Exploration and Development of Quantum-HPC Hybrid Applications for Life Sciences, Healthcare, and Drug Discovery

Quantum machine learning (QML) is expected to offer capabilities beyond those of conventional machine learning. However, given the limited number of qubits available on current quantum computers, it remains difficult to perform practical machine learning tasks using QML alone. We will therefore explore life science problems in which the combination of GPUs, QPUs, and HPC resources can provide unique value, and investigate ways to demonstrate the utility of such hybrid approaches. One possible approach is to use GPUs to reduce the search space and subsequently apply QML on QPUs. A key challenge in this approach is to develop effective methods for encoding and embedding the data or features generated by classical computation into quantum states or quantum circuits.

We will also investigate a hybrid computational approach in which GPUs are used for machine learning and feature extraction, QPUs for representing and sampling high-dimensional probability distributions, and HPC systems for large-scale data processing, analysis, and validation. Such an approach will enable us not only to evaluate and demonstrate the effectiveness of GPU-QPU-HPC hybrid computing, but also to take full advantage of the advanced quantum-HPC computing platform that has been developed through this project.

Specifically, we will explore the use of quantum computers to address the following issues in the medical field.

In the biomedical sciences, rapid advances in omics technologies have led to the accumulation of large-scale multimodal datasets that integrate omics data with clinical information, medical imaging, and other data modalities, forming an increasingly important source of big data for medicine and life sciences. These datasets are challenging not only because of their sheer size, but also because of their high dimensionality, complex interactions, heterogeneity, sparsity, and nonlinearity. Consequently, there is a growing need for advanced analytical approaches that go beyond conventional computational methods. We propose a new computational paradigm to address these challenges by leveraging the rapidly advancing capabilities of quantum computing and generative AI. Since the data used in biomedical research is essentially classical and does not inherently possess quantum properties, a key challenge is how to effectively encode such data into quantum circuits. We will therefore investigate methods for efficiently integrating quantum computing with high-performance computing (HPC) for large-scale data processing, as well as approaches for designing appropriate quantum circuits based on the characteristics and structure of biomedical data.

Recent advances using the Sample-Based Quantum Diagonalization (SQD) method have enabled highly accurate ground-state energy calculations for biologically relevant molecular systems, such as iron-sulfur clusters. Furthermore, by decomposing large biomolecular systems into fragments and applying SQD calculations to individual fragments, it has become possible to investigate the energies and docking properties of large-scale biomolecular systems. Understanding various phenomena in biomolecules also requires the analysis of electron transport processes. Hybrid approaches are therefore being explored in which the overall biomolecular system is analyzed using conventional HPC applications, while quantum computers are selectively applied to regions where detailed electronic-structure or electron-transport analysis is required. In drug discovery in particular, these approaches need to be further advanced and extended. By integrating quantum-HPC hybrid simulations with GPU-accelerated machine learning, we aim to enable more sophisticated analyses of large-scale biomolecular systems and their interactions with drug candidates.