{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.12.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"none","dataSources":[{"sourceId":118765,"databundleVersionId":15231210,"sourceType":"competition"},{"sourceId":14604295,"sourceType":"datasetVersion","datasetId":9328538},{"sourceId":224830487,"sourceType":"kernelVersion"},{"sourceId":242177530,"sourceType":"kernelVersion"},{"sourceId":290004465,"sourceType":"kernelVersion"},{"sourceId":291182720,"sourceType":"kernelVersion"}],"dockerImageVersionId":31259,"isInternetEnabled":false,"language":"python","sourceType":"notebook","isGpuEnabled":false}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# Stanford RNA 3D Folding Part 2 - TBM Baseline\n\nThis notebook is a template-based modeling (TBM) baseline for the Stanford RNA 3D Folding Part 2 competition.\n\nInputs required (Kaggle notebook):\n- Competition dataset: Stanford RNA 3D Folding Part 2 (already attached)\n- Optional: a dataset containing a `biopython*.whl` wheel (offline install)\n- Optional: `tm-score-permutechains` + `usalign` datasets for local validation scoring\n\nRun order:\n1. Setup and data loading\n2. Baseline TBM prediction\n3. (Optional) Validation scoring","metadata":{}},{"cell_type":"code","source":"import os\nimport sys\nimport glob\nimport time\nimport random\nimport warnings\nimport subprocess\n\nimport numpy as np\nimport pandas as pd\n\nwarnings.filterwarnings(\"ignore\")\n\n# -----------------------------\n# 0) Dependency setup (offline)\n# -----------------------------\n\ndef ensure_biopython():\n    try:\n        import Bio  # noqa: F401\n        return\n    except Exception:\n        pass\n\n    whls = glob.glob(\"/kaggle/input/**/biopython*.whl\", recursive=True)\n    if whls:\n        whl = whls[0]\n        print(f\"Installing Biopython from {whl}\")\n        subprocess.check_call([sys.executable, \"-m\", \"pip\", \"install\", \"--no-index\", whl])\n        return\n\n    raise RuntimeError(\n        \"Biopython not found. Attach a dataset containing a biopython*.whl wheel.\"\n    )\n\nensure_biopython()\n\nfrom Bio.Align import PairwiseAligner","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-02-13T15:20:18.9136Z","iopub.execute_input":"2026-02-13T15:20:18.914091Z","iopub.status.idle":"2026-02-13T15:20:49.725652Z","shell.execute_reply.started":"2026-02-13T15:20:18.914033Z","shell.execute_reply":"2026-02-13T15:20:49.724346Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# -----------------------------\n# 1) Data ingestion\n# -----------------------------\nDATA_PATH = \"/kaggle/input/stanford-rna-3d-folding-2/\"\n\ntrain_seqs   = pd.read_csv(DATA_PATH + \"train_sequences.csv\")\ntrain_labels = pd.read_csv(DATA_PATH + \"train_labels.csv\")\ntest_seqs    = pd.read_csv(DATA_PATH + \"test_sequences.csv\")\n\n# Optional validation files if present\nvalid_seqs_path = DATA_PATH + \"validation_sequences.csv\"\nvalid_labels_path = DATA_PATH + \"validation_labels.csv\"\nvalid_seqs = pd.read_csv(valid_seqs_path) if os.path.exists(valid_seqs_path) else None\nvalid_labels = pd.read_csv(valid_labels_path) if os.path.exists(valid_labels_path) else None\n\nprint(f\"Train sequences: {len(train_seqs)}\")\nprint(f\"Train labels:    {len(train_labels)}\")\nprint(f\"Test sequences:  {len(test_seqs)}\")\nif valid_seqs is not None:\n    print(f\"Valid sequences: {len(valid_seqs)}\")\nif valid_labels is not None:\n    print(f\"Valid labels:    {len(valid_labels)}\")\n\n# Try to import the provided parse_fasta helper\nsys.path.append(os.path.join(DATA_PATH, \"extra\"))\n\ntry:\n    import typing as _typing\n    import builtins as _builtins\n    _builtins.Dict  = getattr(_typing, \"Dict\")\n    _builtins.Tuple = getattr(_typing, \"Tuple\")\n    _builtins.List  = getattr(_typing, \"List\")\n\n    from parse_fasta_py import parse_fasta as _parse_fasta_raw\n\n    def parse_fasta(fasta_content: str):\n        d = _parse_fasta_raw(fasta_content)\n        out = {}\n        for k, v in d.items():\n            out[k] = v[0] if isinstance(v, tuple) else v\n        return out\n\nexcept Exception:\n    def parse_fasta(fasta_content: str):\n        out = {}\n        cur = None\n        seq_parts = []\n        for line in str(fasta_content).splitlines():\n            line = line.strip()\n            if not line:\n                continue\n            if line.startswith(\">\"):\n                if cur is not None:\n                    out[cur] = \"\".join(seq_parts)\n                header = line[1:]\n                cur = header.split()[0]\n                seq_parts = []\n            else:\n                seq_parts.append(line.replace(\" \", \"\"))\n        if cur is not None:\n            out[cur] = \"\".join(seq_parts)\n        return out\n\n\ndef parse_stoichiometry(stoich: str):\n    if pd.isna(stoich) or str(stoich).strip() == \"\":\n        return []\n    out = []\n    for part in str(stoich).split(\";\"):\n        ch, cnt = part.split(\":\")\n        out.append((ch.strip(), int(cnt)))\n    return out\n\n\ndef get_chain_segments(row):\n    \"\"\"\n    Returns list of (start,end) segments in row['sequence'] corresponding to chain copies\n    in stoichiometry order. If parsing fails or mismatch: fallback single segment (0,L).\n    \"\"\"\n    seq = row[\"sequence\"]\n    stoich = row.get(\"stoichiometry\", \"\")\n    all_seq = row.get(\"all_sequences\", \"\")\n\n    if pd.isna(stoich) or pd.isna(all_seq) or str(stoich).strip() == \"\" or str(all_seq).strip() == \"\":\n        return [(0, len(seq))]\n\n    try:\n        chain_dict = parse_fasta(all_seq)  # chain_id -> sequence\n        order = parse_stoichiometry(stoich)\n\n        segs = []\n        pos = 0\n        for ch, cnt in order:\n            base = chain_dict.get(ch)\n            if base is None:\n                return [(0, len(seq))]\n            for _ in range(cnt):\n                L = len(base)\n                segs.append((pos, pos + L))\n                pos += L\n\n        if pos != len(seq):\n            return [(0, len(seq))]\n        return segs\n    except Exception:\n        return [(0, len(seq))]\n\n\ndef build_segments_map(df):\n    seg_map = {}\n    for _, r in df.iterrows():\n        tid = r[\"target_id\"]\n        seg_map[tid] = get_chain_segments(r)\n    return seg_map\n\ntrain_segs_map = build_segments_map(train_seqs)\ntest_segs_map  = build_segments_map(test_seqs)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-02-13T15:20:49.728558Z","iopub.execute_input":"2026-02-13T15:20:49.729119Z","iopub.status.idle":"2026-02-13T15:21:03.131284Z","shell.execute_reply.started":"2026-02-13T15:20:49.729088Z","shell.execute_reply":"2026-02-13T15:21:03.130218Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# -----------------------------\n# 2) Labels to templates\n# -----------------------------\n\ndef process_labels(labels_df: pd.DataFrame):\n    \"\"\"\n    train_coords_dict: {target_id: (L,3) coords from x_1,y_1,z_1}\n    Group key = labels_df['ID'] split at last underscore\n    \"\"\"\n    coords_dict = {}\n    prefixes = labels_df[\"ID\"].str.rsplit(\"_\", n=1).str[0]\n    for id_prefix, group in labels_df.groupby(prefixes, sort=False):\n        coords_dict[id_prefix] = group.sort_values(\"resid\")[[\"x_1\", \"y_1\", \"z_1\"]].values\n    return coords_dict\n\ntrain_coords_dict = process_labels(train_labels)\nprint(f\"Templates available: {len(train_coords_dict)}\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-02-13T15:21:03.133373Z","iopub.execute_input":"2026-02-13T15:21:03.133806Z","iopub.status.idle":"2026-02-13T15:21:19.366967Z","shell.execute_reply.started":"2026-02-13T15:21:03.133766Z","shell.execute_reply":"2026-02-13T15:21:19.365561Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# -----------------------------\n# 3) Similarity search (global alignment)\n# -----------------------------\n\naligner = PairwiseAligner()\naligner.mode = \"global\"\naligner.match_score = 2.0\naligner.mismatch_score = -1.5\naligner.open_gap_score = -8.0\naligner.extend_gap_score = -0.4\n\n# Explicit terminal gap penalties (avoid end-gap sliding)\naligner.query_left_open_gap_score = -8\naligner.query_left_extend_gap_score = -0.4\naligner.query_right_open_gap_score = -8\naligner.query_right_extend_gap_score = -0.4\naligner.target_left_open_gap_score = -8\naligner.target_left_extend_gap_score = -0.4\naligner.target_right_open_gap_score = -8\naligner.target_right_extend_gap_score = -0.4\n\n\ndef find_similar_sequences(query_seq: str, train_seqs_df: pd.DataFrame, train_coords_dict: dict, top_n: int = 5):\n    similar_seqs = []\n    for _, row in train_seqs_df.iterrows():\n        target_id = row[\"target_id\"]\n        train_seq = row[\"sequence\"]\n        if target_id not in train_coords_dict:\n            continue\n\n        # Length filter to keep search tractable\n        if abs(len(train_seq) - len(query_seq)) / max(len(train_seq), len(query_seq)) > 0.3:\n            continue\n\n        raw_score = aligner.score(query_seq, train_seq)\n        normalized_score = raw_score / (2 * min(len(query_seq), len(train_seq)))\n        similar_seqs.append((target_id, train_seq, normalized_score, train_coords_dict[target_id]))\n\n    similar_seqs.sort(key=lambda x: x[2], reverse=True)\n    return similar_seqs[:top_n]","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-02-13T15:21:19.368519Z","iopub.execute_input":"2026-02-13T15:21:19.368923Z","iopub.status.idle":"2026-02-13T15:21:19.381353Z","shell.execute_reply.started":"2026-02-13T15:21:19.368888Z","shell.execute_reply":"2026-02-13T15:21:19.379359Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# -----------------------------\n# 4) Template transfer + gap fill\n# -----------------------------\n\n\ndef adapt_template_to_query(query_seq: str, template_seq: str, template_coords: np.ndarray):\n    alignment = next(iter(aligner.align(query_seq, template_seq)))\n    new_coords = np.full((len(query_seq), 3), np.nan, dtype=float)\n\n    # Vectorized chunk mapping via aligned blocks\n    for (q_start, q_end), (t_start, t_end) in zip(*alignment.aligned):\n        t_chunk = template_coords[t_start:t_end]\n        if len(t_chunk) == (q_end - q_start):\n            new_coords[q_start:q_end] = t_chunk\n\n    # Fill unmatched residues by interpolation / edge-fill / fallback line\n    for i in range(len(new_coords)):\n        if np.isnan(new_coords[i, 0]):\n            prev_v = next((j for j in range(i - 1, -1, -1) if not np.isnan(new_coords[j, 0])), -1)\n            next_v = next((j for j in range(i + 1, len(new_coords)) if not np.isnan(new_coords[j, 0])), -1)\n\n            if prev_v >= 0 and next_v >= 0:\n                w = (i - prev_v) / (next_v - prev_v)\n                new_coords[i] = (1 - w) * new_coords[prev_v] + w * new_coords[next_v]\n            elif prev_v >= 0:\n                new_coords[i] = new_coords[prev_v] + [3, 0, 0]\n            elif next_v >= 0:\n                new_coords[i] = new_coords[next_v] + [3, 0, 0]\n            else:\n                new_coords[i] = [i * 3, 0, 0]\n\n    return np.nan_to_num(new_coords)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-02-13T15:21:19.382871Z","iopub.execute_input":"2026-02-13T15:21:19.383281Z","iopub.status.idle":"2026-02-13T15:21:19.457956Z","shell.execute_reply.started":"2026-02-13T15:21:19.383242Z","shell.execute_reply":"2026-02-13T15:21:19.456574Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# -----------------------------\n# 5) Segment-aware refinement\n# -----------------------------\n\n\ndef adaptive_rna_constraints(coordinates: np.ndarray, target_id: str, confidence: float = 1.0, passes: int = 2):\n    \"\"\"\n    Apply within each chain segment only (no smoothing across chain breaks).\n    - i,i+1 bond target ~5.95 A\n    - i,i+2 target ~10.2 A\n    - Laplacian smoothing\n    - light self-avoidance on subsampled points\n    Strength increases when confidence is lower.\n    \"\"\"\n    coords = coordinates.copy()\n    segments = test_segs_map.get(target_id, [(0, len(coords))])\n\n    strength = 0.75 * (1.0 - min(confidence, 0.90))\n    strength = max(strength, 0.02)\n\n    for _ in range(passes):\n        for (s, e) in segments:\n            X = coords[s:e]\n            L = e - s\n            if L < 3:\n                coords[s:e] = X\n                continue\n\n            # (1) bond i,i+1\n            d = X[1:] - X[:-1]\n            dist = np.linalg.norm(d, axis=1) + 1e-6\n            target = 5.95\n            scale = (target - dist) / dist\n            adj = (d * scale[:, None]) * (0.22 * strength)\n            X[:-1] -= adj\n            X[1:]  += adj\n\n            # (2) soft i,i+2\n            d2 = X[2:] - X[:-2]\n            dist2 = np.linalg.norm(d2, axis=1) + 1e-6\n            target2 = 10.2\n            scale2 = (target2 - dist2) / dist2\n            adj2 = (d2 * scale2[:, None]) * (0.10 * strength)\n            X[:-2] -= adj2\n            X[2:]  += adj2\n\n            # (3) Laplacian smoothing\n            lap = 0.5 * (X[:-2] + X[2:]) - X[1:-1]\n            X[1:-1] += (0.06 * strength) * lap\n\n            # (4) light self-avoidance (subsample)\n            if L >= 25:\n                k = min(L, 160) if L > 220 else L\n                idx = np.linspace(0, L - 1, k).astype(int) if k < L else np.arange(L)\n\n                P = X[idx]\n                diff = P[:, None, :] - P[None, :, :]\n                distm = np.linalg.norm(diff, axis=2) + 1e-6\n                sep = np.abs(idx[:, None] - idx[None, :])\n\n                mask = (sep > 2) & (distm < 3.2)\n                if np.any(mask):\n                    force = (3.2 - distm) / distm\n                    vec = (diff * force[:, :, None] * mask[:, :, None]).sum(axis=1)\n                    X[idx] += (0.015 * strength) * vec\n\n            coords[s:e] = X\n\n    return coords","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-02-13T15:21:19.460245Z","iopub.execute_input":"2026-02-13T15:21:19.46117Z","iopub.status.idle":"2026-02-13T15:21:19.487892Z","shell.execute_reply.started":"2026-02-13T15:21:19.461126Z","shell.execute_reply":"2026-02-13T15:21:19.486616Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# -----------------------------\n# 6) Diversity transforms + prediction\n# -----------------------------\n\n\ndef _rotmat(axis, ang):\n    axis = np.asarray(axis, float)\n    axis = axis / (np.linalg.norm(axis) + 1e-12)\n    x, y, z = axis\n    c, s = np.cos(ang), np.sin(ang)\n    C = 1.0 - c\n    return np.array(\n        [\n            [c + x * x * C,     x * y * C - z * s, x * z * C + y * s],\n            [y * x * C + z * s, c + y * y * C,     y * z * C - x * s],\n            [z * x * C - y * s, z * y * C + x * s, c + z * z * C],\n        ],\n        dtype=float,\n    )\n\n\ndef apply_hinge(coords, seg, rng, max_angle_deg=25):\n    s, e = seg\n    L = e - s\n    if L < 30:\n        return coords\n    pivot = s + int(rng.integers(10, L - 10))\n    axis = rng.normal(size=3)\n    ang = np.deg2rad(float(rng.uniform(-max_angle_deg, max_angle_deg)))\n    R = _rotmat(axis, ang)\n    X = coords.copy()\n    p0 = X[pivot].copy()\n    X[pivot + 1 : e] = (X[pivot + 1 : e] - p0) @ R.T + p0\n    return X\n\n\ndef jitter_chains(coords, segments, rng, max_angle_deg=12, max_trans=1.5):\n    X = coords.copy()\n    global_center = X.mean(axis=0, keepdims=True)\n    for (s, e) in segments:\n        axis = rng.normal(size=3)\n        ang = np.deg2rad(float(rng.uniform(-max_angle_deg, max_angle_deg)))\n        R = _rotmat(axis, ang)\n        shift = rng.normal(size=3)\n        shift = shift / (np.linalg.norm(shift) + 1e-12) * float(rng.uniform(0.0, max_trans))\n        c = X[s:e].mean(axis=0, keepdims=True)\n        X[s:e] = (X[s:e] - c) @ R.T + c + shift\n    X -= X.mean(axis=0, keepdims=True) - global_center\n    return X\n\n\ndef smooth_wiggle(coords, segments, rng, amp=0.8):\n    X = coords.copy()\n    for (s, e) in segments:\n        L = e - s\n        if L < 20:\n            continue\n        n_ctrl = 6\n        ctrl_x = np.linspace(0, L - 1, n_ctrl)\n        ctrl_disp = rng.normal(0, amp, size=(n_ctrl, 3))\n        t = np.arange(L)\n        disp = np.vstack([np.interp(t, ctrl_x, ctrl_disp[:, k]) for k in range(3)]).T\n        X[s:e] += disp\n    return X\n\n\ndef predict_rna_structures(row, train_seqs_df, train_coords_dict, n_predictions=5):\n    tid = row[\"target_id\"]\n    seq = row[\"sequence\"]\n\n    segments = test_segs_map.get(tid, [(0, len(seq))])\n\n    # Candidate pool top_n=30 then sample for diversity\n    cands = find_similar_sequences(\n        query_seq=seq, train_seqs_df=train_seqs_df, train_coords_dict=train_coords_dict, top_n=30\n    )\n\n    predictions = []\n    used = set()\n\n    for i in range(n_predictions):\n        seed = (abs(hash(tid)) + i * 10007) % (2**32)\n        rng = np.random.default_rng(seed)\n\n        if not cands:\n            # Hard fallback: straight line per chain segment\n            coords = np.zeros((len(seq), 3), dtype=float)\n            for (s, e) in segments:\n                for j in range(s + 1, e):\n                    coords[j] = coords[j - 1] + [5.95, 0, 0]\n            predictions.append(coords)\n            continue\n\n        # Template choice\n        if i == 0:\n            t_id, t_seq, sim, t_coords = cands[0]\n        else:\n            K = min(12, len(cands))\n            sims = np.array([cands[k][2] for k in range(K)], float)\n            w = np.exp((sims - sims.max()) / 0.08)\n            for k in range(K):\n                if cands[k][0] in used:\n                    w[k] *= 0.10\n            w = w / (w.sum() + 1e-12)\n            k = int(rng.choice(np.arange(K), p=w))\n            t_id, t_seq, sim, t_coords = cands[k]\n\n        used.add(t_id)\n\n        # Transfer coords (no sliding)\n        adapted = adapt_template_to_query(query_seq=seq, template_seq=t_seq, template_coords=t_coords)\n\n        # Diversity transforms, then refinement\n        if i == 0:\n            X = adapted\n        elif i == 1:\n            X = adapted + rng.normal(0, max(0.01, (0.40 - sim) * 0.06), adapted.shape)\n        elif i == 2:\n            longest = max(segments, key=lambda se: se[1] - se[0])\n            X = apply_hinge(adapted, longest, rng, max_angle_deg=22)\n        elif i == 3:\n            X = jitter_chains(adapted, segments, rng, max_angle_deg=10, max_trans=1.0)\n        else:\n            X = smooth_wiggle(adapted, segments, rng, amp=0.7)\n\n        refined = adaptive_rna_constraints(X, tid, confidence=sim, passes=2)\n        predictions.append(refined)\n\n    return predictions","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-02-13T15:21:19.49056Z","iopub.execute_input":"2026-02-13T15:21:19.490898Z","iopub.status.idle":"2026-02-13T15:21:19.515606Z","shell.execute_reply.started":"2026-02-13T15:21:19.490873Z","shell.execute_reply":"2026-02-13T15:21:19.514257Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# -----------------------------\n# 7) Generate submission\n# -----------------------------\n\n# Set MAX_TEST to a small number for quick smoke tests\nMAX_TEST = None  # e.g., 3\n\nall_predictions = []\nstart_time = time.time()\n\nrows = test_seqs if MAX_TEST is None else test_seqs.head(MAX_TEST)\n\nfor idx, row in rows.iterrows():\n    if idx % 10 == 0:\n        print(f\"Processing {idx} | {time.time() - start_time:.1f}s\")\n    tid = row[\"target_id\"]\n    seq = row[\"sequence\"]\n\n    preds = predict_rna_structures(row, train_seqs, train_coords_dict, n_predictions=5)\n\n    # Safety: each prediction must be (L,3)\n    L = len(seq)\n    for p in preds:\n        assert isinstance(p, np.ndarray) and p.shape == (L, 3), f\"Bad pred shape for {tid}: {getattr(p,'shape',None)}\"\n        assert np.isfinite(p).all(), f\"Non-finite coords in {tid}\"\n\n    for j in range(L):\n        res = {\"ID\": f\"{tid}_{j+1}\", \"resname\": seq[j], \"resid\": j + 1}\n        for i in range(5):\n            res[f\"x_{i+1}\"], res[f\"y_{i+1}\"], res[f\"z_{i+1}\"] = preds[i][j]\n        all_predictions.append(res)\n\nsub = pd.DataFrame(all_predictions)\n\ncols = [\"ID\", \"resname\", \"resid\"] + [f\"{c}_{i}\" for i in range(1, 6) for c in [\"x\", \"y\", \"z\"]]\n\n# Clip explicitly (competition clips coords; prevent explosions)\ncoord_cols = [c for c in cols if c.startswith((\"x_\", \"y_\", \"z_\"))]\nsub[coord_cols] = sub[coord_cols].clip(-999.999, 9999.999)\n\nsub[cols].to_csv(\"submission.csv\", index=False)\nprint(\"submission.csv saved\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-02-13T15:21:19.51718Z","iopub.execute_input":"2026-02-13T15:21:19.517565Z","iopub.status.idle":"2026-02-13T15:23:50.28498Z","shell.execute_reply.started":"2026-02-13T15:21:19.517529Z","shell.execute_reply":"2026-02-13T15:23:50.283612Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# -----------------------------\n# 8) Optional: validation scoring (requires extra inputs)\n# -----------------------------\n\nmetric_path = \"/kaggle/input/tm-score-permutechains/metric.py\"\nusalign_path = \"/kaggle/input/usalign/USalign\"\n\nif os.path.exists(metric_path) and os.path.exists(usalign_path) and valid_labels is not None:\n    import runpy\n\n    module_globals = runpy.run_path(metric_path)\n    score = module_globals[\"score\"]\n\n    sol = valid_labels.copy()\n    sub = pd.read_csv(\"/kaggle/working/submission.csv\")\n\n    # target_id is ID without the residue suffix\n    sol[\"target_id\"] = sol[\"ID\"].apply(lambda x: \"_\".join(str(x).split(\"_\")[:-1]))\n    sub[\"target_id\"] = sub[\"ID\"].apply(lambda x: \"_\".join(str(x).split(\"_\")[:-1]))\n\n    results = []\n    for target_id, group_native in sol.groupby(\"target_id\"):\n        group_pred = sub[sub[\"target_id\"] == target_id]\n        if len(group_pred) == 0:\n            continue\n        result = score(group_native, group_pred, \"ID\")\n        print(target_id, result)\n        results.append(result)\n\n    if results:\n        print(\"Mean score:\", float(sum(results)) / len(results), f\"(n={len(results)})\")\nelse:\n    print(\"Validation scoring skipped. Attach tm-score-permutechains and usalign to enable.\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-02-13T15:23:50.286476Z","iopub.execute_input":"2026-02-13T15:23:50.28687Z","iopub.status.idle":"2026-02-13T15:23:50.297947Z","shell.execute_reply.started":"2026-02-13T15:23:50.286834Z","shell.execute_reply":"2026-02-13T15:23:50.296503Z"}},"outputs":[],"execution_count":null}]}