Skip to content

THEMAP: Task Hardness Estimation for Molecular Activity Prediction

THEMAP is a comprehensive Python library designed to aid drug discovery by providing powerful methods for estimating task hardness and computing transferability maps for bioactivity prediction tasks. It enables researchers and chemists to efficiently determine the similarity between molecular datasets and make informed decisions about transfer learning strategies.

Key Features

🧪 Multi-Modal Distance Computation

  • Molecular datasets: OTDD, Euclidean, and Cosine distances
  • Protein sequences: ESM2-based embeddings and similarity metrics
  • Metadata integration: Assay descriptions and experimental conditions
  • Combined analysis: Multi-modal fusion strategies

🎯 Task Hardness Estimation

  • Transfer learning guidance: Identify similar tasks for knowledge transfer
  • Difficulty quantification: Estimate prediction task complexity
  • Resource optimization: Prioritize computational resources effectively

âš¡ Production-Ready Framework

  • Scalable architecture: Handle large-scale dataset comparisons
  • Caching system: Efficient feature storage and reuse
  • Error handling: Robust validation and error recovery
  • GPU acceleration: CUDA support for intensive computations

🔬 Unified Task System

  • Integrated data management: Molecules, proteins, and metadata in one framework
  • Flexible organization: Train/validation/test fold management
  • Feature extraction: Unified API for multi-modal featurization

Installation

# Clone repository
git clone https://github.com/HFooladi/THEMAP.git
cd THEMAP

# Create conda environment
conda env create -f environment.yml
conda activate themap

# Install THEMAP
pip install --no-deps -e .

# Install with all optional dependencies
pip install -e ".[all]"

# Install development dependencies
pip install -e ".[dev,test]"

# Run tests
python run_tests.py

Quick Examples

Molecular Dataset Distance

Compute distances between molecular datasets to understand chemical space similarity:

from themap.data.tasks import Tasks
from themap.distance import MoleculeDatasetDistance

# Load tasks from directory structure
tasks = Tasks.from_directory(
    directory="datasets/",
    task_list_file="datasets/sample_tasks_list.json",
    load_molecules=True,
    load_proteins=True
)

# Compute molecular distances using OTDD
mol_distance = MoleculeDatasetDistance(
    tasks=tasks,
    molecule_method="otdd"
)

distances = mol_distance.get_distance()
print(distances)
# {'CHEMBL2219358': {'CHEMBL1023359': 7.074}}

Unified Multi-Modal Analysis

Combine molecular, protein, and metadata information for comprehensive task comparison:

from themap.distance import TaskDistance

# Compute combined distances from multiple modalities
task_distance = TaskDistance(
    tasks=tasks,
    molecule_method="cosine",
    protein_method="euclidean"
)

# Get all distance types
all_distances = task_distance.compute_all_distances(
    combination_strategy="weighted_average",
    molecule_weight=0.7,
    protein_weight=0.3
)

print(f"Molecule distances: {len(all_distances['molecule'])} tasks")
print(f"Protein distances: {len(all_distances['protein'])} tasks")
print(f"Combined distances: {len(all_distances['combined'])} tasks")

Protein Similarity Analysis

Analyze protein similarity using advanced sequence embeddings:

from themap.distance import ProteinDatasetDistance

# Compute protein distances using ESM2 embeddings
prot_distance = ProteinDatasetDistance(
    tasks=tasks,
    protein_method="euclidean"
)

distances = prot_distance.get_distance()
# Organized as {target_task: {source_task: distance, ...}, ...}

Use Cases

Drug Discovery Workflows

  • Target identification: Find similar protein targets for drug repurposing
  • Chemical space analysis: Understand molecular diversity across datasets
  • Assay development: Identify related bioactivity assays

Transfer Learning Applications

  • Source task selection: Choose optimal training data for new targets
  • Model adaptation: Quantify domain shift between datasets
  • Performance prediction: Estimate model performance on new tasks

Computational Biology

  • Protein function prediction: Leverage sequence similarity for annotation
  • Chemical-protein interaction: Model molecular-target relationships
  • Multi-omics integration: Combine molecular and protein data

Performance

THEMAP is optimized for both accuracy and computational efficiency:

Method Speed Memory Accuracy Best For
OTDD Slower High Highest Small-medium datasets
Euclidean Fast Low Good Large datasets
Cosine Fast Low Good High-dimensional features

Scalability Features

  • Parallel processing: Multi-core distance computations
  • Memory management: Efficient handling of large datasets
  • Caching system: Reuse expensive feature computations
  • Batch processing: Handle thousands of dataset comparisons

Citation

If you use THEMAP in your research, please cite:

@article{fooladi2024quantifying,
  title={Quantifying the hardness of bioactivity prediction tasks for transfer learning},
  author={Fooladi, Hosein and Hirte, Steffen and Kirchmair, Johannes},
  journal={Journal of Chemical Information and Modeling},
  volume={64},
  number={10},
  pages={4031-4046},
  year={2024},
  publisher={ACS Publications}
}

Community and Support

Get Help

  • Documentation: Comprehensive guides and API reference
  • GitHub Issues: Bug reports and feature requests
  • Discussions: Community Q&A and best practices

Contributing

We welcome contributions! See our contribution guidelines for: - Code contributions - Documentation improvements - Bug reports and feature requests - Example workflows and tutorials

License

THEMAP is released under the MIT License. See LICENSE for details.

What's Next?