ETH Zurich and Copenhagen Researchers Compute the Energy of an Arginine-Tryptophan Protein Contact on Quantinuum's Helios, arXiv 2611.22834, 17 November 2026

Quentir Medicine Monitor

Evidence-based insights for quantum medicine. Published by Quentir Systems LLC · November 18, 2026.

Conceptual close-up of a clear glass pharmaceutical capsule with amber end caps, holding two linked faceted crystal rings that recall a molecular model, resting on a gold-patterned chip in a bright room with a window.

Chemists at ETH Zurich and the University of Copenhagen have used Quantinuum's Helios trapped-ion computer to estimate the electronic ground-state energy of an arginine-tryptophan pair, two amino-acid side chains that meet inside proteins, while the pair sat in a model of surrounding water. Their method, sparse quantum prolate diagonalization, measures only a random subset of the quantum signal the calculation needs and reconstructs the rest on an ordinary computer. For a reduced model of the molecule's electrons, the hardware estimates were consistent with chemical accuracy, the 1.6 millihartree margin chemists use as a working standard, within wide statistical uncertainty: the central errors ranged from about 1.5 to 3.4 millihartree.

The result concerns the quantum chemistry that sits underneath drug design and protein science, and it is deliberately small. The quantum part of the calculation used three qubits of a 98-qubit machine, and the molecular model was reduced to two electrons in two orbitals. The study is a preprint and has not yet been peer reviewed. Its value for medicine lies in the cost of the method, which the authors cut by up to an order of magnitude in simulations, and in the choice of test molecule: a protein contact in water was carried through the whole workflow on real hardware, where earlier experiments of this kind mostly used the hydrogen molecule.

Who wrote the 17 November 2026 arXiv study, and who funded it

The preprint, "Molecular energies from sparse quantum prolate diagonalization on a trapped-ion quantum processor" on arXiv, lists Davide Castaldo, Marvin Kronenberger and Markus Reiher of ETH Zurich's Department of Chemistry and Applied Biosciences; Marek Miller and Matthias Christandl of the University of Copenhagen's Department of Mathematical Sciences; Sélène Forget of the Niels Bohr Institute; Gemma C. Solomon of the Niels Bohr Institute and Copenhagen's Department of Chemistry; and the protein scientist Kresten Lindorff-Larsen, also of the University of Copenhagen. The Novo Nordisk Foundation funds the work through its Quantum for Life grant, its quantum computing program at the Niels Bohr Institute and a Synergy grant titled "Molecular Recognition from Quantum Computing". The team also acknowledges a Phase-2 finalist award from XPRIZE and Danish e-infrastructure funding for access to Quantinuum's machines.

The method builds on the original quantum prolate diagonalization paper by Stroschein, Castaldo and Reiher in Physical Review Research, published in 2026 as volume 8, article 033177. That paper showed that the technique can reach chemically accurate energies with evolution times more than ten times shorter than standard quantum phase estimation. A quantum computer running such a method prepares a starting state of the molecule, lets it evolve for a series of time steps, and measures a signal called the autocorrelation function at each step. A classical computer then extracts the energies from that signal. Each time step requires its own circuit, and each circuit must be repeated many times, so the number of time steps largely sets the bill.

How sparse sampling cuts the number of quantum circuits

The new variant exploits a property of weakly correlated molecules, the large class in which the simplest approximate description of the electrons, the Hartree-Fock state, already overlaps strongly with the true ground state. For such molecules the matrix built from the sampled signal is close to a low-rank matrix, and a mathematical technique called Hankel matrix completion can fill in the missing time points from a random subset. The authors report that the number of distinct circuits then grows only with the logarithm of the full number of time points, and that circuit depth does not increase at the target accuracy. In their simulations this reduced the total cost by up to an order of magnitude.

The test case came from protein science. Forget ran molecular dynamics simulations of the arginine-tryptophan pair in water, and the team selected 23 representative structures from that trajectory by clustering. Arginine carries a positive charge; tryptophan carries a large ring system whose electrons attract that charge. Interactions of this kind, known as cation-pi interactions, are common in proteins: in a survey of the Protein Data Bank, Gallivan and Dougherty found that over one-fourth of all tryptophans experience an energetically significant cation-pi interaction. The authors write that interactions involving such ring systems are still generally described with empirical force fields in simulations of large biomolecular complexes, which is the motivation for computing them at higher accuracy. Quentir has followed a comparable hardware experiment in drug metabolism, in which QC Ware and IonQ mapped four orbitals of a cytochrome P450nor active site onto eight qubits of IonQ Forte.

Quantum pillar: computing. Technology readiness: TRL 3 of 9. In Quentir's assessment, the method stands at proof of concept: this study ran it on real quantum hardware for one small model of a single amino-acid contact, using three qubits, and applied it to no whole protein, drug candidate or pharmaceutical workflow.

What Quantinuum's Helios measured for the arginine-tryptophan pair

For the hardware run, the team took one simulation snapshot and built a combined quantum and classical model of it in which the point charges of the surrounding water molecules enter the molecular Hamiltonian, so the solvent was part of the calculation. Helios, which the authors describe as a 98-qubit trapped-ion processor with all-to-all connectivity, ran circuits on a three-qubit register: two qubits encoded the molecule and one ancillary qubit read out the signal. The deepest circuits held 448 two-qubit gates and 1,221 gates in total. The team used no error mitigation, error detection or error correction.

The full calculation gave an error estimate of about 2.9 millihartree. Sparse reconstruction from about 39 percent of the measured points gave about 1.54 millihartree with uniform sampling and about 3.42 millihartree with Gaussian sampling, each with a wide spread across 200 reconstructions. The authors conclude that chemical accuracy lies within the experimental error bars, and they state that the uncertainties do not allow them to say which of the three variants is the most accurate. Earlier phase-estimation experiments on quantum hardware, which they review, mostly concerned the hydrogen molecule; the one experiment beyond it, on the H3+ ion, reported an error of about 25.5 millihartree.

Limits of the Helios result: one run and two orbitals

The authors list the limits plainly. Access to the machine allowed only one measurement run, so the uncertainty of the full calculation could not be estimated, and every reconstruction drew on that single data set. Each circuit was repeated only ten times on Helios. The reduction in cost rests on numerical studies of active spaces up to three orbitals with four electrons, with a scaling analysis extended in simulation to eight orbitals and eight electrons. The approach also depends on the weak-correlation property; molecules whose electrons are strongly correlated fall outside what this paper tests, and the authors leave the trade-offs in that regime to further investigation.

For a pharmaceutical reader, the paper's own conclusion is the useful guide. The authors' long-term aim is to use quantum computers to improve quantum chemical workflows, such as training machine-learned interatomic potentials for chemical reactions and conformational sampling, and they argue that routine weakly correlated chemistry could eventually be handled by quantum computation with an advantage over leading classical methods. That outcome requires far larger active spaces and quantum processors than this experiment used. What the preprint demonstrates is a cheaper sampling scheme and a complete hardware workflow for a solvated protein contact, from molecular dynamics to compiled circuits to classical post-processing.

Sources

Primary source: preprint by Davide Castaldo, Marvin Kronenberger, Marek Miller, Sélène Forget, Gemma C. Solomon, Kresten Lindorff-Larsen, Matthias Christandl and Markus Reiher (ETH Zurich and the University of Copenhagen), arXiv, 17 November 2026. Also drawn on: the original quantum prolate diagonalization paper by Stroschein, Castaldo and Reiher (Physical Review Research, 2026), the Protein Data Bank survey of cation-pi interactions by Gallivan and Dougherty (PNAS, 1999) and earlier Quentir coverage.

  1. "Molecular energies from sparse quantum prolate diagonalization on a trapped-ion quantum processor" on arXiv
  2. quantum prolate diagonalization paper by Stroschein, Castaldo and Reiher in Physical Review Research
  3. over one-fourth of all tryptophans experience an energetically significant cation-pi interaction
Next
Next

EigenOrbit and Trieste Dermatologists Use Quantum-Inspired Butterfly Layers to Certify an Alopecia Image Classifier, arXiv 2611.18775, 12 November 2026