Researchers from Stanford Medicine have created an AI technology to create virtual immune systems for pregnant patients and discover promising drug options for reducing preterm births. The technology has discovered that a drug combination of aspirin plus lansoprazole may be a promising option to reduce early labor.
In the study published online in Science Advances on October 21, 2026, a technique called digital twin modelling was used to create individual immune system models to see how they would react to different drug treatments. These results could provide a new avenue for preterm birth research.
The drug combination was predicted to delay labor significantly more than aspirin alone in the digital twin models. However, this does not guarantee that this combination would prevent preterm births in humans.
Birth before 37 weeks gestation (preterm birth) can cause a variety of complications such as breathing issues, development issues and an increased risk of neonatal death. Preterm birth is currently the leading cause of death in children under five globally.
While the mechanisms behind labor aren’t fully understood, it is known that changes occur in the mothers immune system. During pregnancy, the immune system adapts to help support the fetus. As pregnancy comes to an end, inflammatory signals increase. These can cause labor to begin early.
The Stanford team created a system called SIMCO (Simulated Immune Modeling of Clinical Outcomes). This uses machine learning techniques and large volumes of cellular data to predict how immune cells will react to drugs.
Unlike a digital replica of an entire pregnancy, SIMCO models the immune system at the single-cell level and uses immune-response data to predict treatment effects on the timing of labor.
Immune cells were exposed to different drugs in the lab using samples taken from blood. This provided the researchers with data on how each immune cell reacted to different drugs. The data set included hundreds of millions of immune cells.
The team then added data from 63 pregnancies to help the model identify patterns in immune responses and how they correlate to pregnancy outcomes.
This enabled the researchers to test different treatment and non-treatment options for the individual digital twins.
The team used SIMCO to test nine potential drug options.
A combination of aspirin and lansoprazole appeared to be a promising option. Aspirin has previously been studied as it could reduce the risk of preterm birth in some patients. Lansoprazole is a proton pump inhibitor often prescribed for acid reflux.
The combination was predicted to delay labor much more than aspirin alone. The team at Stanford reported that the combination was predicted to reduce labor by around 15 times more than just aspirin.
They suggested that the combination could have a synergistic effect by affecting immune pathways that help keep pregnancy going and pathways that cause labor.
The study also identified tetrahydrofolate (THF), a form of folate, as a promising candidate for delaying labor in the model. The researchers tested THF in their laboratory experiments rather than directly establishing that routine folic acid supplementation delays labor. Folic acid is already recommended during pregnancy to reduce the risk of neural tube defects, but this study does not establish an additional clinical benefit for preventing preterm birth.
This provides new opportunities to understand how drugs and supplements currently available could affect pregnancy.
The benefit of using digital twins is that drugs can be trialed in silico before being tested in pregnant patients.
There can be ethical concerns when trying to trial new drugs in pregnancy. By identifying promising drugs using biological modelling, researchers can prioritize treatments.
The team also tested the combination in an animal model of inflammatory pregnancy. They found that the combination delayed labor.
They have gained approval from the US Food and Drug Administration to trial the combination in humans.
However, gaining approval from the FDA to trial the combination in humans does not mean that this combination should be used to reduce preterm birth. We need to know how effective this combination is, what the correct doses are, and whether it is safe for pregnant patients.
The digital twin technology may also help identify which patients are more likely to benefit from a particular treatment. If shown to be effective, this could help design more targeted trials.
This study shows how useful AI can be for speeding up drug repurposing. Researchers no longer need to rely solely on traditional drug discovery methods but can use biological data-driven models to find new uses for drugs. However, the predictions from this study cannot be guaranteed to have clinical benefits. The model was based on limited clinical data and there were few preterm births in the validation cohort. Larger studies will be required to understand how accurate the predictions are.
Further research will be needed to understand if the combination can safely reduce preterm births, if there are differences in how it works between different patient groups and if the benefits outweigh the risks.
Overall, this study shows an early but promising use of AI-driven precision medicine. Using detailed immune data and modelling this could help researchers find and prioritize new ways to reduce premature births.
Ultimately, while AI-powered digital twins have identified that a combination of aspirin and lansoprazole may help reduce labor, it is still an investigative technique. Patients should not start, stop or combine these drugs during pregnancy to reduce preterm birth without speaking to their clinician.
References:
Digitale, Erin. “'Digital Twins' of Immune System in Pregnancy Identify Drugs to Prevent Early Labor, Stanford Medicine-Led Study Finds.” Stanford Medicine News Center, October 21, 2026. https://med.stanford.edu/news/all-news/2026/10/preterm-labor-drug.html.
Einhaus, Jakob, et al. “Single-Cell-Level Digital Twins for Preterm Birth Prevention Strategies.” Science Advances, published online October 21, 2026. https://doi.org/10.1126/sciadv.aed2741.
Rh/MF/MSM