ReDrugAI - Drug Repurposing Recommendation System

[Hakathon page] (https://www.kaggle.com/competitions/bigquery-ai-hackathon/writeups/redrugai0421)

Overview

ReDrugAI is an AI-powered drug repurposing recommendation system built on Google BigQuery and BigQuery AI. The system leverages machine learning embeddings and vector similarity search to identify potential therapeutic applications for existing drugs in new disease contexts.

Python Dependencies

• Python: Version >= 3.12.7
• Dependencies: Install required packages from requirements.txt

Architecture Diagram

Simple overview of the ReDrugAI system architecture showing the main data flow and components

Complex detailed view of the system architecture including all technical components, data processing pipelines, and AI model interactions

Core Components

main.ipynb

The main project file - A comprehensive Jupyter notebook that provides complete explanation and step-by-step implementation of the ReDrugAI system:

• Prepare Table: Data preprocessing, embedding table creation with vector index, and similarity matrix construction for drug-disease relationships
• Vector Search: Implementation of embedding-based similarity search for drug recommendations
• Performance Evaluation: Comprehensive testing and validation of recommendation accuracy

score.py

Main recommendation engine containing:

• recommend_for_disease_with_similars(): Primary recommendation function
• MOA similarity scoring algorithms
• Drug candidate filtering and ranking

evaluation.py

Evaluation framework featuring:

• run_disease_evaluation(): Single disease evaluation
• calculate_evaluation_metrics(): Precision/recall calculation
• Leave-one-out cross-validation methodology