INmune Bio Becomes Quantum Computing Inc.'s First Pharma Customer on 14 October 2026: What Dirac-3S Will Search for in the XPro Alzheimer's Trial Data
Quentir Medicine Monitor
Evidence-based insights for quantum medicine. Published by Quentir Systems LLC · October 15, 2026.
On 14 October 2026 the Alzheimer's drug developer INmune Bio and the photonics company Quantum Computing Inc. (QCi) announced that INmune Bio will use QCi's Dirac-3S quantum optimization machine to look for new diseases in which its anti-inflammatory candidate drug might work. The engagement is a first stage with options for more, and QCi describes it as its first commercial customer in the pharmaceutical and biotechnology market.
The drug is XPro1595, which INmune Bio calls XPro. It is designed to neutralize the soluble form of tumor necrosis factor, a signaling protein that drives inflammation, while leaving the membrane-bound form, which the immune system and nerve cells rely on, largely untouched. INmune Bio has tested it in early Alzheimer's disease in patients whose blood shows signs of inflammation.
According to the joint announcement issued from Boca Raton and Hoboken, the two companies will combine INmune Bio's data from preclinical work and four clinical trials with public biomedical datasets. The aim is to find new indications, meaning diseases driven by the same inflammatory biology XPro targets, and to rank them by their chance of success. INmune Bio's chief executive David Moss said the search would start with rare diseases, where trials can be smaller and faster, and where a qualifying pediatric approval may earn a priority review voucher from the US Food and Drug Administration.
The announcement contains one commitment that deserves more attention than the headline. The companies say they will compare the quantum machine's output with conventional statistical and machine-learning methods and judge which approach suits which part of the analysis. Whether that comparison is published will decide how much the collaboration teaches anyone outside the two companies.
What the MINDFuL phase 2 trial of XPro1595 found in 200 patients with early Alzheimer's disease
The largest body of patient data INmune Bio brings to the work comes from MINDFuL, a randomized, double-blind, placebo-controlled phase 2 trial registered as NCT05318976 on ClinicalTrials.gov. The registry lists 207 enrolled participants, a start in February 2022 and completion on 12 May 2025. Patients aged 50 to 85 with mild cognitive impairment or mild dementia and raised inflammatory biomarkers received a weekly injection under the skin of XPro1595 at 1.0 mg/kg or a placebo for 24 weeks, at 35 centers in eight countries.
The results, written up by Judith Jaeger, Christopher Barnum and colleagues in a preprint posted to medRxiv on 27 September 2025, are mixed. In the 200 patients of the modified intent-to-treat population, the drug did not beat placebo on the primary endpoint, a composite of six cognitive tests called the Early and Mild Alzheimer's Cognitive Composite (EMACC), with p = 0.672. The authors write plainly that XPro1595 did not meet the primary endpoint in the overall population.
The company's case rests on a second group that the preprint authors describe as prospectively defined: 100 patients who were amyloid-positive and carried two or more inflammatory markers. In that group the effects favored the drug, with a standardized effect size (Cohen's d) of 0.27 on the cognitive composite, 0.23 on a neuropsychiatric scale and 0.37 on its agitation subscale, with matching small shifts in the blood markers pTau217 and GFAP. The preprint reports no amyloid-related imaging abnormalities, the brain swelling and bleeding seen with antibody drugs that clear amyloid, although about 70 percent of participants carried the APOE4 risk gene. In its company description INmune Bio states that XPro now holds FDA Fast Track designation and that an End-of-Phase 2 meeting supports a combined phase 2b/3 program in early Alzheimer's disease with inflammation.
Quantum pillar: computing. Technology readiness: TRL 2 of 9. For this proposed clinical use of quantum optimization, the companies have agreed a first-stage engagement and described what they intend to analyze; the announcement reports no results from quantum analysis of the XPro data and no patient group or new indication chosen on the strength of it.
How QCi describes its Dirac-3S machine and the kind of problem it accepts
QCi builds what it calls entropy quantum computing. In its product description of Dirac-3, the company says the machine maps an objective function onto the evolution of a physical photonic system and uses loss and noise as a resource, runs at room temperature and fits a standard server rack. It accepts an optimization problem written as an energy function and searches for the lowest-energy configuration. In its base configuration QCi lists capacity for 949 variables in problems with pairwise interactions, falling to 19 variables when five variables interact at once; an expansion module raises those figures to 9,980 and 55.
QCi claims a polynomial advantage for approximate solutions and strong results on a standard benchmark suite. Those are the company's statements about its hardware. The machine solves optimization problems; it does not read clinical data by itself. Someone has to turn the question "which patients respond, and which diseases resemble them" into an energy function first, and much of the scientific judgment in this project will sit in that translation.
Optimization of this kind has a natural place in biomarker work. Choosing a small panel of markers out of hundreds, so that the panel separates likely responders from non-responders, is a combinatorial selection problem. The same holds for grouping diseases by overlapping pathway signatures. Classical solvers handle these problems routinely, which is why the promised head-to-head comparison matters.
Why a few hundred trial patients make indication search a statistics problem first
The XPro data are rich in measurements and modest in patients. MINDFuL's enriched group holds 100 people, split between drug and placebo. Any method that searches many combinations of markers and subgroups in a set that size will find patterns, and some of those patterns will be chance. This applies equally to a quantum optimizer, a gradient-boosted model and a careful statistician with a spreadsheet. The safeguard is the same in every case: fix the analysis in advance, hold out data the search never sees, and confirm promising signals in a new cohort.
The Monitor made a related point when reading a quantum attention model tested on Merck's Alzheimer's candidate verubecestat: a quantum contribution means most when a classical method of the same size has had the same chance on the same data. The INmune Bio and QCi plan states that intent. The public will learn whether it was carried out only if the comparison, including the cases where conventional methods did as well or better, is published.
There is also a difference between the two kinds of output the companies describe. A list of candidate rare diseases is a hypothesis that a later trial must test. A biomarker profile that predicts response is closer to the clinic, because it could decide who is enrolled in the phase 2b/3 program. The second use carries higher stakes and needs stronger validation, ideally on data from patients outside MINDFuL.
What clinicians and hospital buyers should look for once Stage 1 reports
For memory clinics and neurology departments, the near-term relevance of XPro lies in its trial program, which would enrich for neuroinflammation using blood biomarkers. The quantum collaboration may influence how that enrichment is defined, and it may point the drug toward rare inflammatory diseases outside neurology. It does not change anything in current care: XPro1595 is an investigational drug, and the MINDFuL subgroup signals are small effects in a trial that missed its main endpoint.
Three questions will show whether Stage 1 produced something useful. Did the companies publish which methods were compared, on which data, and with what held-out validation? Did the quantum optimizer find a marker panel or a disease grouping that conventional methods missed, or did it reach the same answer faster or at lower cost? And did any candidate indication move into a registered study? Until those answers exist, the announcement establishes a commercial relationship and a research plan, with the scientific result still to come.
Sources
Primary source: INmune Bio Inc. and Quantum Computing Inc., joint announcement of a quantum optimization and AI collaboration for XPro, PR Newswire, Boca Raton and Hoboken, 14 October 2026, quoting chief executives David Moss and Yuping Huang. Also drawn on: the MINDFuL phase 2 preprint by Judith Jaeger, Christopher Barnum and colleagues, the trial's ClinicalTrials.gov entry, QCi's Dirac-3 product description, and the Monitor's earlier coverage of a quantum attention model for verubecestat. The readiness rating is the Monitor's editorial assessment.