revise

NMRGym_both_test.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a999773f00c4cf5688900bc8edd4dc490a83a6f9209d3041d2f5eb682bb6f86f
+size 172546408

NMRGym_both_train.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:627e133032a30e8b03e35f53d560b9a8cb41700371b288802034e3069bd6d59a
+size 1348822022

NMRGym_test_balanced.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:996cb692e86c86de5ac3735274500298b3e80d3e1f57868478270fe472301a6b
+size 207945355

NMRGym_train_balanced.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cd4fdb6f8a54b1b49c27b94f2c32e449e6499fe247ada04651b5ca85716ed54
+size 1041901432

NMRGym_val_balanced.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:deb0c5d1e27527986ef994ac53f75528ddf2dce2874bbbe75012729fbf0976f0
+size 138866740

README.md

@@ -0,0 +1,237 @@
+---
+license: mit
+task_categories:
+- text-to-text
+- tabular-regression
+- tabular-classification
+pretty_name: NMRGym
+language:
+- en
+tags:
+- chemistry
+- nmr
+- spectroscopy
+- molecular-property-prediction
+- drug-discovery
+- cheminformatics
+size_categories:
+- 100K<n<1M
+---
+
+# ‼️ Important ‼️
+"If you need access to the dataset, please email [zhengf723@connect.hkust-gz.edu.cn](mailto:zhengf723@connect.hkust-gz.edu.cn)"
+
+# NMRGym
+Benchmark on NMR Spectrum.
+
+## Data Sources (Before Cleaning)
+
+| Source              |     Records | Unique SMILES | Total Spectrums |      ¹H NMR |     ¹³C NMR |
+| :------------------ | ----------: | ------------: | --------------: | ----------: | ----------: |
+| **CH-NP**           |      12,165 |        12,165 |          24,326 |      12,161 |      12,165 |
+| **HMDB**            |       1,791 |           896 |           3,278 |       1,566 |       1,712 |
+| **NMRBank**         |     148,914 |       142,964 |         297,043 |     148,437 |     148,606 |
+| **NMRShiftDB 2024** |      41,019 |        39,631 |          50,416 |      18,570 |      31,846 |
+| **NP-MRD**          |     489,569 |       243,598 |         950,242 |     462,233 |     488,009 |
+| **PubChem-NMR**     |       1,647 |         1,535 |           2,605 |       1,556 |       1,049 |
+| **SDBS**            |      12,926 |        12,626 |          22,711 |      11,522 |      11,189 |
+| **Total**           | **708,031** |   **430,690** |   **1,350,621** | **656,045** | **694,576** |
+
+---
+
+## Balanced Dataset Statistics (After Cleaning & Processing)
+
+After data cleaning, filtering, quality control, and balanced scaffold splitting, the final NMRGym dataset consists of three splits:
+
+| Split    |     Records | Unique SMILES |      ¹H NMR |     ¹³C NMR |
+| :------- | ----------: | ------------: | ----------: | ----------: |
+| **Train** |     402,676 |       250,822 |     402,676 |     402,676 |
+| **Val**   |      53,672 |        50,716 |      53,672 |      53,672 |
+| **Test**  |      80,069 |        74,568 |      80,069 |      80,069 |
+| **Total** | **536,417** |   **376,106** | **536,417** | **536,417** |
+
+### Dataset Visualizations
+
+<!-- ![Dataset Overview](dataset_overview.png)
+*Figure 1: Overview of dataset statistics including total records, unique SMILES, data duplication, NMR spectra types, and top elements.*
+
+![Functional Groups Distribution](functional_groups.png)
+*Figure 2: Distribution of 22 functional groups across train, validation, and test sets.*
+
+![Element Distribution](element_distribution.png)
+*Figure 3: Distribution of common elements (C, H, O, N, F, Cl, Br, S, P, I) across datasets.* -->
+
+
+
+---
+
+### Quick Checklist
+
+* [✅] Data cleaning: Heavy atom filtering and illegal smiles exclusion.
+* [✅] Data Summary and Visualization.
+* [✅] Data Split.
+* [] 3D coord. generation. Rdkit.
+* [✅] Toxic Property label generation. 
+* [✅] Function Group label generation.
+
+---
+## NMRGym Split Dataset Format (`nmrgym_spec_filtered_both_{train,test}.pkl`)
+
+This dataset contains post-QC, scaffold-split molecular entries with paired NMR spectra, structure fingerprints, functional group annotations, and toxicity labels.
+
+Each `.pkl` file is a Python list of dictionaries. Each dictionary corresponds to a single unique molecule.
+
+### Example record structure
+
+```python
+{
+    "smiles": "CC(=O)Oc1ccccc1C(=O)O",
+    "h_shift": [7.25, 7.32, 2.14, 1.27],
+    "c_shift": [128.5, 130.1, 172.9, 20.7],
+    "fingerprint": np.ndarray(shape=(2048,), dtype=np.uint8),
+    "fg_onehot": np.ndarray(shape=(22,), dtype=np.uint8),
+    "toxicity": np.ndarray(shape=(7,), dtype=np.int8),
+}
+```
+
+---
+
+### Field definitions
+
+**`smiles`** (`str`)
+Canonical SMILES string for this molecule.
+
+**`h_shift`** (`list[float]`)
+Experimental or curated ¹H NMR peak positions in ppm.
+
+**`c_shift`** (`list[float]`)
+Experimental or curated ¹³C NMR peak positions in ppm.
+
+**`fingerprint`** (`np.ndarray(2048,)`, `uint8`)
+RDKit Morgan fingerprint (radius = 2, 2048 bits). Encodes local circular substructures.
+
+**`fg_onehot`** (`np.ndarray(22,)`, `uint8`)
+Binary functional-group vector. `1` means the SMARTS pattern is present in the molecule.
+Index mapping (0→21):
+
+1. Alcohol
+2. Carboxylic Acid
+3. Ester
+4. Ether
+5. Aldehyde
+6. Ketone
+7. Alkene
+8. Alkyne
+9. Benzene
+10. Primary Amine
+11. Secondary Amine
+12. Tertiary Amine
+13. Amide
+14. Cyano
+15. Fluorine
+16. Chlorine
+17. Bromine
+18. Iodine
+19. Sulfonamide
+20. Sulfone
+21. Sulfide
+22. Phosphoric Acid
+
+**`toxicity`** (`np.ndarray(7,)`, `int8`)
+Seven binary toxicity endpoints (1 = toxic / positive, 0 = negative):
+0. AMES (mutagenicity)
+
+1. DILI (Drug-Induced Liver Injury)
+2. Carcinogens_Lagunin (carcinogenicity)
+3. hERG (cardiotoxic channel block)
+4. ClinTox (clinical toxicity)
+5. NR-ER (endocrine / estrogen receptor)
+6. SR-ARE (oxidative stress response)
+
+---
+
+### Train / Test meaning
+
+* `nmrgym_spec_filtered_both_train.pkl`: scaffold-based training set.
+* `nmrgym_spec_filtered_both_test.pkl`: scaffold-disjoint test set.
+
+Scaffold split is computed using Bemis–Murcko scaffolds. This means test scaffolds do not appear in train, simulating generalization to new chemotypes instead of just new stereoisomers.
+
+---
+
+### Minimal usage example
+
+```python
+import pickle
+from rdkit import Chem
+from rdkit.Chem import AllChem
+
+# load the dataset
+test_path = "/gemini/user/private/NMRGym/utils/split_output/nmrgym_spec_filtered_both_test.pkl"
+with open(test_path, "rb") as f:
+    dataset = pickle.load(f)
+
+print("num molecules:", len(dataset))
+print("keys:", dataset[0].keys())
+print("example toxicity vector:", dataset[0]["toxicity"])
+
+# build 3D conformer for the first molecule
+smi = dataset[0]["smiles"]
+mol = Chem.MolFromSmiles(smi)
+mol = Chem.AddHs(mol)  # add hydrogens for 3D
+AllChem.EmbedMolecule(mol, AllChem.ETKDG())
+AllChem.UFFOptimizeMolecule(mol)
+
+# extract 3D coordinates (in Å)
+conf = mol.GetConformer()
+coords = []
+for atom_idx in range(mol.GetNumAtoms()):
+    pos = conf.GetAtomPosition(atom_idx)
+    coords.append([pos.x, pos.y, pos.z])
+
+print("3D coords for first molecule (Angstrom):")
+for i, (x,y,z) in enumerate(coords):
+    sym = mol.GetAtomWithIdx(i).GetSymbol()
+    print(f"{i:2d} {sym:2s}  {x:8.3f} {y:8.3f} {z:8.3f}")
+```
+
+### Notes on 3D coordinates
+
+* We generate a single low-energy conformer using RDKit ETKDG embedding + UFF optimization.
+* Coordinates are in Ångström.
+* These coordinates are not guaranteed to match the experimental NMR conformer in solvent; they are intended for featurization (message passing models, geometry-aware models, etc.), not quantum-accurate geometry.
+
+---
+
+## Downstream Benchmark Tasks
+
+This benchmark suite evaluates AI models on multiple spectroscopy-related prediction tasks. Each task reflects a distinct molecular reasoning aspect — from direct spectrum regression to property and toxicity prediction.
+
+### 1. Spectral Prediction
+
+| **Method / Paper Title** | **Model Type / Approach** | **Metric(s)** | **Model Size** | **Notes** |
+| ------------------------ | ------------------------- | ------------- | -------------- | --------- |
+| x | x |x | x | x |
+
+### 2. Structure Prediction (Inverse NMR)
+
+| **Method / Paper Title** | **Model Type / Approach** | **Metric(s)** | **Model Size** | **Notes** |
+| ------------------------ | ------------------------- | ------------- | -------------- | --------- |
+| x | x |x | x | x |
+### 3. Molecular Fingerprint Prediction (Spec2FP)
+
+| **Method / Paper Title** | **Model Type / Approach** | **Metric(s)** | **Model Size** | **Notes** |
+| ------------------------ | ------------------------- | ------------- | -------------- | --------- |
+| x | x |x | x | x |
+### 4. Functional Group Classification (Spec2Func)
+
+| **Method / Paper Title** | **Model Type / Approach** | **Metric(s)** | **Model Size** | **Notes** |
+| ------------------------ | ------------------------- | ------------- | -------------- | --------- |
+| x | x |x | x | x |
+### 5. Molecular Toxicity Prediction (Spec2Tox / ADMET)
+
+| **Method / Paper Title** | **Model Type / Approach** | **Metric(s)** | **Model Size** | **Notes** |
+| ------------------------ | ------------------------- | ------------- | -------------- | --------- |
+| x | x |x | x | x |
+
+---

analyze_data.py

@@ -0,0 +1,368 @@
+#!/usr/bin/env python3
+"""
+Analyze NMRGym balanced datasets and generate visualizations
+"""
+import pickle
+import numpy as np
+import matplotlib.pyplot as plt
+import seaborn as sns
+from collections import Counter
+from rdkit import Chem
+from rdkit.Chem.Scaffolds import MurckoScaffold
+import json
+
+# Set style
+sns.set_style("whitegrid")
+plt.rcParams['font.size'] = 10
+plt.rcParams['figure.dpi'] = 300
+
+# Functional group names (index 0-21)
+FG_NAMES = [
+    "Alcohol", "Carboxylic Acid", "Ester", "Ether", "Aldehyde", "Ketone",
+    "Alkene", "Alkyne", "Benzene", "Primary Amine", "Secondary Amine",
+    "Tertiary Amine", "Amide", "Cyano", "Fluorine", "Chlorine",
+    "Bromine", "Iodine", "Sulfonamide", "Sulfone", "Sulfide", "Phosphoric Acid"
+]
+
+def load_dataset(pkl_path):
+    """Load a pickle file"""
+    with open(pkl_path, "rb") as f:
+        return pickle.load(f)
+
+def get_scaffold(smiles):
+    """Get Bemis-Murcko scaffold"""
+    try:
+        mol = Chem.MolFromSmiles(smiles)
+        if mol is None:
+            return None
+        scaffold = MurckoScaffold.GetScaffoldForMol(mol)
+        return Chem.MolToSmiles(scaffold)
+    except:
+        return None
+
+def get_element_counts(smiles):
+    """Get element counts from SMILES"""
+    try:
+        mol = Chem.MolFromSmiles(smiles)
+        if mol is None:
+            return {}
+        element_counts = {}
+        for atom in mol.GetAtoms():
+            symbol = atom.GetSymbol()
+            element_counts[symbol] = element_counts.get(symbol, 0) + 1
+        return element_counts
+    except:
+        return {}
+
+def analyze_dataset(dataset, name):
+    """Analyze a single dataset"""
+    print(f"\n{'='*60}")
+    print(f"Analyzing {name}")
+    print(f"{'='*60}")
+
+    stats = {
+        'name': name,
+        'total_records': len(dataset),
+        'unique_smiles': len(set(d['smiles'] for d in dataset)),
+        'scaffolds': [],
+        'functional_groups': np.zeros(22),
+        'elements': Counter(),
+        'h_spectra': 0,
+        'c_spectra': 0,
+    }
+
+    all_smiles = set()
+    scaffold_counter = Counter()
+
+    for record in dataset:
+        smiles = record['smiles']
+        all_smiles.add(smiles)
+
+        # Count spectra
+        if 'h_shift' in record and record['h_shift'] is not None and len(record['h_shift']) > 0:
+            stats['h_spectra'] += 1
+        if 'c_shift' in record and record['c_shift'] is not None and len(record['c_shift']) > 0:
+            stats['c_spectra'] += 1
+
+        # Get scaffold
+        scaffold = get_scaffold(smiles)
+        if scaffold:
+            scaffold_counter[scaffold] += 1
+
+        # Functional groups
+        if 'fg_onehot' in record:
+            stats['functional_groups'] += record['fg_onehot']
+
+        # Element counts
+        elem_counts = get_element_counts(smiles)
+        for elem, count in elem_counts.items():
+            stats['elements'][elem] += count
+
+    stats['unique_scaffolds'] = len(scaffold_counter)
+    stats['top_scaffolds'] = scaffold_counter.most_common(10)
+
+    print(f"Total records: {stats['total_records']:,}")
+    print(f"Unique SMILES: {stats['unique_smiles']:,}")
+    print(f"Unique scaffolds: {stats['unique_scaffolds']:,}")
+    print(f"¹H NMR spectra: {stats['h_spectra']:,}")
+    print(f"¹³C NMR spectra: {stats['c_spectra']:,}")
+    print(f"\nTop 5 elements:")
+    for elem, count in stats['elements'].most_common(5):
+        print(f"  {elem}: {count:,}")
+
+    return stats
+
+def plot_functional_groups(train_stats, val_stats, test_stats, output_path):
+    """Plot functional group distribution"""
+    fig, ax = plt.subplots(figsize=(14, 6))
+
+    x = np.arange(len(FG_NAMES))
+    width = 0.25
+
+    train_fg = train_stats['functional_groups']
+    val_fg = val_stats['functional_groups']
+    test_fg = test_stats['functional_groups']
+
+    ax.bar(x - width, train_fg, width, label='Train', alpha=0.8)
+    ax.bar(x, val_fg, width, label='Val', alpha=0.8)
+    ax.bar(x + width, test_fg, width, label='Test', alpha=0.8)
+
+    ax.set_xlabel('Functional Group')
+    ax.set_ylabel('Count')
+    ax.set_title('Functional Group Distribution Across Datasets')
+    ax.set_xticks(x)
+    ax.set_xticklabels(FG_NAMES, rotation=45, ha='right')
+    ax.legend()
+    ax.grid(axis='y', alpha=0.3)
+
+    plt.tight_layout()
+    plt.savefig(output_path, dpi=300, bbox_inches='tight')
+    plt.close()
+    print(f"Saved: {output_path}")
+
+def plot_element_distribution(train_stats, val_stats, test_stats, output_path):
+    """Plot element distribution for common elements"""
+    # Focus on common organic elements
+    common_elements = ['C', 'H', 'O', 'N', 'F', 'Cl', 'Br', 'S', 'P', 'I']
+
+    fig, ax = plt.subplots(figsize=(12, 6))
+
+    # Get counts for each element
+    train_counts = [train_stats['elements'].get(e, 0) for e in common_elements]
+    val_counts = [val_stats['elements'].get(e, 0) for e in common_elements]
+    test_counts = [test_stats['elements'].get(e, 0) for e in common_elements]
+
+    x = np.arange(len(common_elements))
+    width = 0.25
+
+    ax.bar(x - width, train_counts, width, label='Train', alpha=0.8)
+    ax.bar(x, val_counts, width, label='Val', alpha=0.8)
+    ax.bar(x + width, test_counts, width, label='Test', alpha=0.8)
+
+    ax.set_xlabel('Element')
+    ax.set_ylabel('Total Count')
+    ax.set_title('Element Distribution Across Datasets')
+    ax.set_xticks(x)
+    ax.set_xticklabels(common_elements)
+    ax.legend()
+    ax.grid(axis='y', alpha=0.3)
+    ax.set_yscale('log')
+
+    plt.tight_layout()
+    plt.savefig(output_path, dpi=300, bbox_inches='tight')
+    plt.close()
+    print(f"Saved: {output_path}")
+
+def plot_dataset_overview(train_stats, val_stats, test_stats, output_path):
+    """Plot overview of dataset statistics"""
+    fig, axes = plt.subplots(2, 2, figsize=(14, 10))
+
+    # 1. Total records and unique SMILES
+    ax = axes[0, 0]
+    datasets = ['Train', 'Val', 'Test']
+    total_records = [train_stats['total_records'], val_stats['total_records'], test_stats['total_records']]
+    unique_smiles = [train_stats['unique_smiles'], val_stats['unique_smiles'], test_stats['unique_smiles']]
+
+    x = np.arange(len(datasets))
+    width = 0.35
+
+    ax.bar(x - width/2, total_records, width, label='Total Records', alpha=0.8)
+    ax.bar(x + width/2, unique_smiles, width, label='Unique SMILES', alpha=0.8)
+    ax.set_ylabel('Count')
+    ax.set_title('Dataset Size Comparison')
+    ax.set_xticks(x)
+    ax.set_xticklabels(datasets)
+    ax.legend()
+    ax.grid(axis='y', alpha=0.3)
+
+    # Add value labels on bars
+    for i, (tr, us) in enumerate(zip(total_records, unique_smiles)):
+        ax.text(i - width/2, tr, f'{tr:,}', ha='center', va='bottom', fontsize=8)
+        ax.text(i + width/2, us, f'{us:,}', ha='center', va='bottom', fontsize=8)
+
+    # 2. Unique scaffolds
+    ax = axes[0, 1]
+    unique_scaffolds = [train_stats['unique_scaffolds'], val_stats['unique_scaffolds'], test_stats['unique_scaffolds']]
+
+    ax.bar(datasets, unique_scaffolds, alpha=0.8, color='coral')
+    ax.set_ylabel('Count')
+    ax.set_title('Unique Scaffolds')
+    ax.grid(axis='y', alpha=0.3)
+
+    for i, count in enumerate(unique_scaffolds):
+        ax.text(i, count, f'{count:,}', ha='center', va='bottom', fontsize=9)
+
+    # 3. NMR spectra types
+    ax = axes[1, 0]
+    h_spectra = [train_stats['h_spectra'], val_stats['h_spectra'], test_stats['h_spectra']]
+    c_spectra = [train_stats['c_spectra'], val_stats['c_spectra'], test_stats['c_spectra']]
+
+    x = np.arange(len(datasets))
+    width = 0.35
+
+    ax.bar(x - width/2, h_spectra, width, label='¹H NMR', alpha=0.8)
+    ax.bar(x + width/2, c_spectra, width, label='¹³C NMR', alpha=0.8)
+    ax.set_ylabel('Count')
+    ax.set_title('NMR Spectra Types')
+    ax.set_xticks(x)
+    ax.set_xticklabels(datasets)
+    ax.legend()
+    ax.grid(axis='y', alpha=0.3)
+
+    # 4. Top 5 elements (train set)
+    ax = axes[1, 1]
+    top_elements = train_stats['elements'].most_common(5)
+    elements = [e[0] for e in top_elements]
+    counts = [e[1] for e in top_elements]
+
+    ax.bar(elements, counts, alpha=0.8, color='skyblue')
+    ax.set_ylabel('Total Count')
+    ax.set_title('Top 5 Elements (Train Set)')
+    ax.grid(axis='y', alpha=0.3)
+    ax.set_yscale('log')
+
+    plt.tight_layout()
+    plt.savefig(output_path, dpi=300, bbox_inches='tight')
+    plt.close()
+    print(f"Saved: {output_path}")
+
+def plot_scaffold_diversity(train_stats, val_stats, test_stats, output_path):
+    """Plot scaffold diversity comparison"""
+    fig, ax = plt.subplots(figsize=(10, 6))
+
+    datasets = ['Train', 'Val', 'Test']
+    total_records = [train_stats['total_records'], val_stats['total_records'], test_stats['total_records']]
+    unique_scaffolds = [train_stats['unique_scaffolds'], val_stats['unique_scaffolds'], test_stats['unique_scaffolds']]
+
+    # Calculate scaffold diversity ratio
+    diversity_ratio = [s/r for s, r in zip(unique_scaffolds, total_records)]
+
+    x = np.arange(len(datasets))
+    width = 0.6
+
+    bars = ax.bar(x, diversity_ratio, width, alpha=0.8, color=['#1f77b4', '#ff7f0e', '#2ca02c'])
+    ax.set_ylabel('Scaffold Diversity Ratio')
+    ax.set_title('Scaffold Diversity (Unique Scaffolds / Total Records)')
+    ax.set_xticks(x)
+    ax.set_xticklabels(datasets)
+    ax.grid(axis='y', alpha=0.3)
+    ax.set_ylim(0, 1)
+
+    # Add value labels
+    for i, (bar, ratio) in enumerate(zip(bars, diversity_ratio)):
+        height = bar.get_height()
+        ax.text(bar.get_x() + bar.get_width()/2., height,
+                f'{ratio:.3f}\n({unique_scaffolds[i]:,}/{total_records[i]:,})',
+                ha='center', va='bottom', fontsize=9)
+
+    plt.tight_layout()
+    plt.savefig(output_path, dpi=300, bbox_inches='tight')
+    plt.close()
+    print(f"Saved: {output_path}")
+
+def main():
+    # File paths
+    train_path = "/gemini/code/NMRGym/NMRGym_train_balanced.pkl"
+    val_path = "/gemini/code/NMRGym/NMRGym_val_balanced.pkl"
+    test_path = "/gemini/code/NMRGym/NMRGym_test_balanced.pkl"
+
+    # Load datasets
+    print("Loading datasets...")
+    train_data = load_dataset(train_path)
+    val_data = load_dataset(val_path)
+    test_data = load_dataset(test_path)
+
+    # Analyze each dataset
+    train_stats = analyze_dataset(train_data, "Train (Balanced)")
+    val_stats = analyze_dataset(val_data, "Val (Balanced)")
+    test_stats = analyze_dataset(test_data, "Test (Balanced)")
+
+    # Generate visualizations
+    print("\n" + "="*60)
+    print("Generating visualizations...")
+    print("="*60)
+
+    plot_dataset_overview(train_stats, val_stats, test_stats,
+                         "/gemini/code/NMRGym/dataset_overview.png")
+
+    plot_functional_groups(train_stats, val_stats, test_stats,
+                          "/gemini/code/NMRGym/functional_groups.png")
+
+    plot_element_distribution(train_stats, val_stats, test_stats,
+                             "/gemini/code/NMRGym/element_distribution.png")
+
+    plot_scaffold_diversity(train_stats, val_stats, test_stats,
+                           "/gemini/code/NMRGym/scaffold_diversity.png")
+
+    # Save statistics as JSON
+    summary = {
+        'train': {
+            'total_records': train_stats['total_records'],
+            'unique_smiles': train_stats['unique_smiles'],
+            'unique_scaffolds': train_stats['unique_scaffolds'],
+            'h_spectra': train_stats['h_spectra'],
+            'c_spectra': train_stats['c_spectra'],
+        },
+        'val': {
+            'total_records': val_stats['total_records'],
+            'unique_smiles': val_stats['unique_smiles'],
+            'unique_scaffolds': val_stats['unique_scaffolds'],
+            'h_spectra': val_stats['h_spectra'],
+            'c_spectra': val_stats['c_spectra'],
+        },
+        'test': {
+            'total_records': test_stats['total_records'],
+            'unique_smiles': test_stats['unique_smiles'],
+            'unique_scaffolds': test_stats['unique_scaffolds'],
+            'h_spectra': test_stats['h_spectra'],
+            'c_spectra': test_stats['c_spectra'],
+        }
+    }
+
+    with open('/gemini/code/NMRGym/dataset_stats.json', 'w') as f:
+        json.dump(summary, f, indent=2)
+    print("\nSaved: /gemini/code/NMRGym/dataset_stats.json")
+
+    # Print final summary table
+    print("\n" + "="*60)
+    print("FINAL SUMMARY")
+    print("="*60)
+    print(f"{'Dataset':<15} {'Records':>10} {'Unique SMILES':>15} {'Scaffolds':>12} {'¹H NMR':>10} {'¹³C NMR':>10}")
+    print("-" * 85)
+    for name, stats in [('Train', train_stats), ('Val', val_stats), ('Test', test_stats)]:
+        print(f"{name:<15} {stats['total_records']:>10,} {stats['unique_smiles']:>15,} {stats['unique_scaffolds']:>12,} "
+              f"{stats['h_spectra']:>10,} {stats['c_spectra']:>10,}")
+
+    total_records = train_stats['total_records'] + val_stats['total_records'] + test_stats['total_records']
+    total_unique = train_stats['unique_smiles'] + val_stats['unique_smiles'] + test_stats['unique_smiles']
+    total_scaffolds = train_stats['unique_scaffolds'] + val_stats['unique_scaffolds'] + test_stats['unique_scaffolds']
+    total_h = train_stats['h_spectra'] + val_stats['h_spectra'] + test_stats['h_spectra']
+    total_c = train_stats['c_spectra'] + val_stats['c_spectra'] + test_stats['c_spectra']
+
+    print("-" * 85)
+    print(f"{'Total':<15} {total_records:>10,} {total_unique:>15,} {total_scaffolds:>12,} "
+          f"{total_h:>10,} {total_c:>10,}")
+    print("="*60 + "\n")
+
+if __name__ == "__main__":
+    main()

dataset_stats.json

@@ -0,0 +1,20 @@
+{
+  "train": {
+    "total_records": 402676,
+    "unique_smiles": 250822,
+    "h_spectra": 402676,
+    "c_spectra": 402676
+  },
+  "val": {
+    "total_records": 53672,
+    "unique_smiles": 50716,
+    "h_spectra": 53672,
+    "c_spectra": 53672
+  },
+  "test": {
+    "total_records": 80069,
+    "unique_smiles": 74568,
+    "h_spectra": 80069,
+    "c_spectra": 80069
+  }
+}

quick_analyze.py

@@ -0,0 +1,313 @@
+#!/usr/bin/env python3
+"""
+Quick analysis of NMRGym balanced datasets without expensive scaffold computation
+"""
+import pickle
+import numpy as np
+import matplotlib.pyplot as plt
+import seaborn as sns
+from collections import Counter
+from rdkit import Chem
+import json
+
+# Set style
+sns.set_style("whitegrid")
+plt.rcParams['font.size'] = 10
+plt.rcParams['figure.dpi'] = 300
+
+# Functional group names (index 0-21)
+FG_NAMES = [
+    "Alcohol", "Carboxylic Acid", "Ester", "Ether", "Aldehyde", "Ketone",
+    "Alkene", "Alkyne", "Benzene", "Primary Amine", "Secondary Amine",
+    "Tertiary Amine", "Amide", "Cyano", "Fluorine", "Chlorine",
+    "Bromine", "Iodine", "Sulfonamide", "Sulfone", "Sulfide", "Phosphoric Acid"
+]
+
+def load_dataset(pkl_path):
+    """Load a pickle file"""
+    print(f"Loading {pkl_path}...")
+    with open(pkl_path, "rb") as f:
+        return pickle.load(f)
+
+def get_element_counts(smiles):
+    """Get element counts from SMILES"""
+    try:
+        mol = Chem.MolFromSmiles(smiles)
+        if mol is None:
+            return {}
+        element_counts = {}
+        for atom in mol.GetAtoms():
+            symbol = atom.GetSymbol()
+            element_counts[symbol] = element_counts.get(symbol, 0) + 1
+        return element_counts
+    except:
+        return {}
+
+def analyze_dataset(dataset, name):
+    """Analyze a single dataset (quick version without scaffold)"""
+    print(f"\n{'='*60}")
+    print(f"Analyzing {name}")
+    print(f"{'='*60}")
+
+    stats = {
+        'name': name,
+        'total_records': len(dataset),
+        'unique_smiles': 0,
+        'functional_groups': np.zeros(22),
+        'elements': Counter(),
+        'h_spectra': 0,
+        'c_spectra': 0,
+    }
+
+    unique_smiles = set()
+
+    for i, record in enumerate(dataset):
+        if (i + 1) % 1000 == 0:
+            print(f"  Processed {i+1}/{len(dataset)} records...")
+
+        smiles = record['smiles']
+        unique_smiles.add(smiles)
+
+        # Count spectra
+        if 'h_shift' in record and record['h_shift'] is not None and len(record['h_shift']) > 0:
+            stats['h_spectra'] += 1
+        if 'c_shift' in record and record['c_shift'] is not None and len(record['c_shift']) > 0:
+            stats['c_spectra'] += 1
+
+        # Functional groups
+        if 'fg_onehot' in record:
+            stats['functional_groups'] += record['fg_onehot']
+
+        # Element counts
+        elem_counts = get_element_counts(smiles)
+        for elem, count in elem_counts.items():
+            stats['elements'][elem] += count
+
+    stats['unique_smiles'] = len(unique_smiles)
+
+    print(f"Total records: {stats['total_records']:,}")
+    print(f"Unique SMILES: {stats['unique_smiles']:,}")
+    print(f"¹H NMR spectra: {stats['h_spectra']:,}")
+    print(f"¹³C NMR spectra: {stats['c_spectra']:,}")
+    print(f"\nTop 5 elements:")
+    for elem, count in stats['elements'].most_common(5):
+        print(f"  {elem}: {count:,}")
+
+    return stats
+
+def plot_functional_groups(train_stats, val_stats, test_stats, output_path):
+    """Plot functional group distribution"""
+    fig, ax = plt.subplots(figsize=(14, 6))
+
+    x = np.arange(len(FG_NAMES))
+    width = 0.25
+
+    train_fg = train_stats['functional_groups']
+    val_fg = val_stats['functional_groups']
+    test_fg = test_stats['functional_groups']
+
+    ax.bar(x - width, train_fg, width, label='Train', alpha=0.8)
+    ax.bar(x, val_fg, width, label='Val', alpha=0.8)
+    ax.bar(x + width, test_fg, width, label='Test', alpha=0.8)
+
+    ax.set_xlabel('Functional Group')
+    ax.set_ylabel('Count')
+    ax.set_title('Functional Group Distribution Across Datasets')
+    ax.set_xticks(x)
+    ax.set_xticklabels(FG_NAMES, rotation=45, ha='right')
+    ax.legend()
+    ax.grid(axis='y', alpha=0.3)
+
+    plt.tight_layout()
+    plt.savefig(output_path, dpi=300, bbox_inches='tight')
+    plt.close()
+    print(f"\nSaved: {output_path}")
+
+def plot_element_distribution(train_stats, val_stats, test_stats, output_path):
+    """Plot element distribution for common elements"""
+    # Focus on common organic elements
+    common_elements = ['C', 'H', 'O', 'N', 'F', 'Cl', 'Br', 'S', 'P', 'I']
+
+    fig, ax = plt.subplots(figsize=(12, 6))
+
+    # Get counts for each element
+    train_counts = [train_stats['elements'].get(e, 0) for e in common_elements]
+    val_counts = [val_stats['elements'].get(e, 0) for e in common_elements]
+    test_counts = [test_stats['elements'].get(e, 0) for e in common_elements]
+
+    x = np.arange(len(common_elements))
+    width = 0.25
+
+    ax.bar(x - width, train_counts, width, label='Train', alpha=0.8)
+    ax.bar(x, val_counts, width, label='Val', alpha=0.8)
+    ax.bar(x + width, test_counts, width, label='Test', alpha=0.8)
+
+    ax.set_xlabel('Element')
+    ax.set_ylabel('Total Count')
+    ax.set_title('Element Distribution Across Datasets')
+    ax.set_xticks(x)
+    ax.set_xticklabels(common_elements)
+    ax.legend()
+    ax.grid(axis='y', alpha=0.3)
+    ax.set_yscale('log')
+
+    plt.tight_layout()
+    plt.savefig(output_path, dpi=300, bbox_inches='tight')
+    plt.close()
+    print(f"Saved: {output_path}")
+
+def plot_dataset_overview(train_stats, val_stats, test_stats, output_path):
+    """Plot overview of dataset statistics"""
+    fig, axes = plt.subplots(2, 2, figsize=(14, 10))
+
+    # 1. Total records and unique SMILES
+    ax = axes[0, 0]
+    datasets = ['Train', 'Val', 'Test']
+    total_records = [train_stats['total_records'], val_stats['total_records'], test_stats['total_records']]
+    unique_smiles = [train_stats['unique_smiles'], val_stats['unique_smiles'], test_stats['unique_smiles']]
+
+    x = np.arange(len(datasets))
+    width = 0.35
+
+    ax.bar(x - width/2, total_records, width, label='Total Records', alpha=0.8)
+    ax.bar(x + width/2, unique_smiles, width, label='Unique SMILES', alpha=0.8)
+    ax.set_ylabel('Count')
+    ax.set_title('Dataset Size Comparison')
+    ax.set_xticks(x)
+    ax.set_xticklabels(datasets)
+    ax.legend()
+    ax.grid(axis='y', alpha=0.3)
+
+    # Add value labels on bars
+    for i, (tr, us) in enumerate(zip(total_records, unique_smiles)):
+        ax.text(i - width/2, tr, f'{tr:,}', ha='center', va='bottom', fontsize=8)
+        ax.text(i + width/2, us, f'{us:,}', ha='center', va='bottom', fontsize=8)
+
+    # 2. Data duplicates (Total vs Unique SMILES)
+    ax = axes[0, 1]
+    duplication_ratio = [1 - (u/t) if t > 0 else 0 for u, t in zip(unique_smiles, total_records)]
+
+    bars = ax.bar(datasets, duplication_ratio, alpha=0.8, color='coral')
+    ax.set_ylabel('Duplication Ratio')
+    ax.set_title('Data Duplication (1 - Unique/Total)')
+    ax.grid(axis='y', alpha=0.3)
+    ax.set_ylim(0, max(duplication_ratio) * 1.2 if max(duplication_ratio) > 0 else 1)
+
+    for i, (bar, ratio) in enumerate(zip(bars, duplication_ratio)):
+        height = bar.get_height()
+        ax.text(bar.get_x() + bar.get_width()/2., height,
+                f'{ratio:.2%}',
+                ha='center', va='bottom', fontsize=9)
+
+    # 3. NMR spectra types
+    ax = axes[1, 0]
+    h_spectra = [train_stats['h_spectra'], val_stats['h_spectra'], test_stats['h_spectra']]
+    c_spectra = [train_stats['c_spectra'], val_stats['c_spectra'], test_stats['c_spectra']]
+
+    x = np.arange(len(datasets))
+    width = 0.35
+
+    ax.bar(x - width/2, h_spectra, width, label='¹H NMR', alpha=0.8)
+    ax.bar(x + width/2, c_spectra, width, label='¹³C NMR', alpha=0.8)
+    ax.set_ylabel('Count')
+    ax.set_title('NMR Spectra Types')
+    ax.set_xticks(x)
+    ax.set_xticklabels(datasets)
+    ax.legend()
+    ax.grid(axis='y', alpha=0.3)
+
+    # 4. Top 5 elements (train set)
+    ax = axes[1, 1]
+    top_elements = train_stats['elements'].most_common(5)
+    elements = [e[0] for e in top_elements]
+    counts = [e[1] for e in top_elements]
+
+    ax.bar(elements, counts, alpha=0.8, color='skyblue')
+    ax.set_ylabel('Total Count')
+    ax.set_title('Top 5 Elements (Train Set)')
+    ax.grid(axis='y', alpha=0.3)
+    ax.set_yscale('log')
+
+    plt.tight_layout()
+    plt.savefig(output_path, dpi=300, bbox_inches='tight')
+    plt.close()
+    print(f"Saved: {output_path}")
+
+def main():
+    # File paths
+    train_path = "/gemini/code/NMRGym/NMRGym_train_balanced.pkl"
+    val_path = "/gemini/code/NMRGym/NMRGym_val_balanced.pkl"
+    test_path = "/gemini/code/NMRGym/NMRGym_test_balanced.pkl"
+
+    # Load datasets
+    train_data = load_dataset(train_path)
+    val_data = load_dataset(val_path)
+    test_data = load_dataset(test_path)
+
+    # Analyze each dataset
+    train_stats = analyze_dataset(train_data, "Train (Balanced)")
+    val_stats = analyze_dataset(val_data, "Val (Balanced)")
+    test_stats = analyze_dataset(test_data, "Test (Balanced)")
+
+    # Generate visualizations
+    print("\n" + "="*60)
+    print("Generating visualizations...")
+    print("="*60)
+
+    plot_dataset_overview(train_stats, val_stats, test_stats,
+                         "/gemini/code/NMRGym/dataset_overview.png")
+
+    plot_functional_groups(train_stats, val_stats, test_stats,
+                          "/gemini/code/NMRGym/functional_groups.png")
+
+    plot_element_distribution(train_stats, val_stats, test_stats,
+                             "/gemini/code/NMRGym/element_distribution.png")
+
+    # Save statistics as JSON
+    summary = {
+        'train': {
+            'total_records': train_stats['total_records'],
+            'unique_smiles': train_stats['unique_smiles'],
+            'h_spectra': train_stats['h_spectra'],
+            'c_spectra': train_stats['c_spectra'],
+        },
+        'val': {
+            'total_records': val_stats['total_records'],
+            'unique_smiles': val_stats['unique_smiles'],
+            'h_spectra': val_stats['h_spectra'],
+            'c_spectra': val_stats['c_spectra'],
+        },
+        'test': {
+            'total_records': test_stats['total_records'],
+            'unique_smiles': test_stats['unique_smiles'],
+            'h_spectra': test_stats['h_spectra'],
+            'c_spectra': test_stats['c_spectra'],
+        }
+    }
+
+    with open('/gemini/code/NMRGym/dataset_stats.json', 'w') as f:
+        json.dump(summary, f, indent=2)
+    print("\nSaved: /gemini/code/NMRGym/dataset_stats.json")
+
+    # Print final summary table
+    print("\n" + "="*60)
+    print("FINAL SUMMARY")
+    print("="*60)
+    print(f"{'Dataset':<15} {'Records':>10} {'Unique SMILES':>15} {'¹H NMR':>10} {'¹³C NMR':>10}")
+    print("-" * 70)
+    for name, stats in [('Train', train_stats), ('Val', val_stats), ('Test', test_stats)]:
+        print(f"{name:<15} {stats['total_records']:>10,} {stats['unique_smiles']:>15,} "
+              f"{stats['h_spectra']:>10,} {stats['c_spectra']:>10,}")
+
+    total_records = train_stats['total_records'] + val_stats['total_records'] + test_stats['total_records']
+    total_unique = train_stats['unique_smiles'] + val_stats['unique_smiles'] + test_stats['unique_smiles']
+    total_h = train_stats['h_spectra'] + val_stats['h_spectra'] + test_stats['h_spectra']
+    total_c = train_stats['c_spectra'] + val_stats['c_spectra'] + test_stats['c_spectra']
+
+    print("-" * 70)
+    print(f"{'Total':<15} {total_records:>10,} {total_unique:>15,} "
+          f"{total_h:>10,} {total_c:>10,}")
+    print("="*60 + "\n")
+
+if __name__ == "__main__":
+    main()