Researchers at Utrecht University have built a computer simulation of an early embryo to study how everyday chemicals, such as pesticides, might disrupt the formation of the spinal cord, and to estimate the risk of spina bifida. Utrecht University announced the work on 13 October 2026, and says the model is promising but not yet ready to stand on its own.

What the model does

According to the university, Job Berkhout developed the model for his PhD thesis. It represents embryonic cells that move, divide and change shape. Berkhout also studied in the lab how environmental chemicals interfere with the instructions inside cells, and fed that knowledge into the simulation.

He then tested the model on neural tube closure, the early step in which the tissue that becomes the brain and spinal cord folds shut. Spina bifida is a defect that arises when this closure goes wrong. In the simulation, the neural tube closed properly when no harmful substances were present, and closure was disrupted when chemicals were introduced.

Because cell behaviour is partly random, the model does not give a yes-or-no answer. It gives the probability that a defect occurs, the university says.

Berkhout also built an AI model that compares the structure of a substance with chemicals whose effects are already known, producing a ranking of substances that are likely to be harmful. The university quotes the researcher as saying the AI "does not have the final say, but it gives us a starting point for our search."

A related published paper

The university's announcement does not cite a journal article. A paper in the open-access journal Current Research in Toxicology, published in December 2024 with Berkhout as first author, describes a computational model of neural tube closure built on the same idea. Its authors are affiliated with the Dutch national institute for public health and the environment (RIVM), Utrecht University and US institutions. It is not stated in the university's release that this paper and the thesis are identical, so readers should treat it as related background.

The paper reports that the model reproduced several types of spina bifida and midline fusion defects when particular genes were disrupted in the simulation. It also lists its own limits: the model is two-dimensional and covers only a slice of the mid-spinal region, many parameters are standardised rather than fitted to measurements, and some predictions have not been checked experimentally. The work was carried out within the EU-funded ONTOX project (Horizon 2020).

Why it matters, and what is still missing

The stated hope is that models like this can help identify substances that are harmful to an embryo and eventually reduce reliance on animal testing. The university is clear that this is not yet possible: "We first need to test the model further," Berkhout says, before it can assess risk independently.

The announcement does not name the chemicals tested or say how closely the simulated probabilities match real-world incidence of spina bifida, so the model's predictive accuracy for people remains to be shown.

Prize nomination

Berkhout is nominated for the Hugo van Poelgeest Prize, which the university describes as the Netherlands' oldest award for animal-free scientific research. An expert panel chooses the main winner, and the public can vote for an audience award until 1 November 2026. The winner will be announced on 13 November 2026 at the event "The Future is Human(e)".

After his thesis, the university says, Berkhout continues as a postdoctoral researcher at Vrije Universiteit Amsterdam, where he is building a similar model of the synapse as part of the international SynGO consortium.