alterlab-molfeat

Files (6)

• SKILL.md
• evals.json · evals/
• api_reference.md · references/
• available_featurizers.md · references/
• examples.md · references/
• workflows_and_patterns.md · references/

SKILL.md (first lines)

# Molfeat - Molecular Featurization Hub

## Overview

Molfeat is a comprehensive Python library for molecular featurization that unifies 100+ pre-trained embeddings and hand-crafted featurizers. Convert chemical structures (SMILES strings or RDKit molecules) into numerical representations for machine learning tasks including QSAR modeling, virtual screening, similarity searching, and deep learning applications. Features fast parallel processing, scikit-learn compatible transformers, and built-in caching.

## When to Use This Skill

This skill should be used when working with:
- **Molecular machine learning**: Building QSAR/QSPR models, property prediction
- **Virtual screening**: Ranking compound libraries for biological activity
- **Similarity searching**: Finding structurally similar molecules
- **Chemical space analysis**: Clustering, visualization, dimensionality reduction
- **Deep learning**: Training neural networks on molecular data
- **Featurization pipelines**: Converting SMILES to ML-ready representations
- **Cheminformatics**: Any task requiring molecular feature extraction

### Does NOT Trigger

| Scenario | Use Instead |
|----------|-------------|
| Training/evaluating models end-to-end on MoleculeNet benchmarks with built-in loaders and GNNs | `alterlab-deepchem` |
| Sourcing a labeled benchmark dataset with scaffold/cold splits | `alterlab-pytdc` |
| Low-level fingerprint or descriptor primitives, custom sanitization, SMARTS | `alterlab-rdkit` |
| Standardizing and cleaning molecule tables before featurization | `alterlab-datamol` |

## Installation

```bash
uv pip install molfeat            # molfeat 1.0.0 (current as of 2026-09); Python >= 3.11, pulls torch + datamol

# Optional extras (molfeat 1.x)
uv pip install "molfeat[transformer]"   # Hugging Face models: ChemBERTa, ChemGPT, MolT5, ...
uv pip install "molfeat[mordred]"       # Mordred descriptors (mordredcommunity)
uv pip install "molfeat[fcd]"           # FCD / ChemNet embeddings
uv pip install "molfeat[pyg]"           # PyTorch Geometric (used by Mol-JEPA)
uv pip install "molfeat[all]"
```

**molfeat 1.0 breaking changes.** The DGL-based pretrained GNNs (`gin_supervised_*`, `jtvae_zinc_no_kl`), Graphormer, and the protein featurizers were removed, along with the `dgl` and `graphormer` extras (there has never been a `map4` extra); loading those model-store entries now fails. New foundation-model featurizers are `CheMeleonTransformer` (2,048-d, weights fetched from Zenodo and checksum-verified) and `MolJEPATransformer` (CC BY-NC 4.0; requires `trust_remote_code=True` and `accept_noncommercial_license=True`). If you must reproduce legacy GIN/Graphormer embeddings, pin `molfeat<1` (0.11.x requires Python ≤ 3.10) in a separate environment. MAP4 needs the `map4` package from https://github.com/reymond-group/map4 (not on PyPI).

## Core Concepts

Molfeat organizes featurization into three hierarchical classes:

### 1. Calculators (`molfeat.calc`)

Callable objects that convert individual molecules into feature vectors. Accept RDKit `Chem.Mol` objects or SMILES strings.

**Use calculators for:**
- Single molecule featurization
- Custom processing loops
- Direct feature computation

**Example:**
```python
from molfeat.calc import FPCalculator

calc = FPCalculator("ecfp", radius=3, fpSize=2048)
features = calc("CCO")  # Returns numpy array (2048,)
```

### 2. Transformers (`molfeat.trans`)

Scikit-learn compatible transformers that wrap calculators for batch processing with parallelization.

**Use transformers for:**
- Batch featurization of molecular datasets
- Integration with scikit-learn pipelines
- Parallel processing (automatic CPU utilization)

**Example:**
```python
import numpy as np
from molfeat.trans import MoleculeTransformer
from molfeat.calc import FPCalculator

transformer = MoleculeTransformer(FPCalculator("ecfp"), n_jobs=-1, dtype=np.float32)
features = transformer(smiles_list)  # (n_mols, 2048) array; without dtype you get a list of arrays
```

### 3. Pretrained Transformers (`molfeat.trans.pretrained`)

Specialized transformers for deep learning models with batched inference and caching.

**Use pretrained transformers for:**
- State-of-the-art molecular embeddings
- Transfer learning from large chemical datasets
- Deep learning feature extraction

**Example** (`PretrainedMolTransformer` is the abstract base class — instantiate a concrete subclass):
```python
import numpy as np
from molfeat.trans.pretrained import PretrainedHFTransformer

transformer = PretrainedHFTransformer(kind="ChemBERTa-77M-MLM", notation="smiles", dtype=np.float32)
embeddings = transformer(smiles_list)  # (n_mols, 384) mean-pooled embeddings
```

## Quick Start Workflow

### Basic Featurization

```python
import numpy as np
from molfeat.calc import FPCalculator
from molfeat.trans import MoleculeTransformer

# Load molecular data
smiles = ["CCO", "CC(=O)O", "c1ccccc1", "CC(C)O"]

# Create calculator and transformer (dtype makes the output a single array)
calc = FPCalculator("ecfp", radius=3)
transformer = MoleculeTransformer(calc, n_jobs=-1, dtype=np.float32)

# Featurize molecules
features = transformer(smiles)
print(f"Shape: {features.shape}")  # (4, 2048)
```

From AlterLab-IEU/AlterLab-Academic-Skills, Apache-2.0.

More from AlterLab-IEU/AlterLab-Academic-Skills