Skip to main content
DA / EN

Quantum Computing for Clinical Research

Investigator Team:

 
Shan Shan (SCIENCE), Jørgen Ellegaard Andersen (SCIENCE), Maja Sofie Thiele (HEALTH), Maria Kjærgaard (HEALTH) Aleksander Krag (HEALTH) Moustapha Kassem (HEALTH) and Maria Timofeeva (HEALTH).

 

1.  Introduction

Big data in clinical research. The exponential growth of healthcare data – such as disease registries, electronic health records (EHRs), results of medical examinations and tests – pointed to an overwhelming need for developing methods that address the challenges of analyzing and interpreting big data to build healthier societies. Technical challenges include a range of issues from data collection, visualization of high-dimensional data, simulation of clinical trials, interpretability, feature selection, robustness, prediction, optimization, uncertainty quantification, model validation and evaluation. With recent advances in machine learning, artificial intelligence, deep learning, and high-performance cloud computing, significant progress has been made in data-based clinical research to accelerate diagnoses, optimize pricing, enhance patient experiences, and improve healthcare practitioner work lives. However, given the rapid pace of data collected and generated from individuals, devices and systems, the need for greater computing power is stretching the capabilities of classical computing systems. 

The promise of quantum computers. Quantum computers can exponentially advance computational power and solve problems that cannot be tackled by classical machines. In health and biomedical research, quantum computers have the potential to enable numerous computation-intensive applications, such as supersonic drug design, in silico clinical trials simulation, medical imaging analysis. However,  recent efforts in the field have  mainly focused on  the problems of protein folding and protein design using either a quantum annealing approach and the quantum gate approach. The quantum annealing approach recasts the research question as a Quadratic Unconstrained Binary Optimization (QUBO) program and solves the reformulated problem on an abiadiatic quantum device (e.g., The Advantage quantum computer developed by D-Wave). The quantum gate approach maps the classical algorithm to its corresponding quantum circuit and runs the circuit on a universal quantum computer, of which we are starting to see the first few prototypes in existence now and the technological development is expected to strongly accelerate over the coming years.

Advantage of Gaussian Boson Samplers. There has been little work that addresses other aspects of the above- mentioned challenges in healthcare research in the era of big data, nor using other emerging quantum technologies, such as Gaussian Boson Samplers (GBS). GBS is a photonic quantum model that has been demonstrated with quantum advantage on a special sampling task. In 2021, Zhong et al. reported a 144-mode GBS that yields a sampling rate approximately 1024 times faster than using brute-force simulation on classical supercomputers. Compared to other leading quantum efforts, the biggest advantage of GBS is that GBS can work in room temperature, making it a promising candidate for wide use in the future. The objective of this proposal is to develop machine learning methods in clinical research based on a new quantum technology – Gaussian boson samplers (GBS). The significance of this proposal is that enabling usage of GBS in healthcare would help exploit the near-term quantum technologies to their full potential, and opens a vast and uncharted space of quantum methods in health and clinical research.

 

2 Specific Research Aims

As a first concrete target, we propose to explore use cases of GBS with the following research aims (A):

A1: Optimize pathways for accurate referral of fatty liver disease patients. We will develop a new GBS- based classification algorithm that identify patients with advanced liver fibrosis from three non-invasive bio-markers using simple blood tests and ultrasound-based screening tools. The new classification method will exploit the graph- theoretical framework for robust fitting as we have developed in, and will be suitable to deal with noisy data – for example, due to measurement error or missing tests. The new method can potentially improve and accelerate the diagnoses of advanced liver fibrosis using non-invasive tests.

A2: Improve the clinical efficacy of transplanted human bone marrow stromal cells. We will develop a new quantum-enhanced feature selection method to identify a subset from donor- and cell-related features for the ability of stem cells to form bone in vitro. We will first encode the information about the importance and correlation of features into a weighted graph, and then use GBS to find important features through clustering on the weighted graph. The proposed new feature selection technique could enhance classification accuracy and offer greater interpretability for the classification model.

A3: Enhance the prediction accuracy for colorectal cancer risk using genetic, biochemical and clinical data from UK Biobank. We will use the method proposed in A2 to identify key features in genetic, biochemical and clinical data to improve the prediction accuracy for colorectal cancer risk. We propose a new hybrid framework that enables the quantum-enhanced feature selection method of A2 to be implementable on a super large graph. The framework consists of first using a classical trimming algorithm to filter out inactive vertices, and then using GBS to cluster the remaining active vertices. The proposed technique enables application of GBS algorithms to massive graphs stemmed from real problems that often have millions of vertices and hundreds of millions of edges, and further improve the accuracy for colorectal cancer risk prediction.

Interdisciplinary aspect. The research aims (A1-3) can only be achieved through an interdisciplinary study that combines quantum computing, statistics, machine learning, health and biomedical research in order to explore the full potential of quantum technologies for real-world health challenges. The proposed research will build on joint efforts of the following personnel. Professor Jørgen Ellegaard Andersen (DIAS Chair of Quantum Mathematics) and Post-doc Shan Shan from Centre for Quantum Mathematics, whose expertise are in photonic quantum computing technologies and toplogical quantum field theory.Professor Aleksander Krag (DIAS Chair of Health Sciences), Professor Maja Sofie Thiele and PhD student Maria Kjærgaar from the Center of Liver Research.Professor Moustapha Kassem (DIAS Chair of Health Sciences) from KI, Endocrinology.Professor Maria Timofeeva (DIAS Fellow of Health Sciences) from the Department of Public Health.

Potential impact to science and society. One of the biggest challenges in the near-term quantum computing technologies is to identify specific problems of practical interest for which these devices can prove advantageous. The success of the new quantum algorithms developed in the proposal will significantly enlarge the applicability of the current and emergent quantum computing technologies. Healthcare and medicine are key fields for future high-impact quantum application because of their large and increasing demand for computing power. A quantum workforce within these fields will bring groundbreaking change to our everyday life. In addition to the three problems – diagnosing advanced liver fibrosis, identifying key features for osteoblastic differentiation in hBMSCs, and predicting colorectal cancer risk – as we proposed to study above, our methods also enjoy applications other problems. Density estimation, feature selection and classification are fundamental tasks in statistical machine learning. A quantum solution to these problem would potentially help address other important issues in society.

Relevancy to DIAS. The proposed research is truly interdisciplinary, because it tackles a challenging problem – whose solutions are beyond the scope of a single discipline – through integrating information, data, techniques, tools, perspectives, concepts from both clinical research and quantum computing. Thus, this proposal aligns with the mission of DIAS that “brings together outstanding researchers together in an interdisciplinary center for fundamental research and intellectual fora”. While the proposed research focuses on applications in clinical research, such as liver fibrosis, bone formation, and colorectal cancer, the quantum-based machine learning techniques we will be developing in this research have a wide application in all data-intensive fields. This project will be a first step towards promoting a cross- disciplinary approach to quantum computing which brings together experts in quantum mathematics, engineering, computer science, psychology, history, and other research fields at DIAS. We will discuss this aspect in more details through the planned activities in Section 4.

 

You can find out more about the research programme here

Sidst opdateret: 09.05.2023