Quantum-Enhanced Molecular Binding Affinity Prediction

## Background Molecular binding affinity prediction is a cornerstone of computational drug discovery. Accurately predicting how strongly a candidate drug molecule binds to its target protein determines which compounds advance through the drug development pipeline — and which are discarded early, saving years of laboratory work and hundreds of millions in failed clinical trials. Current state-of-the-art classical methods, primarily Density Functional Theory (DFT) and molecular dynamics simulations, achieve moderate accuracy (MAE ~1.5–2.0 kcal/mol) but face fundamental scaling limitations. Screening a library of 10,000+ compounds against a single protein target requires weeks of compute on traditional HPC clusters, creating a bottleneck that limits the breadth of drug candidate exploration. Quantum computing offers a theoretically grounded advantage for molecular simulation: the electronic structure of molecules is inherently quantum mechanical. Hybrid quantum-classical algorithms — particularly Variational Quantum Eigensolver (VQE), Quantum Approximate Optimization (QAOA), and quantum kernel methods — have shown promising results on small molecular systems (< 20 atoms). The open question is whether these advantages translate to pharmaceutical-scale molecules (50–200+ atoms) in a cost-effective manner. ## Problem Statement Develop a hybrid quantum-classical approach that predicts the binding affinity (in kcal/mol) between candidate drug molecules and target proteins with at least 30% lower Mean Absolute Error compared to the classical DFT baseline, while maintaining computational cost within 5x of the classical approach for equivalent problem sizes. Submissions will be evaluated against experimentally validated binding energy data for 50+ diverse molecule-protein complexes spanning four therapeutic areas: oncology, neurodegenerative disease, infectious disease, and cardiovascular targets. ## Objectives 1. **Accuracy improvement**: Achieve MAE ≤ 1.26 kcal/mol (30% reduction from classical DFT baseline of 1.80 kcal/mol) across the full 50-molecule test set 2. **Correlation quality**: Maintain R² ≥ 0.85 between predicted and experimental binding energies, demonstrating consistent predictive power across molecular diversity 3. **Cost efficiency**: Demonstrate quantum resource usage (circuit depth × shots) that scales sub-exponentially with molecular size, with wall-clock time within 5x of classical DFT for equivalent accuracy 4. **Reproducibility**: All quantum circuits must execute deterministically under fixed random seed (seed=42) with variance < 0.05 kcal/mol across 5 independent runs ## Organizational Context Helix Molecular's Board of Directors has allocated an **$850K pilot budget** for quantum computing investment in the drug discovery pipeline, contingent on empirical evidence that quantum-enhanced methods provide material improvement over classical approaches. The **Q3 2026 board review** serves as the decision gate. This challenge is the primary evidence-generation mechanism: results will directly inform the go/no-go decision for a $12M multi-year quantum computing program. Dr. Sarah Chen (VP Computational Chemistry) has decision authority, with CTO Office as board sponsor. The governance framework follows a **Stage-Gate model with external validation** — all winning approaches will undergo independent verification by an external computational chemistry review panel before investment recommendations are finalized. ## Scope & Constraints **In scope:** - Small to medium molecules (drug-like, MW 200–800 Da) binding to protein targets with known crystal structures - Hybrid quantum-classical algorithms using gate-based quantum computing (simulator or hardware) - Any quantum computing framework: PennyLane, Qiskit, Cirq, Braket, CUDA-Q, OpenQASM **Out of scope:** - Pure classical ML approaches (must include a quantum computing component) - Quantum annealing approaches (gate-based only for this challenge) - Covalent binding or allosteric interactions - Metalloprotein binding sites (reserved for Phase 2) **Hardware constraints:** - Evaluation runs on classical hardware with quantum simulation (no QPU access required) - Maximum 32 GB memory, 64 CPU cores, 1x NVIDIA A100 80GB GPU - Maximum 1-hour wall-clock time per molecule prediction