clinicaltrials-interact 0.1.5

A Python package to interact with ClinicalTrials.gov API v2

Clinical Trials Interact

A Python package for interacting with and analyzing clinical trial data from ClinicalTrials.gov API.

Features

• Search clinical trials using flexible search expressions
• Build similarity graphs between related clinical trials
• Navigate through clinical trial networks using graph traversal
• Extract detailed trial information and metadata
• Analyze relationships between trials based on various criteria

Installation

Install using pip:

pip install clinicaltrials_interact

Note: For an isolated environment, consider using a virtual environment or Conda:

conda create -n ct_interact_env python=3.10
conda activate ct_interact_env
pip install clinicaltrials_interact

Requirements

• Python 3.7+
• requests
• networkx
• pandas
• matplotlib
• sentence-transformers
• scikit-learn
• seaborn
• plotly

Quick Start

from clinicaltrials_interact import ClinicalTrialsAPI

# Initialize the API client
ct_api = ClinicalTrialsAPI()

# Search for clinical trials related to "breast cancer"
trials = ct_api.query_studies("breast cancer", max_results=100)

# Get detailed information about a specific trial
trial_detail = ct_api.get_study_details("NCT04303169")

Examples

Example 1: Analyzing COVID-19 Vaccine Trials

from clinicaltrials_interact import ClinicalTrialsNavigator
import matplotlib.pyplot as plt

# 1. Define the navigator
navigator = ClinicalTrialsNavigator()
    
# 2. Search for studies and build a similarity graph in one step
print("\nSearching for diabetes studies and building graph...")
graph = navigator.search_and_build_graph(
    search_expr="diabetes type 2", 
    max_studies=200,  # Limit to 50 studies for this example
    similarity_threshold=0.5,  # Lower threshold to get more connections
    max_edges_per_node=20
)

# 3. Perform breadth-first search traversal
print("\nPerforming Breadth-First Search traversal:")
bfs_results = navigator.breadth_first_search(
    start_id=first_trial_id,
    max_depth=2,
    visited_limit=10
)

print(f"BFS found {len(bfs_results)} trials:")
for i, result in enumerate(bfs_results[:5]):  # Show first 5 results
    print(f"{i+1}. {result['NCTId']} (depth {result['depth']}): {result['title'][:50]}...")

# 4. Perform depth-first search traversal
print("\nPerforming Depth-First Search traversal:")
dfs_results = navigator.depth_first_search(
    start_id=first_trial_id,
    max_depth=2,
    visited_limit=10
)

print(f"DFS found {len(dfs_results)} trials:")
for i, result in enumerate(dfs_results[:5]):  # Show first 5 results
    print(f"{i+1}. {result['NCTId']} (depth {result['depth']}): {result['title'][:50]}...")

# 5. Find a path between two trials
if len(trial_ids) >= 2:
    target_id = trial_ids[5]  # Pick the 6th trial as the target
    print(f"\nFinding path from {first_trial_id} to {target_id}:")
    path = navigator.find_path(first_trial_id, target_id)
    
    if path:
        print(f"Path found with {len(path)} nodes:")
        for i, node_id in enumerate(path):
            node_data = graph.nodes[node_id]
            print(f"{i+1}. {node_id}: {node_data.get('title', '')[:50]}...")
    else:
        print("No path found.")

# 6. Find connected component
print(f"\nFinding connected component for trial {first_trial_id}:")
connected = navigator.get_connected_component(first_trial_id)
print(f"Connected component has {len(connected)} trials.")

# 7. Visualize the graph with highlighted path and nodes
print("\nVisualizing the graph (a plot window should appear)...")
if len(trial_ids) >= 2:
    navigator.visualize_graph(
        highlight_nodes=[first_trial_id, target_id],
        highlight_path=path,
        figsize=(10, 8)
    )
else:
    navigator.visualize_graph(highlight_nodes=[first_trial_id])

Example 2: Finding Similar Trials

import clinicaltrials_interact
from clinicaltrials_interact import ClinicalTrialsAPI
import clinicaltrials_interact.clustering as clustering 

ct_api = ClinicalTrialsAPI()

# 1a. Retrieve candidate studies related to "leukemia" and compute embeddings. 
df, embeddings, model = clustering.get_candidate_embeddings("leukemia", max_studies=100)

# 1b. Plot the cosine similarity matrix for a subset (e.g., 50 studies) of the candidate pool.
# Saved to figures directory
clustering.plot_similarity_matrix(embeddings, subset_size=50)

# 1c. Output the index to id mapping to understand which studies are most to each other. 
clustering.plot_index_to_id_mapping(df, subset_size=50)

# 1d. Perform spectral clustering on the embeddings.
labels, sim_matrix = clustering.perform_spectral_clustering(embeddings, n_clusters=3)
    
# 1e. Visualize the clusters with a t-SNE plot.
clustering.plot_clusters(embeddings, labels, filename="leukemia_clusters.png", keyword = "leukemia")

# 2a. For a given query, return the IDs of the top 10 most similar studies.
query = "A study about heart attacks"
top_similar_ids = clustering.get_top_similar_ids(query, "", candidate_pool_size=500, top_n=10)
print(f"Top similar study IDs related to your query: {top_similar_ids}\n")

# 3. Perform general clustering and extract keywords for each cluster. Takes the top 500 or so studies and clusters them into 5 clusters. Returns the keywords for each cluster
clustering.print_cluster_keywords()

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add some amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Citation

If you use this package in your research, please cite:

KeyVuLee Innovations. (2025). Clinical Trials Interact: A Python package for analyzing clinical trial data.
GitHub: https://github.com/hsph-bst236/midterm-project-keyvulee-innovations

Contact

For questions and support, please open an issue or contact the maintainer at [e-mail].