EAGER-QAC-QCH: Hybrid Quantum Classical Algortithm for NMR Inference

Harvard University, Cambridge MA

Investigators

Abstract

Eugene Demler of Harvard University is supported by an EAGER award from the Chemical Theory, Models and Computational Methods program in the Division of Chemistry to study Hybrid Quantum-Classical Algorithms for NMR Inference. The Condensed Matter and Materials program in the Division of Materials Research also cofunds this award. The proposal was submitted in response to the Quantum Algorithm Challenge Dear Colleague Letter, NSF 20-056. NMR is one of the most powerful analytical techniques available to medicine and biology. It is suited for both in vivo and in vitro studies. Yet it is difficult to interpret the data in NMR experiments. Spectral profiling of compounds is a complex pattern recognition problem. Spectroscopic analysis and interpretation for NMR compound identification is cumbersome and slow. It can be inconsistent across platforms. It is generally hard to scale it to novel compounds, an important obstacle to drug discovery and identification of mechanism of action. Therefore, compound identification is a major challenge for implementing these technologies in medicine, engineering, and science. Quantum correlations and entanglement rather than traditional correlations define NMR spectra. Therefore, quantum approaches to solve spectroscopic inference problems are well-suited to achieve a sought-after quantum advantage. Eugene Demler is developing hybrid approaches that combine the tools of data science, deep learning methods, with quantum computing to address the problem of inference of spectral analysis. This work can have a wide impact in many applications of NMR in medicine, science and engineering, with an immediate application in metabolomics compound identification, which focuses on profiling small molecules inside cells, organs and body fluids. NMR is one of the most powerful analytical techniques available to medicine and biology, as it is suited for both in vivo and in vitro studies. Yet it is difficult to interpret the data in NMR experiments. One directly observes only the magnetic spectrum of a biological sample, whereas the ultimate goal is to learn about the underlying microscopic Hamiltonian and ultimately identify and quantify chemical compounds. Eugene Demler is developing a hybrid approach that combines quantum computing, quantum simulations, and classical machine learning to address the problem of NMR inference. The novelty of this approach is in using quantum simulators to compute spectra for hypothetical values of the Hamiltonians, and then using classical deep-learning algorithms to optimize these parameters. Eugene Demler’s work address the following specific questions: (i) Finding optimal protocols for sampling NMR spectra on currently available quantum computing platforms with a focus on ion chains and Rydberg arrays; (ii) improving hybrid algorithms through finding more efficient methods of Trotterizing the Hamiltonian evolution; (iii) improving the classical optimization procedure using LASSO regularization of the variational Bayesian Gaussian method; (iv) establishing the theoretical limits to quantum assisted NMR inference ; (v) collaborating with experimental groups to provide a proof of principle experimental demonstration of quantum assisted NMR inference. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.