Qure AI is a hybrid AI + Quantum pipeline designed to discover and validate potential drug molecules targeting the bacteria that cause meningitis.

• AI/ML Component: A machine learning model (Support Vector Machine) is trained on molecular datasets (sourced from ChEMBL) to identify effective drug candidates. We also propose novel SMILES strings as potential cures.
• Quantum Component: The Variational Quantum Eigensolver (VQE) algorithm is applied to calculate molecular ground-state energies. Lower energies indicate greater molecular stability, providing a validation step for both existing and AI-generated molecules.

This project demonstrates the synergy of machine learning and quantum simulation for next-generation drug discovery.

Meningitis remains a critical global health burden:

• Every year, 236,000 people die of meningitis.
• Over 2 million new cases are diagnosed annually.
• That’s 1 death every 2 minutes and 1 diagnosis every 15 seconds.
• 1 in 5 survivors live with permanent disabilities (brain damage, limb amputations, kidney failure).
• Within 1–2 years, the majority of antibodies decrease rapidly.

By integrating AI and quantum computing, Qure AI aims to accelerate drug discovery and validation, potentially reducing timelines and costs compared to traditional lab methods.

We studied molecules that attack four major bacterial pathogens responsible for meningitis:

• Neisseria meningitidis
• Streptococcus pneumoniae
• Haemophilus influenzae
• Streptococcus agalactiae

• Source: ChEMBL Database
• Features used:
  • SMILES strings
  • Molecular weight
  • MIC (Minimum Inhibitory Concentration)
  • MBC (Minimum Bactericidal Concentration)
  • Selective Ratios
  • IZ (Inhibition Zone)
  • IC50 values
  • Log₁₀CFU
  • AlogP
  • Lipinski’s Rule of 5

• Model Used: Support Vector Machine (SVM)
• Objective: Classify molecules as effective/ineffective against target bacteria
• Generative AI: Produce novel SMILES molecules with drug-like properties

• Framework: Qiskit + PySCF
• Process:
  1. Convert molecules into Hamiltonians
  2. Choose an appropriate Ansatz (EfficientSU2)
  3. Implement a classical optimizer (COBYLA)
  4. Run Variational Quantum Eigensolver (VQE)
  5. Compare ground-state energies in Hartree (1 Hartree = 27.2 eV)
• Interpretation:
  • Lower energy → more stable molecule
  • Validates both existing and AI-generated molecules

flowchart TD
    A[📂 Clean Dataset] --> B[🤖 Train SVM Model on Molecular Features]
    B --> C[⚛️ Run VQE Simulation for Quantum Validation]
    C --> D[🧬 Generate Novel Molecules via Machine Learning]
    D --> E[📊 Compare Novel vs Existing Molecules by Calculating Energy Levels]

Python, HTML, CSS, Qiskit, PySCF, Pandas, RDKit, NumPy, scikit-learn, Matplotlib, three.js, chart.js, vanta.js

The VQE quantum simulation results demonstrate successful energy convergence for both existing and AI-generated molecules:

• The generated molecule is 1.98x more stable than existing molecule

Quantum Computing & VQE:

• Hamiltonian Construction
• Ansatz
• Classical Optimizers
• Ground State

Machine Learning & ChEMBL:

• pChEMBL Value Definitions

Meningitis Statistics & General Information:

• Global Meningitis Burden
• Meningitis Statistics
• CDC Vaccine Information
• Antibiotic Resistance Data

Target Bacteria (ChEMBL):

• Neisseria meningitidis
• Streptococcus pneumoniae
• Haemophilus influenzae
• Streptococcus agalactiae

Parameter Research:

• MIC (Minimum Inhibitory Concentration)
• IC50
• MBC (Minimum Bactericidal Concentration)
• AlogP