Drug Target Interaction prediction using Variational Quantum classifier

The need to discover therapeutic compounds has intensified in recent years due to the rise of antibiotic resistance and complexity of modern diseases. The conventional drug discovery pipeline is resource-intensive. The field of drug development has evolved into interdisciplinary science. Several computational techniques have been devised to assist researchers, in the identification of new drugs. These tools often employ machine learning algorithms that are typically executed on classical computing machines. However, recent advancements in the field of quantum computing have offered new possibilities to improve machine learning models. Quantum computing leverages unique quantum-mechanical characteristics to achieve more efficient computations. In this paper, we introduce a framework that utilizes Variational Quantum Classifier (VQC) techniques for the classification of drug-protein interactions, as shown in Figure 1. This approach exploit quantum computing to produce better results than classical computing methods. To validate the efficacy of our proposed framework, we utilized the KIBA dataset to assess the drug-target affinity score. The model was trained at a learning rate of 0.0001 over 500. Through a 3-fold cross-validation, we achieved a concordance index (CI) of 0.802. This performance surpasses that of linear regression technique (kronLRS), which has a concordance index of 0.68 and is comparable to deep learning methods, such as DeepCPI and DeepDTA, which have a CI of 0.86. Notably, standard VQC methods typically attain a CI of approximately 0.75, which is a metric that can be further optimized by adjusting the number of qubits or wires in the circuit.