alethiotx 2.1.5

Alethio Therapeutics Python Toolkit

alethiotx

Alethio Therapeutics Python Toolkit - A growing collection of open-source computational tools used by Alethio Therapeutics.

Overview

alethiotx is a modular Python package providing specialized tools for therapeutic research and drug discovery. Currently, the package features the Artemis module for drug target prioritization using public knowledge graphs. Additional modules and capabilities will be added in future releases.

Current Modules

Artemis Module (alethiotx.artemis)

The Artemis module enables accessible and scalable drug target prioritization by integrating drug molecule and target data from ChEMBL (including clinical trial phases and approvals), MeSH disease hierarchies, HGNC gene families, pathway information from GeneShot, and machine learning pipelines. It leverages public knowledge graphs to prioritize therapeutic targets across multiple disease areas.

Artemis Module Features

• ChEMBL Integration: Query and process ChEMBL bioactive molecule database with clinical trial information and automatic parent molecule normalization
• MeSH Hierarchy: Retrieve MeSH disease trees and descendants for comprehensive disease coverage
• HGNC Gene Families: Download and analyze gene family data to identify and filter over-represented families
• Clinical Scoring: Calculate clinical validation scores for drug targets based on trial phases, approvals, and family representation
• Pathway Genes: Retrieve and analyze disease-associated genes using Ma'ayan Lab's GeneShot API
• Machine Learning Pipeline: Built-in cross-validation with configurable classifiers for target prediction
• UpSet Plots: Visualize gene set intersections across multiple diseases
• Multi-Disease Support: Pre-configured for breast, lung, prostate, melanoma, bowel cancer, diabetes, and cardiovascular disease

Future Modules

Additional modules for various aspects of drug discovery and therapeutic research are planned for future releases. Stay tuned!

Installation

pip install alethiotx

Quick Start

Note: The examples below demonstrate the Artemis module functionality. As new modules are added to the package, they will have their own usage examples.

1. Query ChEMBL and Compute Clinical Scores

from alethiotx.artemis.chembl import molecules
from alethiotx.artemis.clinical import compute

# Query ChEMBL for parent molecules with clinical trial data
chembl_data = molecules(version='36', top_n_activities=1)

# Compute clinical validation scores for specific diseases
results = compute(
    mesh_headings=['Breast Neoplasms', 'Lung Neoplasms'],
    chembl_version='36',
    trials_only_last_n_years=6,
    filter_families=True
)

# Access results for each disease
breast_targets = results['Breast Neoplasms']
print(breast_targets.head())

2. Load Pre-computed Clinical Scores

from alethiotx.artemis.clinical import load

# Load pre-computed clinical scores for multiple diseases from S3
breast, lung, prostate, melanoma, bowel, diabetes, cardio = load(date='2025-12-08')

3. Pathway Gene Analysis

from alethiotx.artemis.pathway import get, load

# Query GeneShot API for disease-associated genes
aml_genes = get("acute myeloid leukemia", rif='generif')
print(aml_genes.loc["FLT3", ["gene_count", "rank"]])

# Load pre-computed pathway genes for multiple diseases
breast_pg, lung_pg, prostate_pg, melanoma_pg, bowel_pg, diabetes_pg, cardio_pg = load(date='2025-11-11', n=100)

4. Machine Learning Pipeline

from alethiotx.artemis.cv import prepare, run
import pandas as pd

# Prepare your knowledge graph features (X) and clinical scores (y)
result = prepare(
    X, 
    y, 
    pathway_genes=pathway_genes, 
    known_targets=known_targets,
    bins=3,
    rand_seed=12345
)

# Run cross-validation pipeline
scores = run(
    result['X'], 
    result['y_binary'], 
    n_splits=5, 
    n_iterations=10, 
    classifier='rf',
    scoring='roc_auc'
)
print(f"Mean AUC: {sum(scores)/len(scores):.3f}")

5. Visualize Gene Overlaps with UpSet Plots

from alethiotx.artemis.upset import prepare, create
from alethiotx.artemis.clinical import load
from alethiotx.artemis.pathway import load as load_pathway

# Load clinical scores for multiple diseases
breast, lung, prostate, melanoma, bowel, diabetes, cardio = load(date='2025-12-08')

# Prepare data for UpSet plot (mode='ct' for clinical targets)
upset_data = prepare(breast, lung, prostate, melanoma, bowel, diabetes, cardio, mode='ct')

# Create and display the UpSet plot
plot = create(upset_data, min_subset_size=5)
plot.plot()

# For pathway genes, use mode='pg'
breast_pg, lung_pg, prostate_pg, melanoma_pg, bowel_pg, diabetes_pg, cardio_pg = load_pathway(date='2025-11-11', n=100)
upset_data_pg = prepare(breast_pg, lung_pg, prostate_pg, melanoma_pg, bowel_pg, diabetes_pg, cardio_pg, mode='pg')
plot_pg = create(upset_data_pg, min_subset_size=10)
plot_pg.plot()

Supported Disease Indications (Artemis Module)

The Artemis module includes built-in pre-computed data for:

• Breast Cancer (Breast Neoplasms)
• Lung Cancer (Lung Neoplasms)
• Prostate Cancer (Prostatic Neoplasms)
• Melanoma (Skin Neoplasms)
• Bowel Cancer (Intestinal Neoplasms)
• Diabetes Mellitus Type 2
• Cardiovascular Disease

The module supports querying any disease with MeSH headings via the compute() function.

Artemis Module API Reference

ChEMBL Module (alethiotx.artemis.chembl)

• molecules(version, top_n_activities) - Query ChEMBL for parent molecules with clinical trial data
• infer_nct_year(nct_id) - Infer registration year from ClinicalTrials.gov NCT identifier

Clinical Scores Module (alethiotx.artemis.clinical)

• compute(mesh_headings, chembl_version, trials_only_last_n_years, filter_families) - Compute clinical validation scores for drug targets
• load(date) - Load pre-computed clinical scores from S3
• lookup_drug_family_representation(chembl) - Create drug-disease-family representation lookup table
• filter_overrepresented_families(targets_df, drug_chembl_id, mesh_heading, lookup_table) - Filter over-represented gene families
• unique(scores, overlap, common_genes) - Remove overlapping genes from clinical scores
• approved(scores) - Filter to include only approved targets
• all_targets(scores) - Extract all unique target genes from score lists

Pathway Genes Module (alethiotx.artemis.pathway)

• get(search, rif) - Query Ma'ayan Lab's GeneShot API for disease-associated genes
• load(date, n) - Load pre-computed pathway genes from S3
• unique(genes, overlap, common_genes) - Remove overlapping genes from pathway lists

MeSH Module (alethiotx.artemis.mesh)

• tree(s3_base, url_base, file_base) - Retrieve MeSH tree structure
• descendants(heading, s3_base, file_base, url_base) - Get all descendant MeSH headings

HGNC Module (alethiotx.artemis.hgnc)

• download(gene_has_family_url, family_url, hgnc_complete_url) - Download HGNC gene family data
• process(gene_has_family, family, hgnc_data) - Process HGNC data and create gene-family mappings

Machine Learning Module (alethiotx.artemis.cv)

• prepare(X, y, pathway_genes, known_targets, term_num, bins, rand_seed) - Prepare datasets for ML model training
• run(X, y, n_splits, n_iterations, classifier, scoring) - Cross-validation pipeline with configurable classifiers

Visualization Module (alethiotx.artemis.upset)

• prepare(breast, lung, prostate, melanoma, bowel, diabetes, cardiovascular, mode) - Prepare data for UpSet plot
• create(indications, min_subset_size) - Create UpSet plots for visualizing gene set intersections

Utilities (alethiotx.artemis.utils)

• find_overlapping_genes(genes, overlap, common_genes) - Find genes that overlap across multiple gene lists

Data Storage (Artemis Module)

The Artemis module uses AWS S3 for storing pre-computed data:

s3://alethiotx-artemis/data/
├── clinical_scores/{date}/{disease}.csv
├── pathway_genes/{date}/{disease}.csv
├── chembl/{version}/molecules.csv
└── mesh/d{year}.pkl

Requirements

• Python >= 3.9
• requests
• scikit-learn
• pandas
• numpy
• setuptools
• fsspec
• s3fs
• upsetplot
• chembl-downloader

Citation

If you use the Artemis module in your research, please cite:

Artemis: public knowledge graphs enable accessible and scalable drug target discovery
Vladimir Kiselev, Alethio Therapeutics

For other modules, citation information will be provided as they are released.

License

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

Author

Vladimir Kiselev
Email: [e-mail]

Links

• Homepage: https://github.com/alethiotx/pypi
• Issues: https://github.com/alethiotx/pypi/issues

Contributing

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

Acknowledgements

• Public knowledge graph providers (Hetionet, BioKG, OpenBioLink, PrimeKG)
• PyKEEN, scikit-learn, and Nextflow communities
• ChEMBL and MeSH data sources
• Portions of this codebase were generated, refactored, and/or cleaned using GitHub Copilot (Claude Sonnet 4.5). The authors reviewed, modified, and validated all AI-assisted code. Responsibility for the correctness, performance, and reproducibility of the code rests entirely with the authors.

---

Current Focus: Artemis - Enabling accessible and scalable drug target discovery through public knowledge graphs.
Coming Soon: Additional modules for expanded drug discovery capabilities.