Researchers led by Stanford Medicine have built AI “digital twins” of the immune system in pregnancy and used them to pick out a drug combination that they say could delay preterm labour, according to a paper in Science Advances and a Stanford release published on 21 October.
The combination is aspirin and lansoprazole, a drug sometimes prescribed in pregnancy for acid reflux. The evidence so far comes from computer simulations, laboratory tests on donated blood and, according to Stanford, an animal experiment. No results in pregnant patients have been reported.
“We’re on the brink of having a real impact for the treatment of preterm birth, a global health problem that affects 10% of pregnancies worldwide,” said Brice Gaudilliere, the senior author and a professor of anesthesiology, perioperative and pain medicine at Stanford. Stanford’s release says prematurity is the leading cause of death in children under five worldwide, and that it has been decades since a new drug for preventing preterm birth emerged.
Stanford Says the FDA Has Approved a Human Trial
Stanford’s release gives this account of the step after the modelling: “The researchers tested the drug combination of aspirin plus lansoprazole in an animal model of an inflammatory state in pregnancy, demonstrating that the combination delays the onset of labor in that context. They subsequently obtained approval from the U.S. Food and Drug Administration to try the combination in a human clinical trial.”
The release says the researchers “are starting a clinical trial in real patients”. It gives no trial size, dose or start date. The animal experiment and the FDA approval are described only in the release; the paper itself does not report them.
The Model Was Built From 218 Million Immune Cells
The system is called SIMCO, short for “Simulated Immunome Modeling of Clinical Outcomes”. According to the paper, the team took blood from 8 non-pregnant and 11 pregnant volunteers and exposed the immune cells in the laboratory to nine drug regimens. The resulting atlas covers 218 million white blood cells. Gaudilliere described it in the release as “a quarter of a billion cells”.
The drugs were chosen through a drug-repurposing pipeline developed by Marina Sirota’s group at the University of California, San Francisco, because, Stanford says, they “were already known to be safe for pregnancy”. They included metformin, pravastatin, a form of folate, and salicylic acid, which the release calls aspirin, both alone and with lansoprazole.
The researchers then applied the model to a separate group of 63 pregnant patients, with 177 blood samples between them. The paper reports that it predicted time to labour with a correlation of 0.64 and an average error of 24.1 days. It separated term from preterm births with an AUROC of 0.75, though only five of the 63 pregnancies ended preterm.
In the Simulations, the Pair Delayed Labour by a Median of 17 Days
Each patient’s digital twin was run with and without each drug. For salicylic acid plus lansoprazole, the model predicted a median extension of pregnancy of 17.3 days, with 87.3% of patients predicted to benefit. Salicylic acid alone was predicted to help 46%. “In the digital twin system, these two drugs in combination had a large effect on the delay of labor, 15 times higher than aspirin alone,” Gaudilliere said in the release. That multiple is his; it is not a figure given in the paper.
Folate scored highest in the simulations, at a median of 24.7 days, and maprotiline came next. The paper says maprotiline’s side effects make it a less attractive candidate. “We’re giving folic acid for spina bifida prevention, but maybe, unknowingly, we’re also helping the maternal immune system adapt to pregnancy,” said Jakob Einhaus, a co-lead author.
The authors write that “the immunomodulatory profile of this combination provides a rationale for its clinical evaluation.”
Simulated Trials Could Shrink Real Ones, but SIMCO Is a Ranking Tool
The main benefit the researchers claim is ethical as well as practical. Testing drugs in pregnancy is hard, and Stanford says SIMCO lets researchers narrow the options “before exposing any patients”. Because the twins capture differences between individuals, the team also says the model could pick out likely responders, so that future trials can be smaller and give clearer answers. TSN has also covered Google Research’s AI agents running wet-lab experiments and a plan by Oracle and AstraZeneca to use AI to find clinical-trial patients.
The paper is plain about what the model cannot do. “SIMCO is best suited for comparative and mechanistic inference rather than precise quantitative prediction,” the authors write. The 17-day figure is a simulated, single-dose estimate from blood exposed in a dish, not a measured result. The validation group contained only five preterm births. Samples came from the second trimester and from blood alone, and the measurements track cell signalling rather than what cells go on to do. The authors call for validation in larger groups, animal studies and prospective trials that “will confirm the safety profile and exclude any increased risk of infection-driven PTL”, meaning preterm labour.
Several authors have commercial links. The paper says Brice Gaudilliere and co-authors D. K. Gaudilliere and Hedou sit on the advisory board of SurgeCare, where Hedou is chief executive, and that five authors are named inventors on a Stanford patent application filed in February 2026. As with Microsoft’s diagnostic AI results in Science, strong model results still have to be tested on patients.
No One Has Yet Taken the Combination in a Trial
Neither source reports anyone receiving aspirin plus lansoprazole to prevent preterm labour in a study. The paper ranks drugs by predicted effect; it does not show that any of them work in pregnancy. Stanford says the next step is a human trial cleared by the FDA, and that is where the 17-day prediction will be tested.
Sources
- Einhaus, J., Neidlinger, P., Fondeur, O., Sato, M. et al., “Single-cell–level digital twins for preterm birth prevention strategies”, Science Advances, published online 21 October 2026: https://www.science.org/doi/10.1126/sciadv.aed2741
- Stanford Medicine, “‘Digital twins’ of immune system in pregnancy identify drugs to prevent early labor, Stanford Medicine-led study finds”, 21 October 2026: https://med.stanford.edu/news/all-news/2026/10/preterm-labor-drug.html
Related stories
- Microsoft Diagnostic AI Scored 83% on Past Cases, Science Paper Says
- Google Research Puts AI Agents on Wet-Lab Robots for Closed-Loop Design
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X
