Five researchers at two Italian universities found that 36 of 53 peer-reviewed experimental studies applying quantum computing to bioinformatics and biomedical data never ran on a quantum computer. They reported the finding in a systematic review published in Frontiers in High Performance Computing on September 24, 2026.
The researchers behind 40 of the studies compared a quantum or quantum-inspired method with at least one classical, or conventional, method. In all 40, they reported performance that matched or exceeded the classical baseline.
The review’s authors also wrote that “none of the considered articles explicitly demonstrates a quantum advantage over classical approaches.” A quantum advantage would mean a quantum computer outperforming the best available classical methods on a task.
Of the 53 studies, 36 ran as simulations or quantum-inspired programs on ordinary computers, 41 used toy or substantially reduced datasets, and none explicitly demonstrated a quantum advantage, according to the authors.
Authors and method
The review’s corresponding authors are Paolo Zuliani of the Department of Computer Science at Sapienza University of Rome and Tiziana Castrignanò of the Department of Ecological and Biological Sciences at the University of Tuscia in Viterbo. Manuel Arcieri of Sapienza is the first author, with Giovanni Carotenuto and Meryam Carrus of Tuscia as co-authors.
The team started from 1,550 records in the Scopus database and dropped preprints and papers from venues without peer review. It also left out papers published before 2014 and purely theoretical proposals. Papers with no Scopus citations were excluded unless they appeared after January 2025.
The authors found no further qualifying studies when they searched four other databases, including PubMed and Web of Science. The authors described the studies they kept as spanning genome assembly, sequence alignment, protein folding, molecular simulation, drug discovery, phylogenetics, and machine learning on clinical and multi-omics data, the combined measurements of genes, proteins, and other molecules.
Hardware, data, and code
Twenty-five studies, or 47%, ran as classical simulations of quantum circuits, the step-by-step programs that quantum computers execute. In another 11, researchers used quantum-inspired algorithms, classical programs that borrow ideas from quantum computing and run on conventional computers.
Eleven studies ran on quantum annealers, special-purpose machines for optimization problems, all of them made by D-Wave. Six ran on the older D-Wave 2X or 2000Q and five on the first-generation Advantage.
The remaining six studies were run on gate-based quantum processors, which execute programs built from basic operations called gates. Two each came from IBM and IonQ, and one each from Rigetti and QuantumCTek.
The authors classified 41 studies, or 77%, as tested only on toy or substantially reduced datasets and 12 as using data close to real-life scale. Of the 17 studies that ran on quantum hardware, 15 were tested on reduced datasets.
Source code was available for 18 studies and data for 37. Sixteen studies, or 30%, had both, the combination the authors called “the minimum requirement for full computational reproducibility.” Fourteen had neither.
Obstacles and recommendations
The authors wrote that access to quantum processors is “restricted and expensive.” Calibration parameters, the settings used to tune each processor, “change continuously, often over hours to days,” so repeated runs of one circuit can produce different outcomes, they wrote. They added that the D-Wave 2X and 2000Q systems used in several of the reviewed studies “are no longer commercially available, making literal reproduction of those experiments impossible.”
The authors recommended standardized benchmark datasets and evaluation protocols, as well as dedicated platforms for negative results. They asked researchers to deposit code and hardware configuration parameters in public repositories along with their experimental data.
They also called for measurements of how methods scale with problem size and for software libraries built for bioinformatics. Researchers should state clearly whether they obtained their results on quantum hardware, with circuit simulations, or with quantum-inspired heuristics, they wrote.
Conditions for practical advantage
The authors wrote that practical quantum advantage in bioinformatics “will require several milestones to be met simultaneously.” The first is enough high-fidelity logical qubits, error-protected units built from many physical qubits, to encode biologically relevant problems.
The second is logical error rates low enough for the circuit depth involved, the number of successive operations in a computation. The third is connectivity overheads, the extra operations needed when qubits cannot interact directly, small enough that the theoretical speedup remains.
The fourth is end-to-end runtimes shorter than those of optimized classical or high-performance computing baselines at comparable accuracy. Fault-tolerant computing, in which errors are corrected faster than they accumulate, “is widely believed to be necessary for demonstrating quantum advantage in practically relevant applications,” the authors wrote, citing a 2025 paper by the physicist John Preskill.
They described the current state of the field as “credible prototypes.” The authors also cited a 2026 systematic review by Khalaj, Rayan, and Jin in Briefings in Bioinformatics as related recent work.
