ramakrishna7288/explainable-drug-env

🧬 Explainable AI Drug Screening Environment

🚀 Overview

This project implements a multi-step AI environment for drug screening and repurposing using protein sequence similarity.

The system simulates how an AI agent selects the best drug for a given protein target using filtering, ranking, and explainable reasoning.

🌍 Real-World Motivation

Drug discovery is expensive and time-consuming. This environment mimics real-world computational drug screening where AI can:

• —Identify suitable drugs
• —Rank candidate molecules
• —Suggest optimal drug choices
• —Provide explanations for decisions

🧠 Environment Design

🔹 Observation Space

• —Protein name
• —Target protein sequence
• —List of drug candidates (sequence-based)

🔹 Action Space

• —filter → remove weak drugs
• —rank → rank drugs by similarity
• —select → choose best drug
• —explain → justify decision

🔄 Workflow

1. 1.AI receives protein + drug sequences
2. 2.Filters low-similarity drugs
3. 3.Ranks remaining candidates
4. 4.Selects best drug
5. 5.Provides explanation

🎯 Tasks

🎁 Reward System

The reward is multi-step:

• —+0.3 → correct filtering
• —+0.3 → correct ranking
• —+0.4 → correct selection
• —+0.2 → explanation

🧪 Example

Input

Protein: EGFR Sequence: MTEYKLVVVGAGG

Drugs:

• —A → MTEYKLVVV
• —B → GGGAVVTKLSA
• —C → TTYYGGCCAAA

Output

Best Drug: A Reason: Highest sequence similarity

⚙️ Setup

pip install -r requirements.txt
python inference.py