Quantum Machine Learning for Efficient Drug Discovery

Abstract

Computational development of a drug requires the analysis of complex molecular structures, the prediction of molecular properties and identifying promising drug candidates in large chemical spaces. However, the conventional computational approaches typically have scalability problems, computational costs and processing efficiency in the case of high dimensional molecular data. In the present paper, a hybrid QML framework is proposed, in which the QNNs and VQCs are combined with classical ML techniques to speed up the drug discovery process. We propose to encode the molecular information into quantum data encoding, learn the features by variational quantum circuits, and predict chemical properties by the quantum neural networks. We use a hybrid quantum-classical optimisation technique to break the limits of current quantum hardware and boost the accuracy of predictions. The system is implemented using Python with Qiskit, PennyLane, TensorFlow Quantum and Scikit-learn to enable efficient molecular analysis and candidate generation. Experimental results show the efficiency of the proposed architecture for the prediction of molecular properties and binding affinity estimation, which is scalable to a quantum-assisted computational chemistry. The proposed method highlights the potential of quantum computing in solving complex problems, enhancing predictive accuracy, and expediting the future of drug discovery.