_to_public_records(df: pd.DataFrame) -> List[Dict[str, Any]]:
+    records = []
+    for i, row in df.reset_index(drop=True).iterrows():
+        item = clean_record(row.to_dict())
+        item["rank"] = i + 1
+        records.append(item)
+    return records
+
+
+def run_cmd(cmd: List[str], cwd: Optional[Path] = None, timeout: Optional[int] = None):
+    result = subprocess.run(
+        cmd,
+        cwd=str(cwd) if cwd else None,
+        text=True,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.STDOUT,
+        timeout=timeout,
+        check=False,
+    )
+    return result
+
+
+def canonicalize_smiles(smiles: str):
+    mol = Chem.MolFromSmiles(smiles)
+    if mol is None:
+        raise HTTPException(status_code=400, detail="Invalid SMILES.")
+    canonical = Chem.MolToSmiles(mol, isomericSmiles=True)
+    return mol, canonical
+
+
+def make_3d_mol(smiles: str):
+    mol, canonical = canonicalize_smiles(smiles)
+    mol = Chem.AddHs(mol)
+
+    params = AllChem.ETKDGv3()
+    params.randomSeed = 42
+
+    status = AllChem.EmbedMolecule(mol, params)
+    if status != 0:
+        raise HTTPException(status_code=500, detail="3D embedding failed.")
+
+    try:
+        AllChem.UFFOptimizeMolecule(mol, maxIters=500)
+    except Exception:
+        pass
+
+    return mol, canonical
+
+
+def write_sdf_for_conversion(mol, sdf_path: Path):
+    writer = Chem.SDWriter(str(sdf_path))
+    writer.write(mol)
+    writer.close()
+
+
+def export_pdb(mol, out_path: Path):
+    Chem.MolToPDBFile(mol, str(out_path))
+
+
+def export_pdbqt(sdf_path: Path, out_path: Path):
+    mk_prepare = shutil.which("mk_prepare_ligand.py") or shutil.which("mk_prepare_ligand")
+    if mk_prepare is None:
+        raise HTTPException(
+            status_code=500,
+            detail="Meeko mk_prepare_ligand.py was not found in PATH. Required for PDBQT export."
+        )
+
+    result = run_cmd([
+        mk_prepare,
+        "-i", str(sdf_path),
+        "-o", str(out_path),
+    ])
+
+    if result.returncode != 0 or not out_path.exists():
+        raise HTTPException(
+            status_code=500,
+            detail=f"PDBQT export failed: {result.stdout[:1000]}"
+        )
+
+
+def export_mol2(sdf_path: Path, out_path: Path):
+    obabel = shutil.which("obabel")
+    if obabel is None:
+        raise HTTPException(
+            status_code=500,
+            detail="Open Babel obabel was not found in PATH. Required for MOL2 export."
+        )
+
+    result = run_cmd([
+        obabel,
+        "-isdf", str(sdf_path),
+        "-omol2",
+        "-O", str(out_path),
+    ])
+
+    if result.returncode != 0 or not out_path.exists():
+        raise HTTPException(
+            status_code=500,
+            detail=f"MOL2 export failed: {result.stdout[:1000]}"
+        )
+
+
+def count_docked(df: pd.DataFrame) -> int:
+    if "docking_status" not in df.columns:
+        return 0
+    return int((df["docking_status"] == "completed").sum())
+
+def job_status_path(job_id: str) -> Path:
+    return JOBS_DIR / job_id / "job_status.json"
+
+
+def make_api_url(path: str) -> str:
+    if not path.startswith("/"):
+        path = "/" + path
+    return f"{PUBLIC_BASE_URL}{path}"
+
+
+def write_job_status(job_id: str, payload: Dict[str, Any]):
+    job_dir = JOBS_DIR / job_id
+    job_dir.mkdir(parents=True, exist_ok=True)
+
+    base = {
+        "job_id": job_id,
+        "status_url": make_api_url(f"/jobs/{job_id}"),
+        "result_url": make_api_url(f"/jobs/{job_id}/result"),
+    }
+    base.update(payload)
+
+    path = job_status_path(job_id)
+    tmp_path = path.with_suffix(path.suffix + ".tmp")
+
+    tmp_path.write_text(
+        json.dumps(base, indent=2, ensure_ascii=False),
+        encoding="utf-8"
+    )
+
+    tmp_path.replace(path)
+
+
+def read_job_status(job_id: str) -> Dict[str, Any]:
+    path = job_status_path(job_id)
+    if not path.exists():
+        raise HTTPException(status_code=404, detail="Job not found.")
+
+    last_error = None
+
+    for _ in range(5):
+        try:
+            text = path.read_text(encoding="utf-8").strip()
+            if text:
+                return json.loads(text)
+        except json.JSONDecodeError as exc:
+            last_error = exc
+
+        time.sleep(0.1)
+
+    if last_error:
+        return {
+            "job_id": job_id,
+            "status_url": make_api_url(f"/jobs/{job_id}"),
+            "result_url": make_api_url(f"/jobs/{job_id}/result"),
+            "status": "running",
+            "stage": "status_update",
+            "message": "Job status is being updated. Retry shortly."
+        }
+
+    return {
+        "job_id": job_id,
+        "status_url": make_api_url(f"/jobs/{job_id}"),
+        "result_url": make_api_url(f"/jobs/{job_id}/result"),
+        "status": "running",
+        "stage": "status_update",
+        "message": "Job status is not ready yet."
+    }
+
+def run_generate_job(job_id: str, req_data: Dict[str, Any]):
+    job_dir = JOBS_DIR / job_id
+
+    try:
+        req = GenerateRequest(**req_data)
+
+        write_job_status(job_id, {
+            "status": "running",
+            "stage": "sampling",
+            "message": "Running REINVENT molecule generation",
+            "request": req_data,
+        })
+
+        runtime_config = write_runtime_sampling_config(job_dir, req.num_molecules)
+
+        run_reinvent_sampling(runtime_config, job_dir)
+
+        write_job_status(job_id, {
+            "status": "running",
+            "stage": "enrichment",
+            "message": "Running RDKit descriptors and DeepPurpose predictions",
+            "request": req_data,
+        })
+
+        enriched_csv = run_enrichment(job_dir, req.return_top_k)
+
+        write_job_status(job_id, {
+            "status": "running",
+            "stage": "docking",
+            "message": f"Running docking mode: {req.docking_mode}",
+            "request": req_data,
+        })
+
+        final_csv = run_optional_docking(job_dir, enriched_csv, req)
+
+        df = pd.read_csv(final_csv)
+
+        if "docking_pose_file" in df.columns:
+            df = df.drop(columns=["docking_pose_file"])
+
+        clean_csv = job_dir / "generated_results.csv"
+        clean_json = job_dir / "generated_results.json"
+
+        df.to_csv(clean_csv, index=False)
+
+        results = dataframe_to_public_records(df)
+
+        result_payload = {
+            "job_id": job_id,
+            "status": "completed",
+            "preset": req.preset,
+            "docking_mode": req.docking_mode,
+            "summary": {
+                "num_requested": req.num_molecules,
+                "num_generated": int(len(pd.read_csv(job_dir / "generated_smiles.csv"))),
+                "num_valid": int((df["valid"] == True).sum()) if "valid" in df.columns else None,
+                "num_returned": int(len(df)),
+                "num_docked": count_docked(df),
+            },
+            "files": {
+                "csv": make_file_meta(job_id, clean_csv),
+                "json": make_file_meta(job_id, clean_json),
+            },
+            "results": results,
+            "warnings": [
+                "Outputs are computational predictions only."
+            ],
+        }
+
+        clean_json.write_text(
+            json.dumps(result_payload, indent=2, ensure_ascii=False),
+            encoding="utf-8"
+        )
+
+        write_job_status(job_id, {
+            "status": "completed",
+            "stage": "completed",
+            "message": "Generation job completed",
+            "request": req_data,
+            "summary": result_payload["summary"],
+            "files": result_payload["files"],
+        })
+
+    except Exception as e:
+        if isinstance(e, HTTPException):
+            detail = e.detail
+        else:
+            detail = str(e)
+
+        write_job_status(job_id, {
+            "status": "failed",
+            "stage": "failed",
+            "message": "Generation job failed",
+            "request": req_data,
+            "error": detail,
+        })
+
+# -----------------------------
+# Endpoints
+# -----------------------------
+
+@app.get("/health")
+def health():
+    return {
+        "status": "ok",
+        "bundle_exists": BUNDLE_DIR.exists(),
+        "models_generator_exists": (BUNDLE_DIR / "models" / "generator" / "egfr_generator.chkpt").exists(),
+        "models_affinity_exists": (BUNDLE_DIR / "models" / "affinity" / "model.pt").exists(),
+        "docking_grid_exists": (BUNDLE_DIR / "docking" / "maps_current" / "4WKQ_receptor_v5_SBr.maps.fld").exists(),
+        "public_base_url": PUBLIC_BASE_URL,
+        "reinvent_device": REINVENT_DEVICE,
+        "adgpu_bin": ADGPU_BIN or None,
+    }
+
+
+@app.post("/reinvent_predict")
+async def reinvent_predict(request: Request):
+    if not HAS_AFFINITY:
+        return JSONResponse(
+            status_code=503,
+            content={"error": "Affinity prediction service not available (DeepPurpose not installed)"}
+        )
+    payload = await request.json()
+    smiles = payload.get("smiles", []) if isinstance(payload, dict) else []
+    preds = predict_smiles_list(smiles)
+    return {"pred_pAff_mean": preds}
+
+
+@app.get("/files/jobs/{job_id}/{path:path}")
+def get_job_file(job_id: str, path: str):
+    root = (JOBS_DIR / job_id).resolve()
+    target = (root / path).resolve()
+
+    if not str(target).startswith(str(root)):
+        raise HTTPException(status_code=400, detail="Invalid file path.")
+
+    if not target.exists() or not target.is_file():
+        raise HTTPException(status_code=404, detail="File not found.")
+
+    return FileResponse(
+        path=str(target),
+        filename=target.name,
+        media_type="application/octet-stream"
+    )
+
+
+@app.post("/ligands/export")
+def export_ligand(req: LigandExportRequest):
+    job_id = new_job_id("lig")
+    job_dir = JOBS_DIR / job_id
+    job_dir.mkdir(parents=True, exist_ok=True)
+
+    mol, canonical = make_3d_mol(req.smiles)
+
+    sdf_path = job_dir / "ligand_3d.sdf"
+    write_sdf_for_conversion(mol, sdf_path)
+
+    if req.format == "pdb":
+        out_path = job_dir / "ligand_3d.pdb"
+        export_pdb(mol, out_path)
+
+    elif req.format == "pdbqt":
+        out_path = job_dir / "ligand_3d.pdbqt"
+        export_pdbqt(sdf_path, out_path)
+
+    elif req.format == "mol2":
+        out_path = job_dir / "ligand_3d.mol2"
+        export_mol2(sdf_path, out_path)
+
+    else:
+        raise HTTPException(status_code=400, detail="Unsupported format.")
+
+    meta = make_file_meta(job_id, out_path)
+    meta["format"] = req.format
+
+    return {
+        "job_id": job_id,
+        "status": "completed",
+        "canonical_smiles": canonical,
+        "format": req.format,
+        "file": meta,
+    }
+
+
+def write_runtime_sampling_config(job_dir: Path, num_molecules: int) -> Path:
+    """
+    Create a job-specific REINVENT sampling config.
+    Keeps bundle config untouched.
+    """
+    template_path = BUNDLE_DIR / "configs" / "sampling_model.bundle.toml"
+    if not template_path.exists():
+        raise HTTPException(status_code=500, detail=f"Sampling config not found: {template_path}")
+
+    text = template_path.read_text(encoding="utf-8")
+
+    # Force runtime device and job-specific outputs.
+    text = text.replace('device = "cuda:0"', f'device = "{REINVENT_DEVICE}"')
+    text = text.replace('json_out_config = "outputs/generate/sampling_resolved.json"', f'json_out_config = "{(job_dir / "sampling_resolved.json").as_posix()}"')
+    text = text.replace('model_file = "models/generator/egfr_generator.chkpt"', f'model_file = "{(BUNDLE_DIR / "models" / "generator" / "egfr_generator.chkpt").as_posix()}"')
+    text = text.replace('output_file = "outputs/generate/generated_smiles.csv"', f'output_file = "{(job_dir / "generated_smiles.csv").as_posix()}"')
+
+    # Replace num_smiles line safely.
+    lines = []
+    replaced = False
+    for line in text.splitlines():
+        if line.strip().startswith("num_smiles"):
+            lines.append(f"num_smiles = {int(num_molecules)}")
+            replaced = True
+        else:
+            lines.append(line)
+
+    if not replaced:
+        lines.append(f"num_smiles = {int(num_molecules)}")
+
+    runtime_config = job_dir / "sampling_runtime.toml"
+    runtime_config.write_text("\n".join(lines) + "\n", encoding="utf-8")
+    return runtime_config
+
+
+def find_reinvent_command() -> List[str]:
+    """
+    Try common REINVENT4 entrypoints.
+    Docker will normally expose `reinvent`.
+    Local fallback may use python -m reinvent.Reinvent.
+    """
+    if shutil.which("reinvent"):
+        return ["reinvent"]
+
+    return [os.getenv("PYTHON", "python"), "-m", "reinvent.Reinvent"]
+
+
+def run_reinvent_sampling(runtime_config: Path, job_dir: Path):
+    cmd = find_reinvent_command() + ["-l", str(job_dir / "reinvent_sampling.log"), str(runtime_config)]
+
+    result = run_cmd(cmd, cwd=BUNDLE_DIR, timeout=1800)
+
+    log_path = job_dir / "reinvent_command_output.log"
+    log_path.write_text(result.stdout or "", encoding="utf-8", errors="replace")
+
+    if result.returncode != 0:
+        raise HTTPException(
+            status_code=500,
+            detail={
+                "message": "REINVENT sampling failed.",
+                "command": " ".join(cmd),
+                "log_tail": (result.stdout or "")[-2000:],
+            }
+        )
+
+
+def run_enrichment(job_dir: Path, return_top_k: int) -> Path:
+    input_csv = job_dir / "generated_smiles.csv"
+    enriched_csv = job_dir / "generated_smiles_enriched.csv"
+
+    if not input_csv.exists():
+        raise HTTPException(status_code=500, detail=f"Generated SMILES CSV not found: {input_csv}")
+
+    script = BUNDLE_DIR / "tools" / "enrich_generated.py"
+    if not script.exists():
+        raise HTTPException(status_code=500, detail=f"Enrichment script not found: {script}")
+
+    cmd = [
+        os.getenv("PYTHON", "python"),
+        str(script),
+        "--input", str(input_csv),
+        "--output", str(enriched_csv),
+        "--affinity-url", DEEPPURPOSE_URL,
+        "--top-k", str(return_top_k),
+    ]
+
+    result = run_cmd(cmd, cwd=BUNDLE_DIR, timeout=1800)
+
+    (job_dir / "enrichment.log").write_text(result.stdout or "", encoding="utf-8", errors="replace")
+
+    if result.returncode != 0:
+        raise HTTPException(
+            status_code=500,
+            detail={
+                "message": "Enrichment failed.",
+                "command": " ".join(cmd),
+                "log_tail": (result.stdout or "")[-2000:],
+            }
+        )
+
+    return enriched_csv
+
+
+def run_optional_docking(job_dir: Path, enriched_csv: Path, req: GenerateRequest) -> Path:
+    final_csv = job_dir / "generated_results.csv"
+
+    script = BUNDLE_DIR / "tools" / "dock_enriched.py"
+    if not script.exists():
+        raise HTTPException(status_code=500, detail=f"Docking script not found: {script}")
+
+    grid_file = BUNDLE_DIR / "docking" / "maps_current" / "4WKQ_receptor_v5_SBr.maps.fld"
+
+    cmd = [
+        os.getenv("PYTHON", "python"),
+        str(script),
+        "--input", str(enriched_csv),
+        "--output", str(final_csv),
+        "--docking-mode", req.docking_mode,
+        "--dock-top-k", str(req.dock_top_k),
+        "--grid-file", str(grid_file),
+        "--work-dir", str(job_dir / "docking"),
+    ]
+
+    if req.docking_mode != "off":
+        if not ADGPU_BIN:
+            raise HTTPException(
+                status_code=500,
+                detail="Docking requested, but ADGPU_BIN is not set. Set AutoDock-GPU binary path in production."
+            )
+        cmd += ["--adgpu-bin", ADGPU_BIN]
+
+    result = run_cmd(cmd, cwd=BUNDLE_DIR, timeout=7200)
+
+    (job_dir / "docking.log").write_text(result.stdout or "", encoding="utf-8", errors="replace")
+
+    if result.returncode != 0:
+        raise HTTPException(
+            status_code=500,
+            detail={
+                "message": "Docking step failed.",
+                "command": " ".join(cmd),
+                "log_tail": (result.stdout or "")[-2000:],
+            }
+        )
+
+    return final_csv
+
+
+@app.post("/generate", status_code=202)
+def submit_generate(req: GenerateRequest, background_tasks: BackgroundTasks):
+    job_id = new_job_id("gen")
+    job_dir = JOBS_DIR / job_id
+    job_dir.mkdir(parents=True, exist_ok=True)
+
+    req_data = req.model_dump()
+
+    write_job_status(job_id, {
+        "status": "queued",
+        "stage": "queued",
+        "message": "Generation job accepted and queued",
+        "request": req_data,
+    })
+
+    background_tasks.add_task(run_generate_job, job_id, req_data)
+
+    return {
+        "job_id": job_id,
+        "status": "queued",
+        "message": "Generation job accepted. Poll status_url until completed.",
+        "status_url": make_api_url(f"/jobs/{job_id}"),
+        "result_url": make_api_url(f"/jobs/{job_id}/result"),
+    }
+
+
+@app.get("/jobs/{job_id}")
+def get_job_status(job_id: str):
+    return read_job_status(job_id)
+
+
+@app.get("/jobs/{job_id}/result")
+def get_job_result(job_id: str):
+    status = read_job_status(job_id)
+
+    if status.get("status") == "failed":
+        return JSONResponse(status_code=500, content=status)
+
+    if status.get("status") != "completed":
+        return JSONResponse(status_code=202, content=status)
+
+    result_path = JOBS_DIR / job_id / "generated_results.json"
+    if not result_path.exists():
+        raise HTTPException(status_code=404, detail="Result file not found.")
+
+    return json.loads(result_path.read_text(encoding="utf-8"))

bundle/.gitattributes

@@ -0,0 +1,62 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.avro filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text
+models/generator/egfr_generator.chkpt filter=lfs diff=lfs merge=lfs -text
+models/generator/reinvent.prior filter=lfs diff=lfs merge=lfs -text

bundle/README.md

@@ -0,0 +1,18 @@
+# EGFR Backend Bundle
+
+This folder contains the packaged backend assets required by the Ailixir EGFR generation pipeline.
+
+## Components
+
+- Generator checkpoint
+- Sampling configuration
+- DeepPurpose affinity model files
+- EGFR target sequence
+- Docking grid/maps
+- Backend utility tools
+
+## Notes
+
+This bundle is intended for backend integration and Docker-based execution.
+
+AutoDock-GPU docking is supported by the API container when running in an NVIDIA GPU Docker environment.

bundle/configs/backend_contract.json

@@ -0,0 +1,80 @@
+{
+  "project": "EGFR AI backend bundle",
+  "version": "v1-minimal",
+  "modes": {
+    "generate": {
+      "type": "sampling_from_packaged_generator",
+      "config": "configs/sampling_model.bundle.toml",
+      "checkpoint": "models/generator/egfr_generator.chkpt",
+      "allowed_user_inputs": [
+        "preset",
+        "num_molecules",
+        "return_top_k"
+      ],
+      "do_not_expose": [
+        "raw TOML",
+        "checkpoint paths",
+        "training stages",
+        "RL scoring strategy",
+        "batch_size",
+        "bucket_size"
+      ],
+      "postprocess": {
+        "script": "tools/enrich_generated.py",
+        "input": "outputs/generate/generated_smiles.csv",
+        "output": "outputs/generate/generated_smiles_enriched.csv",
+        "adds_columns": [
+          "valid",
+          "canonical_smiles",
+          "mw",
+          "logp",
+          "tpsa",
+          "hbd",
+          "hba",
+          "rot_bonds",
+          "qed",
+          "sa_score",
+          "pred_pAff_mean"
+        ],
+        "sort_note": "Display sorting may use pred_pAff_mean desc then QED desc. This is not a final scientific ranking."
+      }
+    },
+    "score_smiles": {
+      "type": "affinity_api",
+      "run_command": "uvicorn services.deeppurpose.serve_affinity:app --host 127.0.0.1 --port 8001",
+      "health_endpoint": "GET /health",
+      "predict_endpoint": "POST /reinvent_predict",
+      "request_example": {
+        "smiles": [
+          "CCO",
+          "c1ccccc1"
+        ]
+      },
+      "response_key": "pred_pAff_mean"
+    },
+    "docking": {
+      "enabled_by_default": false,
+      "allowed_modes": [
+        "off",
+        "top_k",
+        "all"
+      ],
+      "recommended_mode": "top_k",
+      "top_k_default": 10,
+      "maps_dir": "docking/maps_current",
+      "grid_file": "docking/maps_current/4WKQ_receptor_v5_SBr.maps.fld",
+      "binary_env_var": "ADGPU_BIN",
+      "result_columns": [
+        "docking_score",
+        "docking_status",
+        "docking_pose_file"
+      ],
+      "when_disabled": {
+        "docking_score": null,
+        "docking_status": "not_run",
+        "docking_pose_file": null
+      },
+      "note": "Docking is optional and backend-controlled. Use top_k for demos because docking is slower than descriptor and affinity scoring."
+    }
+  }
+}

bundle/configs/sampling_model.bundle.toml

@@ -0,0 +1,16 @@
+# Minimal deployment sampling config
+# Samples molecules from the packaged EGFR generator checkpoint.
+# Does not expose the original training/RL strategy.
+
+run_type = "sampling"
+device = "cuda:0"
+
+json_out_config = "outputs/generate/sampling_resolved.json"
+
+[parameters]
+model_file = "models/generator/egfr_generator.chkpt"
+output_file = "outputs/generate/generated_smiles.csv"
+
+num_smiles = 100
+unique_molecules = true
+randomize_smiles = true

bundle/docking/maps_current/4WKQ_receptor.A.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54b59003346b5d5df1c52329a01a0792259cf10cddaaf3cbf45aeda792470e4a
+size 2049653

bundle/docking/maps_current/4WKQ_receptor.Br.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a39ee75ecf7590c47b095ae19325c5ddf83b1a5843049ae8b2f19904986821d
+size 2069949

bundle/docking/maps_current/4WKQ_receptor.C.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78b2d1b94f8b266ba805d37376b010bad13411c43900d5d5d71899f0ac8f5fd8
+size 2052665

bundle/docking/maps_current/4WKQ_receptor.Cl.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59adaf85853d0ea362160d08d5af5516fbfaf5d25743884f0c2a8dcbecb14f38
+size 2082114

bundle/docking/maps_current/4WKQ_receptor.F.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5ea3924fbd0f7b8acd4b7f44d90f4e6474fead2ce0e9ab79d07de50b628a48c
+size 1954132

bundle/docking/maps_current/4WKQ_receptor.HD.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9c9269b5133820343a14551159aa1f6eefea6308cf1b47245a57fec326b7f6f
+size 1820642

bundle/docking/maps_current/4WKQ_receptor.N.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06b3ba3e6dc84c6d2b5f91d2026c18626cb80861fb8d7ea63e9d0c21fef6863e
+size 2012924

bundle/docking/maps_current/4WKQ_receptor.NA.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f55ea020098731262b5965462d77bb5d678411d5dcf22dd1c5bde218331fd50
+size 2014019

bundle/docking/maps_current/4WKQ_receptor.OA.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69594facd6e979219f24134b0dede413b257b2b4043c746e8349f092a9800a40
+size 2000447

bundle/docking/maps_current/4WKQ_receptor.S.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1ce99441b16d1c0b3cd87ecb4488e7fbff4bb55a830e0e904bf34747e90c167
+size 2065476

bundle/docking/maps_current/4WKQ_receptor.SA.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f10253e8b726f975329cc64e16308ddf1a24cd48f6a68f02d60725a484a7e39c
+size 2115792

bundle/docking/maps_current/4WKQ_receptor.d.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:880b0e603daa5b8bd200c44753c43d9ce1209c98a68fa1954d23e78d5715a1d1
+size 1599272

bundle/docking/maps_current/4WKQ_receptor.e.map

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54a19cf5caa31d2be5fa3936c7f1bb205360022447094637ac46bea02b05191a
+size 1839643

bundle/docking/maps_current/4WKQ_receptor_v5_SBr.maps.fld

@@ -0,0 +1,52 @@
+# AVS field file
+#
+# AutoDock Atomic Affinity and Electrostatic Grids
+#
+# Created by autogrid4.
+#
+#SPACING 0.375
+#NELEMENTS 64 64 64
+#CENTER 7.375 194.927 16.540
+#MACROMOLECULE receptorH_rigid.pdbqt
+#GRID_PARAMETER_FILE receptorH.gpf
+#
+ndim=3			# number of dimensions in the field
+dim1=65			# number of x-elements
+dim2=65			# number of y-elements
+dim3=65			# number of z-elements
+nspace=3		# number of physical coordinates per point
+veclen=13		# number of affinity values at each point
+data=float		# data type (byte, integer, float, double)
+field=uniform		# field type (uniform, rectilinear, irregular)
+coord 1 file=4WKQ_receptor_v5_SBr.maps.xyz filetype=ascii offset=0
+coord 2 file=4WKQ_receptor_v5_SBr.maps.xyz filetype=ascii offset=2
+coord 3 file=4WKQ_receptor_v5_SBr.maps.xyz filetype=ascii offset=4
+label=A-affinity	# component label for variable 1
+label=C-affinity	# component label for variable 2
+label=F-affinity	# component label for variable 3
+label=N-affinity	# component label for variable 4
+label=NA-affinity	# component label for variable 5
+label=Cl-affinity	# component label for variable 6
+label=OA-affinity	# component label for variable 7
+label=HD-affinity	# component label for variable 8
+label=S-affinity	# component label for variable 9
+label=SA-affinity	# component label for variable 10
+label=Br-affinity	# component label for variable 11
+label=Electrostatics	# component label for variable 11
+label=Desolvation	# component label for variable 12
+#
+# location of affinity grid files and how to read them
+#
+variable 1 file=4WKQ_receptor.A.map filetype=ascii skip=6
+variable 2 file=4WKQ_receptor.C.map filetype=ascii skip=6
+variable 3 file=4WKQ_receptor.F.map filetype=ascii skip=6
+variable 4 file=4WKQ_receptor.N.map filetype=ascii skip=6
+variable 5 file=4WKQ_receptor.NA.map filetype=ascii skip=6
+variable 6 file=4WKQ_receptor.Cl.map filetype=ascii skip=6
+variable 7 file=4WKQ_receptor.OA.map filetype=ascii skip=6
+variable 8 file=4WKQ_receptor.HD.map filetype=ascii skip=6
+variable 9 file=4WKQ_receptor.S.map filetype=ascii skip=6
+variable 10 file=4WKQ_receptor.Br.map filetype=ascii skip=6
+variable 11 file=4WKQ_receptor.SA.map filetype=ascii skip=6
+variable 12 file=4WKQ_receptor.e.map filetype=ascii skip=6
+variable 13 file=4WKQ_receptor.d.map filetype=ascii skip=6

bundle/docking/maps_current/4WKQ_receptor_v5_SBr.maps.xyz

@@ -0,0 +1,3 @@
+-4.625 19.375
+182.927 206.927
+4.540 28.540

bundle/docking/maps_current/ligand.sdf

@@ -0,0 +1,217 @@
+
+     RDKit          3D
+
+103108  0  0  0  0  0  0  0  0999 V2000
+   14.9159    0.7454   -1.8018 C   0  0  0  0  0  0  0  0  0  0  0  0
+   13.5628    0.2305   -1.3211 C   0  0  0  0  0  0  0  0  0  0  0  0
+   13.2828    0.6808    0.0443 N   0  0  0  0  0  0  0  0  0  0  0  0
+   12.0685    0.3583    0.7486 C   0  0  0  0  0  0  0  0  0  0  0  0
+   11.8744    0.8586    2.0450 C   0  0  0  0  0  0  0  0  0  0  0  0
+   10.7055    0.5671    2.7510 C   0  0  0  0  0  0  0  0  0  0  0  0
+    9.7139   -0.2293    2.1744 C   0  0  0  0  0  0  0  0  0  0  0  0
+    8.5844   -0.4901    2.8801 F   0  0  0  0  0  0  0  0  0  0  0  0
+    9.8899   -0.7448    0.8778 C   0  0  0  0  0  0  0  0  0  0  0  0
+    8.8269   -1.5940    0.2244 C   0  0  0  0  0  0  0  0  0  0  0  0
+    8.2308   -0.8951   -0.9367 N   0  0  0  0  0  0  0  0  0  0  0  0
+    7.5068   -1.8162   -1.8550 C   0  0  0  0  0  0  0  0  0  0  0  0
+    5.9734   -1.6512   -1.8400 C   0  0  0  0  0  0  0  0  0  0  0  0
+    5.4710   -1.3601   -0.4180 C   0  0  0  0  0  0  0  0  0  0  0  0
+    3.9299   -1.3365   -0.3465 C   0  0  0  0  0  0  0  0  0  0  0  0
+    3.3592   -0.3912   -1.3213 N   0  0  0  0  0  0  0  0  0  0  0  0
+    1.9654   -0.7133   -1.6653 C   0  0  0  0  0  0  0  0  0  0  0  0
+    0.9998   -0.4757   -0.4934 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -0.3856   -0.6700   -0.9425 N   0  0  0  0  0  0  0  0  0  0  0  0
+   -1.3427   -0.3567    0.1329 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -1.4734    1.1548    0.3590 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -2.9341    1.3957    0.7147 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -3.6481    0.2147    0.2574 N   0  0  0  0  0  0  0  0  0  0  0  0
+   -5.0813    0.0701    0.2617 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -5.9154    1.0564    0.8098 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -7.3025    0.8593    0.7886 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -8.1997    1.8486    1.3268 N   0  0  0  0  0  0  0  0  0  0  0  0
+   -7.9960    2.3693    2.6856 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -8.5366    3.8019    2.7919 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -9.9217    3.8869    2.3059 N   0  0  0  0  0  0  0  0  0  0  0  0
+   -9.9981    3.4632    0.8964 C   0  0  0  0  0  0  0  0  0  0  0  0
+  -11.4008    3.7222    0.3029 C   0  0  0  0  0  0  0  0  0  0  0  0
+  -12.5458    2.9790    0.9967 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -9.5337    2.0031    0.7368 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -7.8044   -0.2700    0.2210 N   0  0  0  0  0  0  0  0  0  0  0  0
+   -6.9908   -1.2213   -0.3080 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -7.5507   -2.4045   -0.8952 N   0  0  0  0  0  0  0  0  0  0  0  0
+   -6.7501   -3.4561   -1.4955 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -7.7480   -4.4199   -2.1245 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -9.1043   -4.0854   -1.5056 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -8.9763   -2.6722   -0.9520 C   0  0  0  0  0  0  0  0  0  0  0  0
+   -5.6436   -1.0439   -0.2798 N   0  0  0  0  0  0  0  0  0  0  0  0
+   -2.7549   -0.8224   -0.2303 C   0  0  0  0  0  0  0  0  0  0  0  0
+    6.1124   -0.0702    0.1478 C   0  0  0  0  0  0  0  0  0  0  0  0
+    7.4361    0.2941   -0.5539 C   0  0  0  0  0  0  0  0  0  0  0  0
+   11.0666   -0.4459    0.1707 C   0  0  0  0  0  0  0  0  0  0  0  0
+   14.9260    1.8562   -1.7958 H   0  0  0  0  0  0  0  0  0  0  0  0
+   15.7254    0.3669   -1.1420 H   0  0  0  0  0  0  0  0  0  0  0  0
+   15.1037    0.3914   -2.8373 H   0  0  0  0  0  0  0  0  0  0  0  0
+   12.7725    0.6077   -2.0055 H   0  0  0  0  0  0  0  0  0  0  0  0
+   13.5712   -0.8804   -1.3511 H   0  0  0  0  0  0  0  0  0  0  0  0
+   13.9929    1.2726    0.5345 H   0  0  0  0  0  0  0  0  0  0  0  0
+   12.6304    1.4791    2.5097 H   0  0  0  0  0  0  0  0  0  0  0  0
+   10.5677    0.9629    3.7491 H   0  0  0  0  0  0  0  0  0  0  0  0
+    8.0450   -1.9017    0.9497 H   0  0  0  0  0  0  0  0  0  0  0  0
+    9.3298   -2.5288   -0.1089 H   0  0  0  0  0  0  0  0  0  0  0  0
+    7.8669   -1.6304   -2.8891 H   0  0  0  0  0  0  0  0  0  0  0  0
+    7.7422   -2.8821   -1.6378 H   0  0  0  0  0  0  0  0  0  0  0  0
+    5.6891   -0.8178   -2.5198 H   0  0  0  0  0  0  0  0  0  0  0  0
+    5.5018   -2.5799   -2.2298 H   0  0  0  0  0  0  0  0  0  0  0  0
+    5.7995   -2.2085    0.2238 H   0  0  0  0  0  0  0  0  0  0  0  0
+    3.6142   -1.0626    0.6851 H   0  0  0  0  0  0  0  0  0  0  0  0
+    3.5669   -2.3705   -0.5447 H   0  0  0  0  0  0  0  0  0  0  0  0
+    3.3926    0.5782   -0.9291 H   0  0  0  0  0  0  0  0  0  0  0  0
+    1.8881   -1.7648   -2.0201 H   0  0  0  0  0  0  0  0  0  0  0  0
+    1.6591   -0.0640   -2.5140 H   0  0  0  0  0  0  0  0  0  0  0  0
+    1.2321   -1.1828    0.3338 H   0  0  0  0  0  0  0  0  0  0  0  0
+    1.1398    0.5639   -0.1255 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -0.4920   -1.6854   -1.1794 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -1.0441   -0.8489    1.0884 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -1.2285    1.7061   -0.5763 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -0.8011    1.5012    1.1738 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -3.3115    2.3047    0.1977 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -3.0549    1.5023    1.8146 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -5.5015    1.9648    1.2243 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -6.9223    2.3719    2.9653 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -8.5321    1.7164    3.4082 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -7.8874    4.4932    2.2104 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -8.5009    4.1277    3.8539 H   0  0  0  0  0  0  0  0  0  0  0  0
+  -10.5062    3.2500    2.8966 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -9.3002    4.0960    0.3005 H   0  0  0  0  0  0  0  0  0  0  0  0
+  -11.6112    4.8127    0.3544 H   0  0  0  0  0  0  0  0  0  0  0  0
+  -11.3963    3.4405   -0.7727 H   0  0  0  0  0  0  0  0  0  0  0  0
+  -12.4097    1.8807    0.9312 H   0  0  0  0  0  0  0  0  0  0  0  0
+  -13.5009    3.2320    0.4901 H   0  0  0  0  0  0  0  0  0  0  0  0
+  -12.6355    3.2857    2.0588 H   0  0  0  0  0  0  0  0  0  0  0  0
+  -10.2297    1.3063    1.2488 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -9.5089    1.7490   -0.3462 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -6.0724   -3.0348   -2.2697 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -6.1548   -3.9670   -0.7078 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -7.4656   -5.4776   -1.9310 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -7.7900   -4.2490   -3.2228 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -9.9194   -4.1525   -2.2585 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -9.3165   -4.7935   -0.6745 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -9.4662   -1.9399   -1.6299 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -9.4247   -2.6080    0.0633 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -2.9803   -1.7931    0.2631 H   0  0  0  0  0  0  0  0  0  0  0  0
+   -2.8810   -0.9236   -1.3309 H   0  0  0  0  0  0  0  0  0  0  0  0
+    5.4300    0.7995    0.0518 H   0  0  0  0  0  0  0  0  0  0  0  0
+    6.2968   -0.2059    1.2356 H   0  0  0  0  0  0  0  0  0  0  0  0
+    8.0273    0.9758    0.0939 H   0  0  0  0  0  0  0  0  0  0  0  0
+    7.2083    0.8779   -1.4730 H   0  0  0  0  0  0  0  0  0  0  0  0
+   11.1885   -0.8433   -0.8282 H   0  0  0  0  0  0  0  0  0  0  0  0
+  1  2  1  0
+  2  3  1  0
+  3  4  1  0
+  4  5  2  0
+  5  6  1  0
+  6  7  2  0
+  7  8  1  0
+  7  9  1  0
+  9 10  1  0
+ 10 11  1  0
+ 11 12  1  0
+ 12 13  1  0
+ 13 14  1  0
+ 14 15  1  0
+ 15 16  1  0
+ 16 17  1  0
+ 17 18  1  0
+ 18 19  1  0
+ 19 20  1  0
+ 20 21  1  0
+ 21 22  1  0
+ 22 23  1  0
+ 23 24  1  0
+ 24 25  2  0
+ 25 26  1  0
+ 26 27  1  0
+ 27 28  1  0
+ 28 29  1  0
+ 29 30  1  0
+ 30 31  1  0
+ 31 32  1  0
+ 32 33  1  0
+ 31 34  1  0
+ 26 35  2  0
+ 35 36  1  0
+ 36 37  1  0
+ 37 38  1  0
+ 38 39  1  0
+ 39 40  1  0
+ 40 41  1  0
+ 36 42  2  0
+ 23 43  1  0
+ 14 44  1  0
+ 44 45  1  0
+  9 46  2  0
+ 46  4  1  0
+ 45 11  1  0
+ 43 20  1  0
+ 42 24  1  0
+ 34 27  1  0
+ 41 37  1  0
+  1 47  1  0
+  1 48  1  0
+  1 49  1  0
+  2 50  1  0
+  2 51  1  0
+  3 52  1  0
+  5 53  1  0
+  6 54  1  0
+ 10 55  1  0
+ 10 56  1  0
+ 12 57  1  0
+ 12 58  1  0
+ 13 59  1  0
+ 13 60  1  0
+ 14 61  1  0
+ 15 62  1  0
+ 15 63  1  0
+ 16 64  1  0
+ 17 65  1  0
+ 17 66  1  0
+ 18 67  1  0
+ 18 68  1  0
+ 19 69  1  0
+ 20 70  1  0
+ 21 71  1  0
+ 21 72  1  0
+ 22 73  1  0
+ 22 74  1  0
+ 25 75  1  0
+ 28 76  1  0
+ 28 77  1  0
+ 29 78  1  0
+ 29 79  1  0
+ 30 80  1  0
+ 31 81  1  0
+ 32 82  1  0
+ 32 83  1  0
+ 33 84  1  0
+ 33 85  1  0
+ 33 86  1  0
+ 34 87  1  0
+ 34 88  1  0
+ 38 89  1  0
+ 38 90  1  0
+ 39 91  1  0
+ 39 92  1  0
+ 40 93  1  0
+ 40 94  1  0
+ 41 95  1  0
+ 41 96  1  0
+ 43 97  1  0
+ 43 98  1  0
+ 44 99  1  0
+ 44100  1  0
+ 45101  1  0
+ 45102  1  0
+ 46103  1  0
+M  END
+$$$$

bundle/docking/maps_current/receptorH.gpf

@@ -0,0 +1,22 @@
+npts 64 64 64                              # num.grid points in xyz
+gridfld 4WKQ_receptor_v5_SBr.maps.fld      # grid_data_file
+spacing 0.375                              # spacing(A)
+receptor_types A C NA OA N SA HD           # receptor atom types
+ligand_types A C F N NA Cl OA HD S SA Br   # ligand atom types
+receptor receptorH_rigid.pdbqt             # macromolecule
+gridcenter 7.375 194.927 16.540        # xyz-coordinates or auto
+smooth 0.5                             # store minimum energy w/in rad(A)
+map 4WKQ_receptor.A.map                # atom-specific affinity map
+map 4WKQ_receptor.C.map                # atom-specific affinity map
+map 4WKQ_receptor.F.map                # atom-specific affinity map
+map 4WKQ_receptor.N.map                # atom-specific affinity map
+map 4WKQ_receptor.NA.map               # atom-specific affinity map
+map 4WKQ_receptor.Cl.map               # atom-specific affinity map
+map 4WKQ_receptor.OA.map               # atom-specific affinity map
+map 4WKQ_receptor.HD.map               # atom-specific affinity map
+map 4WKQ_receptor.S.map                # atom-specific affinity map
+map 4WKQ_receptor.Br.map               # atom-specific affinity map
+map 4WKQ_receptor.SA.map               # atom-specific affinity map
+elecmap 4WKQ_receptor.e.map            # electrostatic potential map
+dsolvmap 4WKQ_receptor.d.map           # desolvation potential map
+dielectric -0.1465                     # <0, AD4 distance-dep.diel;>0, constant

bundle/docking/maps_current/receptorH_flex.pdbqt

@@ -0,0 +1,13 @@
+BEGIN_RES CYS A 797
+REMARK INDEX MAP 3 1 15 2 18 3 19 4
+ROOT
+ATOM      1  CA  CYS A 797       5.927 197.219  16.594  1.00  0.00     0.181 C 
+ENDROOT
+BRANCH   1   2
+ATOM      2  CB  CYS A 797       6.611 196.143  17.446  1.00  0.00     0.101 C 
+BRANCH   2   3
+ATOM      3  SG  CYS A 797       7.375 194.927  16.540  1.00  0.00    -0.177 SA
+ATOM      4  HG  CYS A 797       7.947 194.016  17.389  1.00  0.00     0.102 HD
+ENDBRANCH   2   3
+ENDBRANCH   1   2
+END_RES CYS A 797

bundle/docs/API_RESPONSE_SPEC.md

@@ -0,0 +1,111 @@
+# API Response Specification — EGFR AI Backend Bundle
+
+This document describes the expected response shape for the backend/frontend integration.
+
+## 1. Generate endpoint
+
+### Request
+
+```json
+{
+  "preset": "egfr_generator",
+  "num_molecules": 100,
+  "return_top_k": 20,
+  "docking_mode": "off"
+}
+
+Allowed docking modes
+off
+top_k
+all
+
+Meaning:
+
+off: no docking is run. Docking columns are returned as empty/null.
+top_k: docking is run only for the top-k generated molecules after property + affinity scoring.
+all: docking is run for all returned molecules. This is slower.
+##
+
+2. Response
+**Review ai_artifacts_release\examples\generate_response.example.json
+{
+  "job_id": "gen_001",
+  "preset": "egfr_generator",
+  "status": "completed",
+  "docking_mode": "off",
+  "summary": {
+    "num_generated": 100,
+    "num_valid": 100,
+    "num_returned": 20
+  },
+  "columns": [
+    "SMILES",
+    "SMILES_state",
+    "NLL",
+    "valid",
+    "canonical_smiles",
+    "mw",
+    "logp",
+    "tpsa",
+    "hbd",
+    "hba",
+    "rot_bonds",
+    "qed",
+    "sa_score",
+    "pred_pAff_mean",
+    "docking_score",
+    "docking_status",
+    "docking_pose_file"
+  ],
+  "results": [
+    {
+      "SMILES": "CNCCC(=O)Nc1ccc2nncc(-c3ccc4cncnc4c3)c2c1",
+      "SMILES_state": 1,
+      "NLL": 5.71,
+      "valid": true,
+      "canonical_smiles": "CNCCC(=O)Nc1ccc2nncc(-c3ccc4cncnc4c3)c2c1",
+      "mw": 358.405,
+      "logp": 2.788,
+      "tpsa": 92.69,
+      "hbd": 2,
+      "hba": 6,
+      "rot_bonds": 5,
+      "qed": 0.5698,
+      "sa_score": 2.5204,
+      "pred_pAff_mean": 11.0104,
+      "docking_score": null,
+      "docking_status": "not_run",
+      "docking_pose_file": null
+    }
+  ],
+  "warnings": [
+    "Outputs are computational predictions only.",
+    "Docking was not run for this job."   ## depend on the user's preference
+  ]
+}
+
+
+
+2. Score SMILES endpoint
+Request
+{
+  "smiles": [
+    "CCO",
+    "c1ccccc1"
+  ],
+  "docking_mode": "off"
+}
+
+
+
+4. Frontend/filter recommendations
+
+Recommended filters:
+
+valid == true
+mw between 250 and 600
+logp between 0 and 6
+tpsa between 40 and 140
+qed >= 0.3
+pred_pAff_mean available
+docking_status in ["not_run", "completed"]

bundle/examples/generate_response.example.json

@@ -0,0 +1,131 @@
+{
+  "job_id": "example_generation_job",
+  "preset": "egfr_generator",
+  "status": "completed",
+  "docking_mode": "top_k",
+  "summary": {
+    "num_generated": 100,
+    "num_valid": 5,
+    "num_returned": 5
+  },
+  "columns": [
+    "SMILES",
+    "SMILES_state",
+    "NLL",
+    "valid",
+    "canonical_smiles",
+    "mw",
+    "logp",
+    "tpsa",
+    "hbd",
+    "hba",
+    "rot_bonds",
+    "qed",
+    "sa_score",
+    "pred_pAff_mean",
+    "docking_score",
+    "docking_status",
+    "docking_pose_file"
+  ],
+  "results": [
+    {
+      "SMILES": "CNCCC(=O)Nc1ccc2nncc(-c3ccc4cncnc4c3)c2c1",
+      "SMILES_state": 1,
+      "NLL": 5.71,
+      "valid": true,
+      "canonical_smiles": "CNCCC(=O)Nc1ccc2nncc(-c3ccc4cncnc4c3)c2c1",
+      "mw": 358.4050000000001,
+      "logp": 2.7880000000000003,
+      "tpsa": 92.69,
+      "hbd": 2,
+      "hba": 6,
+      "rot_bonds": 5,
+      "qed": 0.5698238625255986,
+      "sa_score": 2.52037562103958,
+      "pred_pAff_mean": 11.01038932800293,
+      "docking_score": -8.86,
+      "docking_status": "completed",
+      "docking_pose_file": "outputs/docking/ligand_0001/ligand_0001.xml"
+    },
+    {
+      "SMILES": "CN1CCN(CCC(=O)Nc2ccc3nncc(-c4ccc5cncnc5c4)c3c2)C1",
+      "SMILES_state": 1,
+      "NLL": 6.86,
+      "valid": true,
+      "canonical_smiles": "CN1CCN(CCC(=O)Nc2ccc3nncc(-c4ccc5cncnc5c4)c3c2)C1",
+      "mw": 413.4850000000002,
+      "logp": 2.7735000000000003,
+      "tpsa": 87.14,
+      "hbd": 1,
+      "hba": 7,
+      "rot_bonds": 5,
+      "qed": 0.5382597382382679,
+      "sa_score": 2.7266538303986785,
+      "pred_pAff_mean": 10.975865364074709,
+      "docking_score": -8.93,
+      "docking_status": "completed",
+      "docking_pose_file": "outputs/docking/ligand_0002/ligand_0002.xml"
+    },
+    {
+      "SMILES": "CN1CCN(CCC(=O)Nc2ccc3nncc(-c4ccc5c(Nc6ccc7cncnc7c6)ncnc5c4)c3c2)CC1",
+      "SMILES_state": 1,
+      "NLL": 12.81,
+      "valid": true,
+      "canonical_smiles": "CN1CCN(CCC(=O)Nc2ccc3nncc(-c4ccc5c(Nc6ccc7cncnc7c6)ncnc5c4)c3c2)CC1",
+      "mw": 570.6610000000001,
+      "logp": 4.502800000000002,
+      "tpsa": 124.95,
+      "hbd": 2,
+      "hba": 10,
+      "rot_bonds": 7,
+      "qed": 0.2833630346138082,
+      "sa_score": 2.9094321238244607,
+      "pred_pAff_mean": 10.92349624633789,
+      "docking_score": -11.13,
+      "docking_status": "completed",
+      "docking_pose_file": "outputs/docking/ligand_0003/ligand_0003.xml"
+    },
+    {
+      "SMILES": "CN1CCCN(CCC(=O)Nc2ccc3nncc(-c4ccc5cncnc5c4)c3c2)C1",
+      "SMILES_state": 1,
+      "NLL": 5.03,
+      "valid": true,
+      "canonical_smiles": "CN1CCCN(CCC(=O)Nc2ccc3nncc(-c4ccc5cncnc5c4)c3c2)C1",
+      "mw": 427.5120000000002,
+      "logp": 3.1636000000000006,
+      "tpsa": 87.14,
+      "hbd": 1,
+      "hba": 7,
+      "rot_bonds": 5,
+      "qed": 0.5234216428527144,
+      "sa_score": 2.7645745083533857,
+      "pred_pAff_mean": 10.726749420166016,
+      "docking_score": null,
+      "docking_status": "not_run",
+      "docking_pose_file": null
+    },
+    {
+      "SMILES": "CN1CCN(CCC(=O)Nc2ccc3nncc(-c4ccc5ncncc5c4)c3c2)CC1",
+      "SMILES_state": 1,
+      "NLL": 6.0,
+      "valid": true,
+      "canonical_smiles": "CN1CCN(CCC(=O)Nc2ccc3nncc(-c4ccc5ncncc5c4)c3c2)CC1",
+      "mw": 427.5120000000002,
+      "logp": 2.8160000000000007,
+      "tpsa": 87.14,
+      "hbd": 1,
+      "hba": 7,
+      "rot_bonds": 5,
+      "qed": 0.5240018720180243,
+      "sa_score": 2.5455327734925,
+      "pred_pAff_mean": 10.25698947906494,
+      "docking_score": null,
+      "docking_status": "not_run",
+      "docking_pose_file": null
+    }
+  ],
+  "warnings": [
+    "Outputs are computational predictions only.",
+    "Docking was run only for a top-k subset."
+  ]
+}

bundle/models/affinity/config.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d12e50c5554fe2c7d13e5db818e78d61ff790b84b0435706e94f52bcb94d5498
+size 513

bundle/models/affinity/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38fd867f9e4d7c352c0f7b60866ecb8e6f1d98c235490498a5ccd6871d23aafd
+size 45788831

bundle/models/affinity/target_sequence.txt

@@ -0,0 +1 @@
+MRPSGTAGAALLALLAALCPASRALEEKKVCQGTSNKLTQLGTFEDHFLSLQRMFNNCEVVLGNLEITYVQRNYDLSFLKTIQEVAGYVLIALNTVERIPLENLQIIRGNMYYENSYALAVLSNYDANKTGLKELPMRNLQEILHGAVRFSNNPALCNVESIQWRDIVSSDFLSNMSMDFQNHLGSCQKCDPSCPNGSCWGAGEENCQKLTKIICAQQCSGRCRGKSPSDCCHNQCAAGCTGPRESDCLVCRKFRDEATCKDTCPPLMLYNPTTYQMDVNPEGKYSFGATCVKKCPRNYVVTDHGSCVRACGADSYEMEEDGVRKCKKCEGPCRKVCNGIGIGEFKDSLSINATNIKHFKNCTSISGDLHILPVAFRGDSFTHTPPLDPQELDILKTVKEITGFLLIQAWPENRTDLHAFENLEIIRGRTKQHGQFSLAVVSLNITSLGLRSLKEISDGDVIISGNKNLCYANTINWKKLFGTSGQKTKIISNRGENSCKATGQVCHALCSPEGCWGPEPRDCVSCRNVSRGRECVDKCNLLEGEPREFVENSECIQCHPECLPQAMNITCTGRGPDNCIQCAHYIDGPHCVKTCPAGVMGENNTLVWKYADAGHVCHLCHPNCTYGCTGPGLEGCPTNGPKIPSIATGMVGALLLLLVVALGIGLFMRRRHIVRKRTLRRLLQERELVEPLTPSGEAPNQALLRILKETEFKKIKVLGSGAFGTVYKGLWIPEGEKVKIPVAIKELREATSPKANKEILDEAYVMASVDNPHVCRLLGICLTSTVQLITQLMPFGCLLDYVREHKDNIGSQYLLNWCVQIAKGMNYLEDRRLVHRDLAARNVLVKTPQHVKITDFGLAKLLGAEEKEYHAEGGKVPIKWMALESILHRIYTHQSDVWSYGVTVWELMTFGSKPYDGIPASEISSILEKGERLPQPPICTIDVYMIMVKCWMIDADSRPKFRELIIEFSKMARDPQRYLVIQGDERMHLPSPTDSNFYRALMDEEDMDDVVDADEYLIPQQGFFSSPSTSRTPLLSSLSATSNNSTVACIDRNGLQSCPIKEDSFLQRYSSDPTGALTEDSIDDTFLPVPEYINQSVPKRPAGSVQNPVYHNQPLNPAPSRDPHYQDPHSTAVGNPEYLNTVQPTCVNSTFDSPAHWAQKGSHQISLDNPDYQQDFFPKEAKPNGIFKGSTAENAEYLRVAPQSSEFIGA

bundle/models/generator/egfr_generator.chkpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b90fc1448a02e5160c784064f5ff8ff846ff56ec2ff50a489d6606342446ade9
+size 23227413

bundle/models/generator/reinvent.prior

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6513ec6dbc54c87ea45cdbf9b4aaefadd7652548b74175366b27f12ec5732fe
+size 23226277

bundle/services/deeppurpose/serve_affinity.py

@@ -0,0 +1,187 @@
+from fastapi import FastAPI, Request, HTTPException
+from fastapi.exceptions import RequestValidationError
+from fastapi.responses import JSONResponse
+from pydantic import BaseModel
+from typing import List
+from pathlib import Path
+from collections import OrderedDict
+from rdkit import Chem
+from DeepPurpose.utils import data_process
+from DeepPurpose import DTI as models
+import traceback
+import csv
+import torch
+
+# --------------------------------------------------
+# Project paths (absolute, derived from this file)
+# --------------------------------------------------
+PROJECT_ROOT = Path(__file__).resolve().parents[2]
+
+MODEL_DIR = PROJECT_ROOT / "models" / "affinity"
+TARGET_SEQUENCE_FILE = PROJECT_ROOT / "models" / "affinity" / "target_sequence.txt"
+
+app = FastAPI(title="EGFR DeepPurpose REINVENT Adapter")
+
+# ---------- Request models ----------
+class PredictRequest(BaseModel):
+    smiles: List[str]
+
+# ---------- Helpers ----------
+def load_default_egfr_sequence(sequence_path: Path) -> str:
+    if not sequence_path.exists():
+        raise FileNotFoundError(f"Target sequence file not found: {sequence_path}")
+
+    seq = sequence_path.read_text(encoding="utf-8").strip()
+    if not seq:
+        raise RuntimeError(f"Target sequence file is empty: {sequence_path}")
+
+    return seq
+
+def canon(smi: str) -> str:
+    try:
+        m = Chem.MolFromSmiles(smi)
+        return Chem.MolToSmiles(m, isomericSmiles=True) if m is not None else smi
+    except Exception:
+        return smi
+
+# ---------- Load model and target ----------
+seq = load_default_egfr_sequence(TARGET_SEQUENCE_FILE)
+model = models.model_pretrained(path_dir=str(MODEL_DIR))
+
+# ---------- Simple cache ----------
+CACHE = OrderedDict()
+CACHE_MAX = 50000
+
+def cache_get(k: str):
+    if k in CACHE:
+        CACHE.move_to_end(k)
+        return CACHE[k]
+    return None
+
+def cache_put(k: str, v: float):
+    CACHE[k] = v
+    CACHE.move_to_end(k)
+    if len(CACHE) > CACHE_MAX:
+        CACHE.popitem(last=False)
+
+# ---------- Shared prediction core ----------
+def predict_smiles_list(smiles_in: List[str]) -> List[float]:
+    if not smiles_in:
+        return []
+
+    out = [None] * len(smiles_in)
+    to_compute = []
+    idx_map = []
+
+    for i, smi in enumerate(smiles_in):
+        k = canon(smi)
+        v = cache_get(k)
+        if v is None:
+            to_compute.append(smi)
+            idx_map.append((i, k))
+        else:
+            out[i] = float(v)
+
+    if to_compute:
+        X_target = [seq] * len(to_compute)
+        y_dummy = [0.0] * len(to_compute)
+
+        ret = data_process(
+            to_compute,
+            X_target,
+            y_dummy,
+            drug_encoding="Morgan",
+            target_encoding="AAC",
+            split_method="no_split",
+        )
+
+        X_pred = ret[0] if isinstance(ret, (tuple, list)) else ret
+
+        with torch.inference_mode():
+            preds = model.predict(X_pred)
+
+        for j, p in enumerate(preds):
+            i, k = idx_map[j]
+            val = float(p)
+            out[i] = val
+            cache_put(k, val)
+
+    return [0.0 if v is None else float(v) for v in out]
+
+# ---------- Helpful debugging ----------
+@app.exception_handler(RequestValidationError)
+async def validation_exception_handler(request: Request, exc: RequestValidationError):
+    body = await request.body()
+    print("\n=== 422 VALIDATION ERROR ===")
+    print("PATH:", request.url.path)
+    print("ERRORS:", exc.errors())
+    print("BODY:", body.decode("utf-8", errors="replace"))
+    print("=== END 422 ===\n")
+    return JSONResponse(status_code=422, content={"detail": exc.errors()})
+
+# ---------- Legacy endpoint ----------
+@app.post("/predict")
+def predict(req: PredictRequest):
+    try:
+        preds = predict_smiles_list(req.smiles or [])
+        return {"pred_pAff_mean": preds}
+    except Exception as e:
+        traceback.print_exc()
+        raise HTTPException(status_code=500, detail=str(e))
+
+# ---------- REINVENT-compatible endpoint ----------
+@app.post("/reinvent_predict")
+async def reinvent_predict(request: Request):
+    try:
+        payload = await request.json()
+
+        # REINVENT in your logs sends {"smiles": [...]}
+        if isinstance(payload, dict) and "smiles" in payload:
+            smiles = payload.get("smiles") or []
+            if not isinstance(smiles, list):
+                raise HTTPException(status_code=422, detail="Field 'smiles' must be a list.")
+            preds = predict_smiles_list(smiles)
+            return {"pred_pAff_mean": preds}
+
+        # Optional compatibility with list-of-items payload
+        if isinstance(payload, list):
+            smiles = []
+            query_ids = []
+            for item in payload:
+                if not isinstance(item, dict):
+                    raise HTTPException(status_code=422, detail="Each list item must be an object.")
+                if "input_string" not in item or "query_id" not in item:
+                    raise HTTPException(
+                        status_code=422,
+                        detail="Each item must contain 'input_string' and 'query_id'."
+                    )
+                smiles.append(item["input_string"])
+                query_ids.append(str(item["query_id"]))
+
+            preds = predict_smiles_list(smiles)
+            successes = [
+                {"query_id": qid, "output_value": float(pred)}
+                for qid, pred in zip(query_ids, preds)
+            ]
+            return {"output": {"successes_list": successes}}
+
+        raise HTTPException(
+            status_code=422,
+            detail="Unsupported request body. Expected either {'smiles': [...]} or a list of {'input_string','query_id'} items."
+        )
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        traceback.print_exc()
+        raise HTTPException(status_code=500, detail=str(e))
+
+# ---------- Health ----------
+@app.get("/health")
+def health():
+    return {
+        "status": "ok",
+        "model_dir": str(MODEL_DIR),
+        "target_sequence_file": str(TARGET_SEQUENCE_FILE),
+        "seq_len": len(seq),
+    }

bundle/tools/dock_enriched.py

@@ -0,0 +1,204 @@
+#!/usr/bin/env python3
+
+import argparse
+import os
+import re
+import shutil
+import subprocess
+from pathlib import Path
+
+import pandas as pd
+from rdkit import Chem
+from rdkit.Chem import AllChem
+
+
+ENERGY_RE = re.compile(
+    r"Estimated Free Energy of Binding\s*=\s*([-+]?\d+(?:\.\d+)?)\s*kcal/mol",
+    re.IGNORECASE,
+)
+
+
+def parse_best_binding_energy(dlg_path: Path):
+    if not dlg_path.exists():
+        return None
+
+    text = dlg_path.read_text(errors="ignore")
+    values = [float(x) for x in ENERGY_RE.findall(text)]
+
+    if not values:
+        return None
+
+    # More negative is better.
+    return min(values)
+
+
+def make_3d_sdf(smiles: str, sdf_path: Path, seed: int = 42):
+    mol = Chem.MolFromSmiles(str(smiles))
+    if mol is None:
+        raise ValueError(f"Invalid SMILES: {smiles}")
+
+    mol = Chem.AddHs(mol)
+
+    params = AllChem.ETKDGv3()
+    params.randomSeed = seed
+
+    status = AllChem.EmbedMolecule(mol, params)
+    if status != 0:
+        raise RuntimeError(f"3D embedding failed for SMILES: {smiles}")
+
+    try:
+        AllChem.UFFOptimizeMolecule(mol, maxIters=500)
+    except Exception:
+        pass
+
+    sdf_path.parent.mkdir(parents=True, exist_ok=True)
+    writer = Chem.SDWriter(str(sdf_path))
+    writer.write(mol)
+    writer.close()
+
+
+def run_cmd(cmd, cwd=None):
+    return subprocess.run(
+        cmd,
+        cwd=str(cwd) if cwd else None,
+        text=True,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.STDOUT,
+        check=False,
+    )
+
+
+def dock_one(smiles, ligand_id, work_dir, adgpu_bin, grid_file, nrun):
+    work_dir.mkdir(parents=True, exist_ok=True)
+
+    sdf_path = work_dir / f"{ligand_id}.sdf"
+    pdbqt_path = work_dir / f"{ligand_id}.pdbqt"
+
+    make_3d_sdf(smiles, sdf_path)
+
+    mk_prepare = shutil.which("mk_prepare_ligand.py")
+    if mk_prepare is None:
+        raise RuntimeError("mk_prepare_ligand.py was not found in PATH.")
+
+    prep = run_cmd([
+        mk_prepare,
+        "-i", str(sdf_path),
+        "-o", str(pdbqt_path),
+    ])
+
+    if prep.returncode != 0 or not pdbqt_path.exists():
+        raise RuntimeError(f"Meeko ligand preparation failed:\n{prep.stdout}")
+
+    dock = run_cmd([
+        adgpu_bin,
+        "--ffile", str(grid_file.resolve()),
+        "--lfile", str(pdbqt_path.resolve()),
+        "--nrun", str(nrun),
+    ], cwd=work_dir)
+
+    if dock.returncode != 0:
+        raise RuntimeError(f"AutoDock-GPU failed:\n{dock.stdout}")
+
+    candidates = sorted(work_dir.glob("*.dlg"))
+    if not candidates:
+        raise RuntimeError(f"No DLG file was produced in {work_dir}")
+
+    dlg_path = candidates[0]
+    score = parse_best_binding_energy(dlg_path)
+
+    if score is None:
+        raise RuntimeError(f"Could not parse binding energy from {dlg_path}")
+
+    xml_candidates = sorted(work_dir.glob("*.xml"))
+    result_file = xml_candidates[0] if xml_candidates else dlg_path
+
+    return score, str(result_file)
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--input", required=True, help="Input enriched CSV.")
+    parser.add_argument("--output", required=True, help="Output CSV with docking columns.")
+    parser.add_argument("--docking-mode", choices=["off", "top_k", "all"], default="off")
+    parser.add_argument("--dock-top-k", type=int, default=10)
+    parser.add_argument("--nrun", type=int, default=8)
+    parser.add_argument("--adgpu-bin", default=os.environ.get("ADGPU_BIN"))
+    parser.add_argument("--grid-file", default="docking/maps_current/4WKQ_receptor_v5_SBr.maps.fld")
+    parser.add_argument("--work-dir", default="outputs/docking")
+    args = parser.parse_args()
+
+    df = pd.read_csv(args.input)
+
+    # Ensure stable dtypes for docking columns.
+    if "docking_score" not in df.columns:
+        df["docking_score"] = None
+    if "docking_status" not in df.columns:
+        df["docking_status"] = "not_run"
+    if "docking_pose_file" not in df.columns:
+        df["docking_pose_file"] = None
+
+    df["docking_status"] = df["docking_status"].astype("object")
+    df["docking_pose_file"] = df["docking_pose_file"].astype("object")
+
+    if args.docking_mode == "off":
+        df["docking_score"] = None
+        df["docking_status"] = "not_run"
+        df["docking_pose_file"] = None
+        Path(args.output).parent.mkdir(parents=True, exist_ok=True)
+        df.to_csv(args.output, index=False)
+        print(f"Docking mode off. Wrote: {args.output}")
+        return
+
+    if not args.adgpu_bin:
+        raise RuntimeError("Set ADGPU_BIN or pass --adgpu-bin.")
+
+    adgpu_bin = Path(args.adgpu_bin)
+    if not adgpu_bin.exists():
+        raise FileNotFoundError(f"AutoDock-GPU binary not found: {adgpu_bin}")
+
+    grid_file = Path(args.grid_file)
+    if not grid_file.exists():
+        raise FileNotFoundError(f"Grid file not found: {grid_file}")
+
+    if "canonical_smiles" not in df.columns:
+        raise ValueError("Input CSV must contain canonical_smiles column.")
+
+    if args.docking_mode == "all":
+        indices = list(df.index)
+    else:
+        indices = list(df.index[: args.dock_top_k])
+
+    work_root = Path(args.work_dir)
+    work_root.mkdir(parents=True, exist_ok=True)
+
+    for count, idx in enumerate(indices, start=1):
+        smiles = df.at[idx, "canonical_smiles"]
+        ligand_id = f"ligand_{count:04d}"
+
+        try:
+            score, result_file = dock_one(
+                smiles=smiles,
+                ligand_id=ligand_id,
+                work_dir=work_root / ligand_id,
+                adgpu_bin=str(adgpu_bin),
+                grid_file=grid_file,
+                nrun=args.nrun,
+            )
+            df.at[idx, "docking_score"] = score
+            df.at[idx, "docking_status"] = "completed"
+            df.at[idx, "docking_pose_file"] = result_file
+            print(f"[OK] {ligand_id}: {score:.2f} kcal/mol")
+
+        except Exception as e:
+            df.at[idx, "docking_score"] = None
+            df.at[idx, "docking_status"] = f"failed: {str(e)[:160]}"
+            df.at[idx, "docking_pose_file"] = None
+            print(f"[FAILED] {ligand_id}: {e}")
+
+    Path(args.output).parent.mkdir(parents=True, exist_ok=True)
+    df.to_csv(args.output, index=False)
+    print(f"Wrote: {args.output}")
+
+
+if __name__ == "__main__":
+    main()

bundle/tools/enrich_generated.py

@@ -0,0 +1,191 @@
+#!/usr/bin/env python3
+
+import argparse
+import json
+from pathlib import Path
+
+import pandas as pd
+import requests
+import os
+import sys
+
+from rdkit import Chem
+from rdkit.Chem import Descriptors, Crippen, Lipinski, rdMolDescriptors, QED
+
+
+def canonicalize(smiles: str):
+    mol = Chem.MolFromSmiles(str(smiles))
+    if mol is None:
+        return None, None
+    can = Chem.MolToSmiles(mol, isomericSmiles=True)
+    return mol, can
+
+
+def try_sa_score(mol):
+    """
+    Compute RDKit SA score.
+
+    RDKit Contrib is installed in different locations depending on the package source.
+    Common conda location:
+        $CONDA_PREFIX/share/RDKit/Contrib/SA_Score/sascorer.py
+
+    Lower score means easier synthesis.
+    Higher score means harder synthesis.
+    """
+    import_paths = []
+
+    # First try the normal Python import.
+    try:
+        from rdkit.Contrib.SA_Score import sascorer
+        return float(sascorer.calculateScore(mol))
+    except Exception:
+        pass
+
+    # Then search common RDKit Contrib locations.
+    conda_prefix = os.environ.get("CONDA_PREFIX")
+    if conda_prefix:
+        import_paths.append(Path(conda_prefix) / "share" / "RDKit" / "Contrib" / "SA_Score")
+
+    import_paths.extend([
+        Path("/opt/conda/envs/ailixir/share/RDKit/Contrib/SA_Score"),
+        Path("/opt/conda/share/RDKit/Contrib/SA_Score"),
+        Path("/usr/share/RDKit/Contrib/SA_Score"),
+        Path("/usr/local/share/RDKit/Contrib/SA_Score"),
+    ])
+
+    for p in import_paths:
+        sascorer_file = p / "sascorer.py"
+        fpscores_file = p / "fpscores.pkl.gz"
+
+        if sascorer_file.exists() and fpscores_file.exists():
+            sys.path.insert(0, str(p))
+            try:
+                import sascorer
+                return float(sascorer.calculateScore(mol))
+            except Exception:
+                continue
+
+    return None
+
+
+def compute_properties(smiles: str):
+    mol, can = canonicalize(smiles)
+
+    if mol is None:
+        return {
+            "valid": False,
+            "canonical_smiles": None,
+            "mw": None,
+            "logp": None,
+            "tpsa": None,
+            "hbd": None,
+            "hba": None,
+            "rot_bonds": None,
+            "qed": None,
+            "sa_score": None,
+        }
+
+    return {
+        "valid": True,
+        "canonical_smiles": can,
+        "mw": float(Descriptors.MolWt(mol)),
+        "logp": float(Crippen.MolLogP(mol)),
+        "tpsa": float(rdMolDescriptors.CalcTPSA(mol)),
+        "hbd": int(Lipinski.NumHDonors(mol)),
+        "hba": int(Lipinski.NumHAcceptors(mol)),
+        "rot_bonds": int(Lipinski.NumRotatableBonds(mol)),
+        "qed": float(QED.qed(mol)),
+        "sa_score": try_sa_score(mol),
+    }
+
+
+def call_affinity_api(smiles_list, url: str):
+    if not smiles_list:
+        return []
+
+    try:
+        response = requests.post(
+            url,
+            json={"smiles": smiles_list},
+            timeout=300,
+        )
+        response.raise_for_status()
+
+        payload = response.json()
+        preds = payload.get("pred_pAff_mean")
+
+        if preds is None:
+            raise RuntimeError(
+                f"Affinity API response missing 'pred_pAff_mean'. Response was: {json.dumps(payload)[:500]}"
+            )
+
+        return [float(x) for x in preds]
+
+    except Exception as e:
+        print(f"Warning: Affinity API call failed ({e}). Returning null predictions.")
+        return [None] * len(smiles_list)
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--input", required=True, help="Input generated_smiles.csv from REINVENT sampling.")
+    parser.add_argument("--output", required=True, help="Output enriched CSV.")
+    parser.add_argument(
+        "--affinity-url",
+        default="http://127.0.0.1:8001/reinvent_predict",
+        help="DeepPurpose/FastAPI affinity endpoint.",
+    )
+    parser.add_argument(
+        "--top-k",
+        type=int,
+        default=0,
+        help="If >0, keep only top K rows after sorting by pred_pAff_mean desc then QED desc.",
+    )
+    args = parser.parse_args()
+
+    input_path = Path(args.input)
+    output_path = Path(args.output)
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+
+    df = pd.read_csv(input_path)
+
+    if "SMILES" not in df.columns:
+        raise ValueError(f"Input file must contain a SMILES column. Found columns: {list(df.columns)}")
+
+    props = [compute_properties(smi) for smi in df["SMILES"].tolist()]
+    props_df = pd.DataFrame(props)
+
+    out = pd.concat([df.reset_index(drop=True), props_df.reset_index(drop=True)], axis=1)
+
+    valid_mask = out["valid"] == True
+    valid_smiles = out.loc[valid_mask, "canonical_smiles"].tolist()
+
+    pred_values = call_affinity_api(valid_smiles, args.affinity_url)
+
+    out["pred_pAff_mean"] = None
+    out.loc[valid_mask, "pred_pAff_mean"] = pred_values
+
+    # Docking is optional in v1. Keep stable columns for frontend/backend filters.
+    out["docking_score"] = None
+    out["docking_status"] = "not_run"
+    out["docking_pose_file"] = None
+
+    # Sort for display only.
+    # This is not a final scientific ranking.
+    out = out.sort_values(
+        by=["pred_pAff_mean", "qed"],
+        ascending=[False, False],
+    )
+
+    if args.top_k and args.top_k > 0:
+        out = out.head(args.top_k)
+
+    out.to_csv(output_path, index=False)
+
+    print(f"Wrote: {output_path}")
+    print(f"Rows: {len(out)}")
+    print("Columns:", ", ".join(out.columns))
+
+
+if __name__ == "__main__":
+    main()

start_hf.sh

@@ -0,0 +1,63 @@
+#!/usr/bin/env bash
+set -euo pipefail
+mkdir -p /app/outputs/jobs
+mkdir -p /app/outputs/deeppurpose
+mkdir -p /home/abdullah/projects/egfr_drug_discovery/runs/deeppurpose
+
+export PORT="${PORT:-7860}"
+export PUBLIC_BASE_URL="${PUBLIC_BASE_URL:-http://localhost:${PORT}}"
+export DEEPPURPOSE_URL="${DEEPPURPOSE_URL:-http://127.0.0.1:8001/reinvent_predict}"
+export REINVENT_DEVICE="${REINVENT_DEVICE:-cpu}"
+export PYTHON="${PYTHON:-python}"
+
+if [ -x "/opt/AutoDock-GPU/bin/autodock_cpu_1wi" ]; then
+  export ADGPU_BIN="${ADGPU_BIN:-/opt/AutoDock-GPU/bin/autodock_cpu_1wi}"
+elif [ -x "/opt/AutoDock-GPU/bin/autodock_gpu_64wi" ]; then
+  export ADGPU_BIN="${ADGPU_BIN:-/opt/AutoDock-GPU/bin/autodock_gpu_64wi}"
+else
+  export ADGPU_BIN="${ADGPU_BIN:-}"
+fi
+
+echo "PORT=${PORT}"
+echo "PUBLIC_BASE_URL=${PUBLIC_BASE_URL}"
+echo "DEEPPURPOSE_URL=${DEEPPURPOSE_URL}"
+echo "REINVENT_DEVICE=${REINVENT_DEVICE}"
+echo "ADGPU_BIN=${ADGPU_BIN}"
+
+echo "Starting DeepPurpose affinity service on 127.0.0.1:8001..."
+conda run --no-capture-output -n dp \
+  uvicorn services.deeppurpose.serve_affinity:app \
+  --app-dir /app/bundle \
+  --host 127.0.0.1 \
+  --port 8001 &
+
+AFFINITY_PID=$!
+
+echo "Waiting for DeepPurpose affinity service..."
+conda run --no-capture-output -n ailixir python - <<'PY'
+import time
+import sys
+import requests
+
+url = "http://127.0.0.1:8001/health"
+
+for i in range(180):
+    try:
+        r = requests.get(url, timeout=3)
+        print(f"Affinity health attempt {i+1}: {r.status_code}")
+        if r.status_code == 200:
+            print("Affinity service is ready.")
+            sys.exit(0)
+    except Exception as e:
+        print(f"Affinity health attempt {i+1} failed: {e}")
+    time.sleep(2)
+
+print("Affinity service did not become ready.")
+sys.exit(1)
+PY
+
+echo "Starting Generation API on 0.0.0.0:${PORT}..."
+exec conda run --no-capture-output -n ailixir \
+  uvicorn api:app \
+  --host 0.0.0.0 \
+  --port "${PORT}"