---
slug: 2026-arxiv-2309.12129-leveraging-analog-quantum-computing-neutral-atoms
title: Leveraging Analog Quantum Computing with Neutral Atoms for Solvent Configuration
  Prediction in Drug Discovery
source: arXiv
link: https://arxiv.org/abs/2309.12129
category: app
triage_date: '2026-06-07'
triager: port_until_julien_parallel
bootstrap_date: '2026-06-07'
---

# Dossier: Leveraging Analog Quantum Computing with Neutral Atoms for Solvent Configuration Prediction in Drug Discovery

| Field        | Value |
|--------------|-------|
| Slug         | `2026-arxiv-2309.12129-leveraging-analog-quantum-computing-neutral-atoms` |
| Source       | arXiv |
| Link         | [https://arxiv.org/abs/2309.12129](https://arxiv.org/abs/2309.12129) |
| Category     | app |
| Triaged      | 2026-06-07 by port_until_julien_parallel |
| Bootstrapped | 2026-06-07 |

## Sourcing (Romain)

### Domain classification

`biology` (primary). This is structure-based drug discovery: predicting the
explicit configuration of water (solvent) molecules inside a protein binding
pocket, demonstrated on the MUP-I pocket complexed with
2-sec-butyl-4,5-dihydrothiazole (PDB 1i06).

Secondary: `optimisation`. The quantum kernel itself is not chemistry. The
water-placement task is recast as a constrained antiferromagnetic Ising model
(equivalently a Maximum Independent Set instance) and solved by analog evolution
of a Rydberg Hamiltonian. Solvation physics enters only through the classical
3D-RISM density that supplies the Ising one-body coefficients $\Gamma_i$, so a
quantum-chemistry reading is adjacent, but the portable primitive is
combinatorial optimisation on a neutral-atom array.

### Expert persona for Julien

A neutral-atom / Rydberg-array physicist who has run analog optimisation
(Maximum Independent Set, antiferromagnetic Ising) on Pasqal-class hardware.
They should be fluent in the ground-Rydberg encoding, the Rydberg blockade and
the $C_6/r^6$ van der Waals interaction scale, global versus local laser
addressing, and adiabatic versus variational (Bayesian-optimised) pulse
schedules. They should be able to read off the device noise budget that the
paper reports for the Fresnel QPU (array-spacing and $\Omega$-homogeneity
calibration, shot-to-shot $\Omega$ fluctuations, effective decay rate, and
false-positive/false-negative detection errors). A working acquaintance with
3D-RISM solvation modelling helps in judging how faithfully the Ising mapping
preserves the underlying chemistry, but the hardware fingerprint is the
deliverable.

### Related prior work (brief bibliography)

- Pichler, Wang, Zhou, Choi & Lukin, "Quantum Optimization for Maximum
  Independent Set Using Rydberg Atom Arrays" (2018), arXiv:1808.10816. Origin of
  the Rydberg-blockade MIS encoding on which this water-placement mapping rests.
- Ebadi et al., "Quantum Optimization of Maximum Independent Set using Rydberg
  Atom Arrays", Science 376, 1209 (2022), DOI 10.1126/science.abo6587,
  arXiv:2202.09372. The same primitive at 289-qubit scale on QuEra hardware; the
  natural cross-vendor reference point for the neutral-atom MIS kernel.
- Henriet et al., "Quantum computing with neutral atoms", Quantum 4, 327 (2020),
  arXiv:2006.12326, DOI 10.22331/q-2020-09-21-327. Pasqal neutral-atom platform
  review; the hardware context for Appendix A.
- Silvério et al., "Pulser: an open-source package for the design of pulse
  sequences in programmable neutral-atom arrays", Quantum 6, 629 (2022),
  arXiv:2104.15044. The emulation and pulse-control stack used throughout the
  paper.
- Sindhikara, Yoshida & Hirata, "Placevent: an algorithm for prediction of
  explicit solvent atom distribution", J. Comput. Chem. 33, 1536 (2012),
  DOI 10.1002/jcc.22984. The classical 3D-RISM water-placement baseline that the
  quantum method is positioned against (cited in-text as Placevent).
- Fusani, Wall, Palmer & Cortes, "Optimal water networks in protein cavities
  with GAsol and 3D-RISM", Bioinformatics 34, 1947 (2018),
  DOI 10.1093/bioinformatics/bty024. The genetic-algorithm classical baseline
  for the same task, cited as the global-optimisation method this work aims to
  improve upon.

### Difficulty estimate

`medium`. The full arXiv LaTeX source (main text, supplementary information, and
figures) is already in the folder, so there is no paywall or fetch risk, and
much of the hardware fingerprint is explicit. Appendix A gives the $^{87}$Rb
ground-Rydberg encoding ($|0\rangle = |5S_{1/2}, F=2, m_F=2\rangle$,
$|1\rangle = |60S_{1/2}, m_J=1/2\rangle$), the two-photon 420 nm / 1013 nm drive
via the $6P_{3/2}$ intermediate state, and the body text gives the noise budget
(around 1% array spacing, 4% $\Omega$ spatial homogeneity, 5% shot-to-shot
$\Omega$, $\Gamma_\text{eff}/2\pi = 0.05$ MHz, detection errors
$\varepsilon = 2\%$ and $\varepsilon' = 18\%$). What lifts it above `low`: the
paper fingerprints two algorithms (a local-addressing quantum adiabatic
evolution run, emulated with Pulser at 4 to 14 qubits and then partially
validated on the Fresnel QPU; and a global-Hamiltonian variational algorithm
with Bayesian optimisation that is emulation-only), the QPU portage substitutes
a constant global pulse for the full adiabatic schedule, and several
numerical-validation details sit in the SI. None of the primitives are exotic
for a neutral-atom specialist, so the extraction is moderate rather than high.