Skip to content

Getting Started Tutorial

This tutorial will walk you through the basic concepts and usage of THEMAP for task hardness estimation and distance computation.

Prerequisites

  • THEMAP installed with basic dependencies
  • Python 3.10+
  • Basic knowledge of molecular datasets

Tutorial Overview

In this tutorial, you will learn how to:

  1. Load molecular and protein datasets
  2. Compute basic distances between datasets
  3. Work with the unified task system
  4. Interpret distance results

Step 1: Loading Your First Dataset

from themap.data import MoleculeDataset
from dpu_utils.utils.richpath import RichPath

# Load a molecular dataset
dataset_path = RichPath.create("datasets/train/CHEMBL1023359.jsonl.gz")
dataset = MoleculeDataset.load_from_file(dataset_path)

print(f"Loaded dataset with {len(dataset)} molecules")
print(f"Task ID: {dataset.task_id}")
print(f"Sample molecule: {dataset[0].smiles}")

Step 2: Computing Simple Distances

from themap.distance import MoleculeDatasetDistance

# Load two datasets to compare
source_dataset = MoleculeDataset.load_from_file(
    RichPath.create("datasets/train/CHEMBL1023359.jsonl.gz")
)
target_dataset = MoleculeDataset.load_from_file(
    RichPath.create("datasets/test/CHEMBL2219358.jsonl.gz")
)

# Create distance calculator
distance_calc = MoleculeDatasetDistance(
    tasks=None,
    molecule_method="euclidean"  # Start with fastest method
)

# Set up the comparison
distance_calc.source_molecule_datasets = [source_dataset]
distance_calc.target_molecule_datasets = [target_dataset]
distance_calc.source_task_ids = [source_dataset.task_id]
distance_calc.target_task_ids = [target_dataset.task_id]

# Compute distance
result = distance_calc.get_distance()
print(f"Distance: {result}")

Step 3: Working with Protein Data

from themap.data import ProteinMetadataDatasets
from themap.distance import ProteinDatasetDistance

# Load protein sequences
proteins = ProteinMetadataDatasets.from_directory("datasets/train/")
print(f"Loaded {len(proteins)} protein sequences")

# Compute protein similarities
protein_distance = ProteinDatasetDistance(
    tasks=None,
    protein_method="euclidean"
)

protein_distance.source_protein_datasets = proteins
protein_distance.target_protein_datasets = proteins

distances = protein_distance.get_distance()
print("Protein distance matrix computed")

Step 4: Understanding Results

Distance values have the following interpretations:

  • Lower values: More similar datasets
  • Higher values: More different datasets
  • Scale: Depends on the method used
# Analyze distance results
for target_id, source_distances in result.items():
    print(f"Target task: {target_id}")
    for source_id, distance in source_distances.items():
        print(f"  Distance from {source_id}: {distance:.3f}")

Next Steps

Now that you understand the basics:

  1. Try different distance methods (cosine, otdd)
  2. Explore the unified task system
  3. Learn about task hardness estimation
  4. Check out the advanced examples

Common Issues

Import Errors

Make sure all dependencies are installed: 

pip install -e ".[ml]"

Memory Issues

Use euclidean distance for large datasets instead of OTDD.

File Not Found

Ensure your data files are in the correct directory structure.