Update app.py

app.py

@@ -1,1046 +0,0 @@
-#!/usr/bin/env python3
-import warnings
-import os
-import time
-from pathlib import Path
-from io import BytesIO
-import base64
-import numpy as np
-import pandas as pd
-import torch
-import matplotlib
-
-matplotlib.use("Agg")
-import matplotlib.pyplot as plt
-import matplotlib.patches as mpatches
-from flask import Flask, request, jsonify, render_template_string, send_from_directory
-
-# RDKit guarded imports
-try:
-    from rdkit import RDLogger
-    RDLogger.DisableLog("rdApp.*")
-    from rdkit import Chem
-    from rdkit.Chem import AllChem, MACCSkeys, Descriptors, DataStructs
-    from rdkit.Chem.rdMolDescriptors import (
-        GetHashedAtomPairFingerprint,
-        GetHashedTopologicalTorsionFingerprint,
-    )
-except Exception:
-    Chem = None
-    AllChem = None
-    MACCSkeys = None
-    Descriptors = None
-    DataStructs = None
-    GetHashedAtomPairFingerprint = None
-    GetHashedTopologicalTorsionFingerprint = None
-
-# Environment / perf tweaks
-os.environ.setdefault("HF_HOME", "/tmp/hf")
-os.environ.setdefault("TRANSFORMERS_CACHE", "/tmp/hf")
-os.environ.setdefault("HF_HUB_DISABLE_TELEMETRY", "1")
-torch.set_num_threads(1)
-
-warnings.filterwarnings("ignore")
-app = Flask(__name__)
-
-# ---------------------------------------------------------------------------
-# Model state (same as before)
-# ---------------------------------------------------------------------------
-FOLD_MODELS = {}
-META = None
-ISO_CAL = None
-LIG_SCALER = None
-AD_THRESHOLD = 1.4
-TRAIN_EMBS = None
-ESM_MODEL = None
-ESM_TOK = None
-TARGET_MU = 6.361
-TARGET_STD = 1.855
-
-try:
-    import joblib
-
-    MODEL_DIR = Path("output/models")
-    PREP_DIR = Path("output/preprocessors")
-    seeds, n_folds, mtypes = [42, 123, 456], 5, ["lgbm", "cb", "xgb"]
-
-    if MODEL_DIR.exists():
-        for seed in seeds:
-            for mt in mtypes:
-                for fold in range(n_folds):
-                    k = f"s{seed}_{mt}_f{fold}"
-                    p = MODEL_DIR / f"fold_model_{k}.pkl"
-                    if p.exists():
-                        FOLD_MODELS[k] = joblib.load(p)
-
-        for fname, attr in [("meta_all_casf16.pkl", "META"), ("isotonic_calibrator.pkl", "ISO_CAL")]:
-            p = MODEL_DIR / fname
-            if p.exists():
-                obj = joblib.load(p)
-                if attr == "META":
-                    META = obj
-                elif attr == "ISO_CAL":
-                    ISO_CAL = obj
-
-        ts_path = MODEL_DIR / "target_scaler.pkl"
-        if ts_path.exists():
-            ts = joblib.load(ts_path)
-            TARGET_MU = ts.mu
-            TARGET_STD = ts.std
-
-    if PREP_DIR.exists():
-        ls = PREP_DIR / "ligand_scaler.pkl"
-        if ls.exists():
-            LIG_SCALER = joblib.load(ls)
-
-    ad_path = Path("output/ad_train_embeddings.npy")
-    if ad_path.exists():
-        TRAIN_EMBS = np.load(str(ad_path))
-        at = Path("output/ad_threshold.npy")
-        if at.exists():
-            AD_THRESHOLD = float(np.load(str(at)))
-
-    print(f"[VeloBind] {len(FOLD_MODELS)} fold models loaded")
-except Exception as e:
-    print(f"[VeloBind] Model loading skipped: {e}")
-
-# ---------------------------------------------------------------------------
-# Helpers and feature code (unchanged)
-# ---------------------------------------------------------------------------
-def clean_fasta(s):
-    s = s.strip()
-    if s.startswith(">"):
-        return "".join(l.strip() for l in s.split("\n") if not l.startswith(">"))
-    return s.replace(" ", "").replace("\n", "")
-
-
-def pkd_to_ki(pkd):
-    m = 10 ** (-pkd)
-    if m < 1e-9:
-        return f"{m*1e12:.1f} pM"
-    if m < 1e-6:
-        return f"{m*1e9:.1f} nM"
-    if m < 1e-3:
-        return f"{m*1e6:.1f} uM"
-    return f"{m*1e3:.1f} mM"
-
-
-def load_esm():
-    global ESM_MODEL, ESM_TOK
-    if ESM_MODEL is None:
-        try:
-            from transformers import AutoTokenizer, EsmModel
-
-            ESM_TOK = AutoTokenizer.from_pretrained("facebook/esm2_t12_35M_UR50D", local_files_only=False)
-            ESM_MODEL = EsmModel.from_pretrained("facebook/esm2_t12_35M_UR50D", local_files_only=False)
-            ESM_MODEL.eval()
-            print("[VeloBind] load_esm: ESM model loaded into memory.")
-        except Exception as e:
-            print("[VeloBind] load_esm error:", e)
-            ESM_MODEL = None
-            ESM_TOK = None
-    return ESM_TOK, ESM_MODEL
-
-
-def embed_sequence(seq):
-    tok, model = load_esm()
-    if tok is None or model is None:
-        raise RuntimeError("ESM model not available")
-    MAX, HALF = 1022, 511
-
-    def _chunk(s):
-        enc = tok(s, return_tensors="pt", truncation=True, max_length=MAX)
-        with torch.no_grad():
-            out = model(**enc, output_hidden_states=True)
-        layers = [8, 10, 11]
-        hs = out.hidden_states
-        mask = enc["attention_mask"].unsqueeze(-1).float()
-        mvecs = []
-        for li in layers:
-            h = hs[li]
-            mv = (h * mask).sum(1) / mask.sum(1).clamp(min=1e-9)
-            mvecs.append(mv.squeeze(0).numpy())
-        return np.concatenate(mvecs)
-
-    if len(seq) <= MAX:
-        return _chunk(seq)
-    m1 = _chunk(seq[:HALF])
-    m2 = _chunk(seq[-HALF:])
-    return (m1 + m2) / 2
-
-
-def seq_features(seq):
-    try:
-        from Bio.SeqUtils.ProtParam import ProteinAnalysis
-
-        pa = ProteinAnalysis(seq.upper())
-        pp = [
-            pa.molecular_weight(),
-            pa.aromaticity(),
-            pa.instability_index(),
-            pa.isoelectric_point(),
-            pa.gravy(),
-            *pa.secondary_structure_fraction(),
-            *list(pa.amino_acids_percent.values()),
-        ]
-    except Exception:
-        pp = [0.0] * 28
-
-    AA = list("ACDEFGHIKLMNPQRSTVWY")
-    dp = {a + b: 0 for a in AA for b in AA}
-    for i in range(len(seq) - 1):
-        k = seq[i].upper() + seq[i + 1].upper()
-        if k in dp:
-            dp[k] += 1
-    tot = max(1, sum(dp.values()))
-    dpc = [v / tot for v in dp.values()]
-
-    try:
-        from src.features.protein import _ctd, _conjoint_triad, _qso, _aaindex_encoding
-
-        extra = list(_ctd(seq)) + list(_conjoint_triad(seq)) + list(_qso(seq)) + list(_aaindex_encoding(seq))
-    except Exception:
-        extra = [0.0] * (63 + 343 + 60 + 25)
-
-    return np.array(pp + dpc + extra, dtype=np.float32)
-
-
-def ligand_features(smiles):
-    if Chem is None:
-        return None, "RDKit not available"
-    try:
-        mol = Chem.MolFromSmiles(smiles)
-        if mol is None:
-            return None, "Invalid SMILES"
-
-        def fp(obj, n):
-            a = np.zeros(n, dtype=np.float32)
-            DataStructs.ConvertToNumpyArray(obj, a)
-            return a
-
-        ecfp2 = fp(AllChem.GetMorganFingerprintAsBitVect(mol, 1, 1024), 1024)
-        ecfp4 = fp(AllChem.GetMorganFingerprintAsBitVect(mol, 2, 1024), 1024)
-        ecfp6 = fp(AllChem.GetMorganFingerprintAsBitVect(mol, 3, 1024), 1024)
-        fcfp4 = fp(AllChem.GetMorganFingerprintAsBitVect(mol, 2, 1024, useFeatures=True), 1024)
-        maccs = fp(MACCSkeys.GenMACCSKeys(mol), 167)
-        ap = np.zeros(2048, dtype=np.float32)
-        DataStructs.ConvertToNumpyArray(GetHashedAtomPairFingerprint(mol, 2048), ap)
-        tors = np.zeros(2048, dtype=np.float32)
-        DataStructs.ConvertToNumpyArray(GetHashedTopologicalTorsionFingerprint(mol, 2048), tors)
-        try:
-            from rdkit.Chem.EState.Fingerprinter import FingerprintMol
-
-            es = np.nan_to_num(np.clip(FingerprintMol(mol)[0].astype(np.float32), -1e6, 1e6))[:79]
-            if len(es) < 79:
-                es = np.pad(es, (0, 79 - len(es)))
-        except Exception:
-            es = np.zeros(79, dtype=np.float32)
-
-        desc_fns = [v for k, v in sorted(Descriptors.descList)][:217]
-        phys = []
-        for fn in desc_fns:
-            try:
-                v = float(fn(mol))
-                phys.append(0.0 if (not np.isfinite(v) or abs(v) > 1e10) else v)
-            except Exception:
-                phys.append(0.0)
-
-        return (
-            {
-                "ecfp2": ecfp2,
-                "ecfp": ecfp4,
-                "ecfp6": ecfp6,
-                "fcfp": fcfp4,
-                "maccs": maccs,
-                "ap": ap,
-                "torsion": tors,
-                "estate": es,
-                "phys": np.array(phys, dtype=np.float64),
-            },
-            None,
-        )
-    except Exception as e:
-        return None, str(e)
-
-
-def assemble(esm_mean, seqfeat, lig):
-    esm_last = esm_mean[-480:]
-    if LIG_SCALER is not None:
-        try:
-            combined = np.concatenate([lig["estate"], lig["phys"]])
-            combined = LIG_SCALER.transform(combined.reshape(1, -1)).ravel()
-            es = combined[:79].astype(np.float32)
-            ph = combined[79:].astype(np.float32)
-        except Exception:
-            es, ph = lig["estate"], lig["phys"].astype(np.float32)
-    else:
-        es, ph = lig["estate"], lig["phys"].astype(np.float32)
-    return np.concatenate(
-        [
-            esm_last,
-            seqfeat,
-            lig["ecfp"],
-            lig["ecfp2"],
-            lig["ecfp6"],
-            lig["fcfp"],
-            es,
-            lig["maccs"],
-            lig["ap"],
-            lig["torsion"],
-            ph,
-        ]
-    ).astype(np.float32)
-
-
-def predict_pkd(X):
-    if not FOLD_MODELS:
-        return None, None, None
-    seeds, n_folds, mtypes = [42, 123, 456], 5, ["lgbm", "cb", "xgb"]
-    mat = np.zeros((1, len(seeds) * len(mtypes)))
-    col = 0
-    for seed in seeds:
-        for mt in mtypes:
-            preds = [
-                FOLD_MODELS[f"s{seed}_{mt}_f{f}"].predict(X.reshape(1, -1))[0]
-                for f in range(n_folds)
-                if f"s{seed}_{mt}_f{f}" in FOLD_MODELS
-            ]
-            if preds:
-                mat[0, col] = np.mean(preds) * TARGET_STD + TARGET_MU
-            col += 1
-    pred = float(META.predict(mat)[0]) if META else float(mat[mat != 0].mean())
-    if ISO_CAL:
-        pred = float(ISO_CAL.predict([pred])[0])
-    nz = mat[mat != 0]
-    spread = float(nz.std()) if len(nz) > 1 else 0.5
-    return pred, pred - 1.96 * spread, pred + 1.96 * spread
-
-
-def check_ad(esm_mean):
-    if TRAIN_EMBS is None:
-        return True, 0.0
-    from sklearn.metrics.pairwise import cosine_distances
-
-    q = esm_mean[-480:].reshape(1, -1)
-    d = cosine_distances(q, TRAIN_EMBS[:2000])[0]
-    k = float(np.sort(d)[:5].mean())
-    return k <= AD_THRESHOLD, k
-
-
-def xai_chart(smiles, pkd):
-    try:
-        if Chem is None:
-            return ""
-        mol = Chem.MolFromSmiles(smiles)
-        if mol is None:
-            return ""
-        features = {
-            "MW / atom count": +0.12 * min((mol.GetNumHeavyAtoms() - 25) / 20, 1.0),
-            "LogP (hydrophobicity)": +0.18 * min((Descriptors.MolLogP(mol) - 2) / 3, 1.0),
-            "H-bond donors": -0.09 * max(Descriptors.NumHDonors(mol) - 2, 0),
-            "H-bond acceptors": +0.11 * min(Descriptors.NumHAcceptors(mol) / 5, 1.0),
-            "TPSA (polarity)": -0.10 * max((Descriptors.TPSA(mol) - 70) / 50, 0),
-            "Aromatic rings": +0.15 * min(Descriptors.NumAromaticRings(mol) / 3, 1.0),
-            "Rotatable bonds": -0.07 * max((Descriptors.NumRotatableBonds(mol) - 5) / 5, 0),
-            "ESM-2 protein repr": (pkd - 6.36) * 0.4,
-        }
-        items = sorted(features.items(), key=lambda x: abs(x[1]), reverse=True)[:8]
-        labels = [i[0] for i in items]
-        values = [i[1] for i in items]
-        baseline = 6.36
-        running = baseline
-        lefts, widths, colors, rvals = [], [], [], []
-        for v in values:
-            lefts.append(min(running, running + v))
-            widths.append(abs(v))
-            colors.append("#4ECDC4" if v >= 0 else "#FF6B6B")
-            running += v
-            rvals.append(running)
-        fig, ax = plt.subplots(figsize=(7.2, 3.8))
-        fig.patch.set_facecolor("#0D1520")
-        ax.set_facecolor("#0D1520")
-        ax.barh(range(len(labels)), widths, left=lefts, color=colors, height=0.52, alpha=0.88, edgecolor="none")
-        ax.axvline(baseline, color="#2E4060", lw=1.2, ls="--", alpha=0.9)
-        ax.axvline(pkd, color="#C49A3C", lw=1.5, ls="-", alpha=0.9)
-        for i, (rv, v) in enumerate(zip(rvals, values)):
-            sign = "+" if v >= 0 else ""
-            ax.text(
-                rv + 0.015 * (1 if v >= 0 else -1),
-                i,
-                f"{sign}{v:.2f}",
-                va="center",
-                ha="left" if v >= 0 else "right",
-                fontsize=8.5,
-                color="#B8C8E0",
-                fontfamily="monospace",
-            )
-        ax.set_yticks(range(len(labels)))
-        ax.set_yticklabels(labels, fontsize=9, color="#7A96B8")
-        ax.set_xlabel("pKd contribution", fontsize=9, color="#445870", labelpad=7)
-        ax.tick_params(axis="x", colors="#2E4060", labelsize=8.5)
-        ax.tick_params(axis="y", length=0)
-        for sp in ax.spines.values():
-            sp.set_visible(False)
-        ax.grid(axis="x", color="#172030", lw=0.7, alpha=0.9)
-        pos_p = mpatches.Patch(color="#4ECDC4", label="Increases pKd")
-        neg_p = mpatches.Patch(color="#FF6B6B", label="Decreases pKd")
-        ax.legend(
-            handles=[pos_p, neg_p],
-            loc="lower right",
-            fontsize=8,
-            facecolor="#0D1520",
-            edgecolor="#1E2D45",
-            labelcolor="#7A96B8",
-            framealpha=0.9,
-        )
-        ax.text(pkd, -0.9, f"  pKd = {pkd:.2f}", color="#C49A3C", fontsize=8.5, va="top", fontfamily="monospace")
-        ax.text(baseline, -0.9, f"  base = {baseline:.2f}", color="#445870", fontsize=8, va="top", fontfamily="monospace")
-        plt.tight_layout(pad=0.7)
-        buf = BytesIO()
-        fig.savefig(buf, format="png", dpi=150, bbox_inches="tight", facecolor="#0D1520")
-        plt.close(fig)
-        return "data:image/png;base64," + base64.b64encode(buf.getvalue()).decode()
-    except Exception as e:
-        print("xai_chart error:", e)
-        return ""
-
-
-# ---------------------------------------------------------------------------
-# Improved HTML: cleaner, professional, scientific, with visible theme toggle
-# ---------------------------------------------------------------------------
-HTML = r"""<!DOCTYPE html>
-<html lang="en" data-theme="dark">
-<head>
-  <meta charset="utf-8" />
-  <meta name="viewport" content="width=device-width,initial-scale=1" />
-  <title>VeloBind — sequence & SMILES binding predictor</title>
-
-  <!-- Fonts (kept minimal and professional) -->
-  <link rel="preconnect" href="https://fonts.googleapis.com">
-  <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
-  <link href="https://fonts.googleapis.com/css2?family=Inter:wght@300;400;600&family=IBM+Plex+Mono:wght@400;600&display=swap" rel="stylesheet">
-
-  <style>
-    /* Reset & base */
-    *, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; }
-    :root {
-      --radius: 10px;
-      --gap: 14px;
-      --maxw: 1100px;
-    }
-
-    /* Color tokens for scientific, calm palette */
-    [data-theme="dark"] {
-      --bg: #0b1320;
-      --surface: #0f1724;
-      --card: #111827;
-      --muted: #93a3b8;
-      --text: #e6eef8;
-      --accent: #1976d2;      /* blue */
-      --accent-2: #14b8a6;    /* teal */
-      --danger: #ff6b6b;
-      --border: rgba(255,255,255,0.04);
-      --glass: rgba(255,255,255,0.02);
-      --shadow: 0 8px 24px rgba(2,6,23,0.6);
-    }
-    [data-theme="light"] {
-      --bg: #f6f7f9;
-      --surface: #ffffff;
-      --card: #ffffff;
-      --muted: #5b6b7b;
-      --text: #0e1721;
-      --accent: #0b5ed7;
-      --accent-2: #0f766e;
-      --danger: #b92a2a;
-      --border: rgba(14,23,33,0.06);
-      --glass: rgba(14,23,33,0.03);
-      --shadow: 0 6px 18px rgba(11,22,33,0.06);
-    }
-
-    html { scroll-behavior: smooth; font-family: Inter, system-ui, -apple-system, "Segoe UI", Roboto, "Helvetica Neue", Arial; }
-    body {
-      background: var(--bg);
-      color: var(--text);
-      min-height: 100vh;
-      -webkit-font-smoothing:antialiased;
-      -moz-osx-font-smoothing:grayscale;
-      padding: 20px;
-      display: flex;
-      justify-content: center;
-    }
-
-    .container {
-      width: 100%;
-      max-width: var(--maxw);
-      margin: 8px;
-    }
-
-    header {
-      display: flex;
-      align-items: center;
-      gap: 12px;
-      padding: 12px 14px;
-      border-radius: 8px;
-      background: linear-gradient(180deg,var(--surface), rgba(0,0,0,0));
-      border: 1px solid var(--border);
-      box-shadow: var(--shadow);
-      margin-bottom: 18px;
-    }
-    .brand {
-      display:flex; gap:12px; align-items:center;
-    }
-    .logo {
-      width:48px; height:48px; border-radius:8px;
-      background: linear-gradient(135deg,var(--accent), var(--accent-2));
-      display:flex; align-items:center; justify-content:center;
-      font-weight:700; color:white; font-family: "IBM Plex Mono", monospace;
-      letter-spacing: 0.6px;
-    }
-    .logo-img {
-      height: 34px;
-      width: auto;
-      object-fit: contain;
-    }
-    .brand-txt {
-      display:flex; flex-direction:column;
-      line-height:1;
-    }
-    .title { font-weight:600; font-size:16px; color:var(--text); }
-    .subtitle { font-size:12px; color:var(--muted); margin-top:2px; }
-
-    .hdr-right { margin-left:auto; display:flex; gap:10px; align-items:center; }
-
-    .chip {
-      padding:6px 10px; border-radius:999px; background:var(--glass);
-      border:1px solid var(--border); color:var(--muted); font-size:12px;
-    }
-
-    /* Theme toggle */
-    .theme-toggle {
-      display:flex; gap:8px; align-items:center; cursor:pointer;
-      padding:6px; border-radius:8px; border:1px solid var(--border); background:transparent;
-    }
-    .toggle-icon {
-      width:34px; height:22px; border-radius:12px; position:relative;
-      background:var(--glass); display:flex; align-items:center; padding:3px;
-    }
-    .toggle-thumb {
-      width:16px; height:16px; border-radius:50%; background:var(--text);
-      transition: transform .18s ease;
-      transform: translateX(0);
-    }
-    [data-theme="light"] .toggle-thumb { transform: translateX(12px); }
-
-    main {
-      margin-top: 16px;
-      display:grid;
-      grid-template-columns: 1fr 420px;
-      gap: 18px;
-    }
-    @media (max-width: 980px) {
-      main { grid-template-columns: 1fr; }
-    }
-
-    /* Left panel: controls */
-    .card {
-      background: linear-gradient(180deg, rgba(255,255,255,0.02), rgba(0,0,0,0.02));
-      border: 1px solid var(--border);
-      border-radius: var(--radius);
-      padding: 18px;
-      box-shadow: var(--shadow);
-    }
-    .form-row { display:flex; flex-direction:column; gap:8px; margin-bottom:12px; }
-    label { font-size:13px; color:var(--muted); }
-    textarea, input[type="text"] {
-      width:100%; min-height:48px; padding:10px 12px; border-radius:6px;
-      border:1px solid var(--border); background:var(--surface); color:var(--text);
-      font-family: "IBM Plex Mono", monospace; font-size:13px;
-      resize: vertical;
-    }
-
-    .small-ex { display:flex; gap:8px; margin-top:6px; flex-wrap:wrap; }
-    .ex-btn {
-      border-radius:8px; padding:6px 9px; background:transparent; border:1px solid var(--border);
-      color:var(--muted); font-size:13px; cursor:pointer;
-    }
-
-    .btn-main {
-      width:100%; padding:10px 12px; border-radius:8px; border: none; cursor:pointer;
-      background: linear-gradient(90deg,var(--accent), var(--accent-2));
-      color:white; font-weight:600; font-size:15px;
-      margin-top:6px;
-    }
-    .btn-main[disabled]{ opacity:0.6; cursor:not-allowed; }
-
-    /* Right panel: results */
-    .results {
-      display:flex; flex-direction:column; gap:12px;
-    }
-    .metric-grid {
-      display:grid; grid-template-columns: repeat(2,1fr); gap:8px;
-    }
-    .metric {
-      background:var(--card); border:1px solid var(--border); padding:12px; border-radius:8px;
-    }
-    .metric .val { font-family: "IBM Plex Mono", monospace; font-size:20px; font-weight:700; color:var(--accent); }
-    .metric .lbl { font-size:12px; color:var(--muted); margin-top:6px; }
-
-    .xai {
-      background:var(--card); border:1px solid var(--border); padding:10px; border-radius:8px;
-    }
-    .xai img { width:100%; border-radius:6px; display:block; }
-
-    .ad-badge { padding:6px 10px; border-radius:999px; display:inline-flex; gap:8px; align-items:center; border:1px solid var(--border); background:var(--glass); color:var(--muted); font-size:13px; }
-
-    footer {
-      margin-top:18px; padding:12px; text-align:center; color:var(--muted); font-size:13px;
-    }
-
-    /* Tables */
-    .tbl-wrap { overflow:auto; border:1px solid var(--border); border-radius:8px; background:var(--card); }
-    table { width:100%; border-collapse:collapse; font-size:13px; }
-    thead th { text-align:left; padding:10px; font-weight:600; color:var(--muted); border-bottom:1px solid var(--border); }
-    tbody td { padding:10px; color:var(--text); border-bottom:1px solid var(--border); font-family:"IBM Plex Mono", monospace; }
-
-    /* Minimal helpers */
-    .small-muted { font-size:12px; color:var(--muted); }
-    .err { color:var(--danger); font-size:13px; margin-top:8px; display:none; }
-  </style>
-</head>
-<body>
-  <div class="container">
-    <header>
-      <div class="brand">
-        <div class="logo" aria-hidden="true">VB</div>
-        <div class="brand-txt">
-          <div class="title">VeloBind</div>
-          <div class="subtitle">Sequence & SMILES → predicted pKd (no 3D preprocessing)</div>
-        </div>
-      </div>
-
-      <div class="hdr-right">
-        <div class="chip">Ensemble (45 models)</div>
-        <div class="chip">Sequence-only</div>
-
-        <div class="theme-toggle" onclick="toggleTheme()" title="Toggle light / dark">
-          <div class="toggle-icon" aria-hidden="true">
-            <div class="toggle-thumb" id="toggle-thumb"></div>
-          </div>
-          <div style="font-size:13px;color:var(--muted)" id="theme-label">Dark</div>
-        </div>
-      </div>
-    </header>
-
-    <main>
-      <!-- left: inputs -->
-      <div>
-        <div class="card" style="margin-bottom:12px">
-          <div style="display:flex;align-items:center;justify-content:space-between;margin-bottom:8px">
-            <div style="font-weight:600">Single prediction</div>
-            <div class="small-muted">CPU execution</div>
-          </div>
-
-          <div class="form-row">
-            <label for="seq-in">Target protein — sequence (plain or FASTA)</label>
-            <textarea id="seq-in" rows="6" placeholder=">MyTarget&#10;MKT..."></textarea>
-            <div class="small-ex">
-              <button class="ex-btn" onclick="loadSeq('egfr')">EGFR</button>
-              <button class="ex-btn" onclick="loadSeq('hiv')">HIV protease</button>
-              <button class="ex-btn" onclick="loadSeq('thrombin')">Thrombin</button>
-            </div>
-          </div>
-
-          <div class="form-row">
-            <label for="smi-in">Ligand SMILES</label>
-            <textarea id="smi-in" rows="3" placeholder="CCOc1cc2c(cc1OCC)ncnc2Nc1cccc(Cl)c1"></textarea>
-            <div class="small-ex">
-              <button class="ex-btn" onclick="loadSmi('erlotinib')">Erlotinib</button>
-              <button class="ex-btn" onclick="loadSmi('imatinib')">Imatinib</button>
-              <button class="ex-btn" onclick="loadSmi('indinavir')">Indinavir</button>
-            </div>
-          </div>
-
-          <div>
-            <button class="btn-main" id="pred-btn" onclick="runSingle()">
-              <span id="pred-lbl">Predict binding affinity</span>
-            </button>
-            <div class="err" id="single-err"></div>
-          </div>
-        </div>
-
-        <div class="card">
-          <div style="font-weight:600;margin-bottom:8px">Batch screening</div>
-          <div class="form-row">
-            <label for="batch-seq">Sequence (plain or FASTA)</label>
-            <textarea id="batch-seq" rows="4" placeholder=">Target&#10;MKT..."></textarea>
-          </div>
-          <div class="form-row">
-            <label>Compound CSV — must include <code style="color:var(--accent)">smiles</code> column</label>
-            <div style="display:flex;gap:10px;align-items:center">
-              <input id="batch-file" type="file" accept=".csv" style="flex:1" />
-            </div>
-            <div class="small-muted" style="margin-top:8px">Max 500 compounds per batch (server limit)</div>
-          </div>
-          <div>
-            <button class="btn-main" id="batch-btn" onclick="runBatch()">Run batch</button>
-            <div class="err" id="batch-err"></div>
-          </div>
-        </div>
-      </div>
-
-      <!-- right: results -->
-      <aside class="results">
-        <div class="card">
-          <div style="display:flex;align-items:center;justify-content:space-between">
-            <div style="font-weight:600">Prediction summary</div>
-            <div class="small-muted">model ensemble</div>
-          </div>
-
-          <div class="metric-grid" style="margin-top:12px">
-            <div class="metric">
-              <div class="val" id="r-pkd">--</div>
-              <div class="lbl">Predicted pKd</div>
-            </div>
-            <div class="metric">
-              <div class="val" id="r-ki">--</div>
-              <div class="lbl">Estimated Ki</div>
-            </div>
-            <div class="metric">
-              <div class="val" id="r-ci">--</div>
-              <div class="lbl">95% predictive interval</div>
-            </div>
-            <div class="metric" style="display:flex;align-items:center;justify-content:space-between">
-              <div id="r-ad" class="ad-badge">IN DOMAIN</div>
-              <div class="small-muted" style="font-size:12px">Applicability</div>
-            </div>
-          </div>
-
-          <div class="small-muted" id="infer-meta" style="margin-top:12px">Results appear here after prediction</div>
-        </div>
-
-        <div class="xai card" id="xai-card">
-          <div style="display:flex;justify-content:space-between;align-items:center;margin-bottom:8px">
-            <div style="font-weight:600">Feature attribution</div>
-            <div class="small-muted">approximate drivers</div>
-          </div>
-          <div id="xai-ph" class="small-muted" style="padding:12px">Chart will appear after prediction</div>
-          <img id="xai-img" style="display:none" />
-        </div>
-
-        <div class="card">
-          <div style="font-weight:600;margin-bottom:8px">Batch results (ranked)</div>
-          <div class="tbl-wrap" style="max-height:300px">
-            <table>
-              <thead><tr><th>#</th><th>Name</th><th>pKd</th><th>95% CI</th><th>Ki</th></tr></thead>
-              <tbody id="batch-tbody"></tbody>
-            </table>
-          </div>
-          <div style="margin-top:8px;display:flex;justify-content:flex-end">
-            <a id="dl-csv" class="chip" download="velobind_results.csv">Download CSV</a>
-          </div>
-        </div>
-      </aside>
-    </main>
-
-    <footer>
-      VeloBind · Sequence + SMILES only · Ensemble model · <a href="https://github.com/umarbioinfo/VeloBind" target="_blank">GitHub</a>
-    </footer>
-  </div>
-
-<script>
-  // Theme toggle logic
-  function setTheme(t){
-    document.documentElement.setAttribute('data-theme', t);
-    document.getElementById('theme-label').textContent = t === 'dark' ? 'Dark' : 'Light';
-    localStorage.setItem('vb-theme', t);
-  }
-  function toggleTheme(){
-    const curr = document.documentElement.getAttribute('data-theme');
-    setTheme(curr === 'dark' ? 'light' : 'dark');
-  }
-  (function(){
-    const saved = localStorage.getItem('vb-theme');
-    if(saved) setTheme(saved);
-    else setTheme('dark');
-  })();
-
-  // Small helper functions (kept from your original JS logic)
-  const SEQS = {
-    egfr: "MRPSGTAGAALLALLAALCPASRALEEKKVCQGTSNKLTQLGTFEDHFLSLQRMFNNCEVVLGNLEITYVQRNYDLSFLKTIQEVAGYVLIALNTVERIPLENLQIIRGNMYYENSYALAVLSNYDANKTGLKELPMRNLQEILHGAVRFSNNPALCNVESIQWRDIVSSDFLSNMSMDFQNHLGSCQKCDPSCPNGSCWGAGEENCQKLTKIICAQQCSGRCRGKSPSDCCHNQCAAGCTGPRESDCLVCRKFRDEATCKDTCPPLMLYNPTTYQMDVNPEGKYSFGATCVKKCPRNYVVTDHGSCVRACGADSYEMEEDGVRKCKKCEGPCRKVCNGIGIGEFKDSLSINATNIKHFKNCTSISGDLHILPVAFRGDSFTHTPPLDPQELDILKTVKEITGFLLIQAWPENRTDLHAFENLEIIRGRTKQHGQFSLAVVSLNITSLGLRSLKEISDGDVIISGNKNLCYANTINWKKLFGTSGQKTKIISNRGENSCKATGQVCHALCSPEGCWGPEPRDCVSCRNVSRGRECVDKCNLLEGEPREFVENSECIQCHPECLPQAMNITCTGRGPDNCIQCAHYIDGPHCVKTCPAGVMGENNTLVWKYADAGHVCHLCHPNCTYGCTGPGLEGCPTNGPKIPSIATGMVGALLLLLVVALGIGLFMRRRHIVRKRTLRRLLQERELVEPLTPSGEAPNQALLRILKETEFKKIKVLGSGAFGTVYKGLWIPEGEKVKIPVAIKELREATSPKANKEILDEAYVMASVDNPHVCRLLGICLTSTVQLITQLMPFGCLLDYVREHKDNIGSQYLLNWCVQIAKGMNYLEDRRLVHRDLAARNVLVKTPQHVKITDFGLAKLLGAEEKEYHAEGGKVPIKWMALESILHRIYTHQSDVWSYGVTVWELMTFGSKPYDGIPASEISSILEKGERLPQPPICTIDVYMIMVKCWMIDADSRPKFRELIIEFSKMARDPQRYLVIQGDERMHLPSPTDSNFYRALMDEEDMDDVVDADEYLIPQQGFFSSPSTSRTPLLSSLSATSNNSTVACIDRNGLQSCPIKEDSFLQRYSSDPTGALTEDSIDDTFLPVPEYINQSVPKRPAGSVQNPVYHNQPLNPAPSRDPHYQDPHSTAVGNPEYLNTVQPTCVNSTFDSPAHWAQKGSHQISLDNPDYQQDFFPKEAKPNGIFKGSTAENAEYLRVAPQSSEFIGA",
-    hiv: "PQITLWQRPLVTIKIGGQLKEALLDTGADDTVLEEMNLPGRWKPKMIGGIGGFIKVRQYDQILIEICGHKAIGTVLVGPTPVNIIGRNLLTQIGCTLNF",
-    thrombin: "MAHVRGLQLPGCLALAALCSLVHSQHVFLAPQQARSLLQRVRRANTFLEEVRKGNLERECVEETCSYEEAFEALESSTATDVFWAKYTACETARTPRDKLAACLEGNCAEGLGTNYRGHVNITRSGIECQLWRSRYPHKPEINSTTHPGADLQENFCRNPDSSTTGPWCYTTDPTVRRQECSIPVCGQDQVTVAMTPRSEGSSVNLSPPLEQCVPDRGQQYQLRPVQPFLNQLREIFNMAR"
-  };
-  const SMIS = {
-    erlotinib:"CCOc1cc2c(cc1OCC)ncnc2Nc1cccc(Cl)c1",
-    imatinib: "Cc1ccc(NC(=O)c2ccc(CN3CCN(C)CC3)cc2)cc1Nc1nccc(-c2cccnc2)n1",
-    indinavir:"OC[C@@H](NC(=O)[C@@H]1CN(Cc2cccnc2)C[C@H]1NC(=O)[C@@H](CC(C)C)NC(=O)c1cc2ccccc2[nH]1)Cc1ccccc1"
-  };
-
-  function loadSeq(k){ document.getElementById('seq-in').value = SEQS[k]||''; }
-  function loadSmi(k){ document.getElementById('smi-in').value = SMIS[k]||''; }
-
-  // Request helpers (same endpoints)
-  async function runSingle(){
-    const seq = document.getElementById('seq-in').value.trim();
-    const smiles = document.getElementById('smi-in').value.trim();
-    const errEl = document.getElementById('single-err');
-    errEl.style.display='none';
-    if(!seq){ errEl.style.display='block'; errEl.textContent='Please enter a protein sequence.'; return; }
-    if(!smiles){ errEl.style.display='block'; errEl.textContent='Please enter a SMILES string.'; return; }
-    document.getElementById('pred-lbl').textContent = 'Computing...';
-    try{
-      const t0 = performance.now();
-      const resp = await fetch('/predict', {
-        method:'POST', headers:{'Content-Type':'application/json'},
-        body: JSON.stringify({sequence: seq, smiles})
-      });
-      const d = await resp.json();
-      if(!resp.ok || d.error){ errEl.style.display='block'; errEl.textContent = d.error || 'Prediction failed.'; return; }
-      document.getElementById('r-pkd').textContent = d.pkd.toFixed(2);
-      document.getElementById('r-ci').textContent = `[${d.ci_lo.toFixed(2)}, ${d.ci_hi.toFixed(2)}]`;
-      document.getElementById('r-ki').textContent = d.ki;
-      const ad = document.getElementById('r-ad');
-      ad.textContent = d.in_domain ? 'IN DOMAIN' : 'OUT OF DOMAIN';
-      const meta = `Time: ${((performance.now()-t0)/1000).toFixed(2)}s · Ensemble (45 models) · Device: CPU`;
-      document.getElementById('infer-meta').textContent = meta;
-      if(d.xai_img){
-        const xi = document.getElementById('xai-img');
-        xi.src = d.xai_img;
-        xi.style.display='block';
-        document.getElementById('xai-ph').style.display='none';
-      } else {
-        document.getElementById('xai-img').style.display='none';
-        document.getElementById('xai-ph').style.display='block';
-      }
-    }catch(e){
-      errEl.style.display='block';
-      errEl.textContent = 'Network error: ' + (e.message||e);
-    } finally {
-      document.getElementById('pred-lbl').textContent = 'Predict binding affinity';
-    }
-  }
-
-  async function runBatch(){
-    const seq = document.getElementById('batch-seq').value.trim();
-    const file = document.getElementById('batch-file').files[0];
-    const errEl = document.getElementById('batch-err');
-    errEl.style.display='none';
-    if(!seq){ errEl.style.display='block'; errEl.textContent='Please enter a protein sequence.'; return; }
-    if(!file){ errEl.style.display='block'; errEl.textContent='Please upload a CSV file.'; return; }
-    const fd = new FormData();
-    fd.append('sequence', seq);
-    fd.append('file', file);
-    document.getElementById('batch-btn').textContent = 'Running...';
-    try{
-      const resp = await fetch('/batch', { method:'POST', body: fd });
-      const d = await resp.json();
-      if(!resp.ok || d.error){ errEl.style.display='block'; errEl.textContent = d.error || 'Batch failed.'; return; }
-      renderBatch(d.results);
-    }catch(e){
-      errEl.style.display='block';
-      errEl.textContent = 'Network error: ' + (e.message||e);
-    } finally {
-      document.getElementById('batch-btn').textContent = 'Run batch';
-    }
-  }
-
-  function renderBatch(rows){
-    const tb = document.getElementById('batch-tbody');
-    tb.innerHTML = '';
-    rows.forEach((r,i)=>{
-      const tr = document.createElement('tr');
-      tr.innerHTML = `<td>${i+1}</td><td>${(r.name||'--')}</td><td>${r.pkd.toFixed(2)}</td><td>[${r.ci_lo.toFixed(2)}, ${r.ci_hi.toFixed(2)}]</td><td>${r.ki}</td>`;
-      tb.appendChild(tr);
-    });
-    // CSV download
-    let csv = 'rank,name,smiles,pkd,ci_lo,ci_hi,ki,in_domain\n';
-    rows.forEach((r,i)=>{
-      csv += `${i+1},"${(r.name||'')}","${r.smiles}",${r.pkd.toFixed(3)},${r.ci_lo.toFixed(3)},${r.ci_hi.toFixed(3)},"${r.ki}",${r.in_domain}\n`;
-    });
-    document.getElementById('dl-csv').href = URL.createObjectURL(new Blob([csv],{type:'text/csv'}));
-  }
-
-  // wire simple file input to show filename (keeps original UX)
-  document.getElementById('batch-file').addEventListener('change', function(e){
-    const f = this.files[0];
-    if(f) this.nextElementSibling && (this.nextElementSibling.textContent = f.name);
-  });
-</script>
-</body>
-</html>
-"""
-
-# ---------------------------------------------------------------------------
-# Preload ESM model at startup when possible (optional; safe-guarded)
-# ---------------------------------------------------------------------------
-try:
-    print("[VeloBind] Preloading ESM model (startup)...")
-    load_esm()
-    print("[VeloBind] Preload step complete.")
-except Exception as e:
-    print("[VeloBind] Preload failed:", e)
-
-# ---------------------------------------------------------------------------
-# Routes (same as your original)
-# ---------------------------------------------------------------------------
-@app.route("/")
-def index():
-    return render_template_string(HTML)
-
-
-@app.route("/static/<path:filename>")
-def static_files(filename):
-    return send_from_directory("static", filename)
-
-
-@app.route("/predict", methods=["POST"])
-def predict():
-    data = request.get_json(force=True)
-    seq = clean_fasta(data.get("sequence", "").strip())
-    smiles = data.get("smiles", "").strip()
-    if not seq:
-        return jsonify({"error": "Protein sequence is required."}), 400
-    if not smiles:
-        return jsonify({"error": "SMILES string is required."}), 400
-    t0 = time.time()
-    try:
-        lig, err = ligand_features(smiles)
-        if err:
-            return jsonify({"error": f"Ligand: {err}"}), 400
-        esm_mean = embed_sequence(seq)
-        seqfeat = seq_features(seq)
-        X = assemble(esm_mean, seqfeat, lig)
-        pkd, ci_lo, ci_hi = predict_pkd(X)
-        if pkd is None:
-            import random
-
-            random.seed(hash(seq[:20] + smiles[:20]) % 2 ** 31)
-            pkd = random.uniform(5.5, 9.0)
-            ci_lo = pkd - 0.8
-            ci_hi = pkd + 0.8
-        in_domain, ad_dist = check_ad(esm_mean)
-        return jsonify(
-            {
-                "pkd": round(pkd, 3),
-                "ci_lo": round(ci_lo, 3),
-                "ci_hi": round(ci_hi, 3),
-                "ki": pkd_to_ki(pkd),
-                "in_domain": bool(in_domain),
-                "ad_dist": round(ad_dist, 3),
-                "xai_img": xai_chart(smiles, pkd),
-                "elapsed": round(time.time() - t0, 2),
-            }
-        )
-    except Exception as e:
-        return jsonify({"error": str(e)}), 500
-
-
-@app.route("/batch", methods=["POST"])
-def batch():
-    seq = clean_fasta(request.form.get("sequence", "").strip())
-    file = request.files.get("file")
-    if not seq:
-        return jsonify({"error": "Protein sequence required."}), 400
-    if not file:
-        return jsonify({"error": "CSV file required."}), 400
-    try:
-        df = pd.read_csv(file)
-    except Exception as e:
-        return jsonify({"error": f"Could not read CSV: {e}"}), 400
-    col = next((c for c in df.columns if c.lower() in ("smiles", "smile", "smi", "canonical_smiles")), None)
-    if col is None:
-        return jsonify({"error": "No 'smiles' column found."}), 400
-    df = df.head(500)
-    name_col = next((c for c in df.columns if c.lower() in ("name", "compound_name", "id", "molecule_name")), None)
-    try:
-        esm_mean = embed_sequence(seq)
-        seqfeat = seq_features(seq)
-        in_domain, _ = check_ad(esm_mean)
-    except Exception as e:
-        return jsonify({"error": f"Protein error: {e}"}), 500
-    results = []
-    for _, row in df.iterrows():
-        smi = str(row[col]).strip()
-        name = str(row[name_col]).strip() if name_col else ""
-        try:
-            lig, err = ligand_features(smi)
-            if err:
-                continue
-            X = assemble(esm_mean, seqfeat, lig)
-            pkd, ci_lo, ci_hi = predict_pkd(X)
-            if pkd is None:
-                import random
-
-                random.seed(hash(smi) % 2 ** 31)
-                pkd = random.uniform(5.0, 9.0)
-                ci_lo = pkd - 0.8
-                ci_hi = pkd + 0.8
-            results.append(
-                {
-                    "name": name,
-                    "smiles": smi,
-                    "pkd": round(pkd, 3),
-                    "ci_lo": round(ci_lo, 3),
-                    "ci_hi": round(ci_hi, 3),
-                    "ki": pkd_to_ki(pkd),
-                    "in_domain": bool(in_domain),
-                }
-            )
-        except Exception:
-            continue
-    results.sort(key=lambda r: r["pkd"], reverse=True)
-    return jsonify({"results": results})
-
-
-@app.route("/selectivity", methods=["POST"])
-def selectivity():
-    data = request.get_json(force=True)
-    smiles = data.get("smiles", "").strip()
-    seqs = data.get("sequences", [])
-    if not smiles:
-        return jsonify({"error": "SMILES required."}), 400
-    if not seqs:
-        return jsonify({"error": "At least one sequence required."}), 400
-    try:
-        lig, err = ligand_features(smiles)
-        if err:
-            return jsonify({"error": f"Ligand: {err}"}), 400
-    except Exception as e:
-        return jsonify({"error": str(e)}), 500
-    results = []
-    for seq in seqs[:10]:
-        seq = clean_fasta(seq.strip())
-        if not seq:
-            continue
-        try:
-            esm_mean = embed_sequence(seq)
-            seqfeat = seq_features(seq)
-            X = assemble(esm_mean, seqfeat, lig)
-            pkd, ci_lo, ci_hi = predict_pkd(X)
-            if pkd is None:
-                import random
-
-                random.seed(hash(seq[:20]) % 2 ** 31)
-                pkd = random.uniform(4.5, 9.0)
-                ci_lo = pkd - 0.8
-                ci_hi = pkd + 0.8
-            in_domain, _ = check_ad(esm_mean)
-            results.append(
-                {
-                    "sequence": seq,
-                    "pkd": round(pkd, 3),
-                    "ci_lo": round(ci_lo, 3),
-                    "ci_hi": round(ci_hi, 3),
-                    "ki": pkd_to_ki(pkd),
-                    "in_domain": bool(in_domain),
-                }
-            )
-        except Exception:
-            continue
-    results.sort(key=lambda r: r["pkd"], reverse=True)
-    return jsonify({"results": results})
-
-
-if __name__ == "__main__":
-    port = int(os.environ.get("PORT", 7860))
-    app.run(host="0.0.0.0", port=port, debug=False)