Fertility sensations: AI will identify substances in the environment that are dangerous for childbirth

Russian scientists have developed a way to identify thousands of chemicals that could potentially imperceptibly affect reproductive health. Artificial intelligence will help identify dangerous compounds even before laboratory tests. Experts have created a system that predicts whether chemicals can disrupt the work of the progesterone receptor, a hormone important for ovulation and pregnancy. The development will make it possible to quickly select substances that require verification and, in the future, strengthen control over chemical safety. In addition, it can be used in the treatment of infertility, experts said.

AI for detecting substances dangerous for the reproductive system

Scientists at Sechenov University have developed a machine learning-based system for the preliminary assessment of chemicals potentially affecting the reproductive system. It predicts whether the compound can bind to the progesterone receptor, activate or block it, and have a general toxic effect on cells.

The traditional toxicological and hygienic assessment of chemicals, including the determination of reproductive toxicity, that is, the ability to disrupt sexual function, fertility, or offspring development, requires significant resources and the use of laboratory animals. At the same time, the number of chemical compounds is so large that a detailed experimental study of each of them is impossible.

One of the important targets of such substances is considered to be the progesterone receptor. Some xenobiotics are able to bind to it, mimicking the action of the hormone or blocking its activity.

— The progesterone receptor plays an important role in the regulation of reproductive processes, however, computational toxicology tools are still much less than for estrogen and androgen receptors. It was important for us to create an approach that would allow us to prioritize chemicals according to the degree of risk of negative effects on the reproductive system and effectively plan further research," said the head of the study, PhD, Associate Professor of the Department of Human Ecology and Environmental Hygiene at the F.F. Institute of Public Health. Erisman Ekaterina Guseva.

Placenta rate: scientists have obtained important data on cancer treatment in pregnant women

How can the study of the characteristics of such patients help to choose an individual therapy

For preventive toxicology, this means the possibility of a preventive assessment of the safety of new or insufficiently studied compounds. It is especially important in this way to check industrial chemicals, plant protection products, as well as other xenobiotics that are massively present in the environment and require operational control over their content, the specialist said.

To create the system, the researchers used the Tox21 database and information on more than 7,000 compounds with known experimental results. The system evaluates whether the new substance interacts with the progesterone receptor, what the effect may be, and whether it is due to a general toxic effect on cells.

The models recognized compounds capable of binding to the progesterone receptor best. The system also takes into account how similar the new substance is to the ones it was trained on: if the differences are too great, the prediction is considered less reliable and requires experimental confirmation.

— There are a lot of chemical compounds that a person comes into contact with, and it is impossible to investigate each one experimentally in detail. We are developing a tool to help determine which connections are worth checking first. It is important for us that the researcher receives not only a forecast, but also an assessment of how reliable it is for a particular substance," said Ivan Nemchenko, one of the authors of the development, a graduate of Sechenov University and an intern at the FAST Center for Digital Strategic Projects.

He received the programming and data analysis skills necessary to work on the project at the joint campus of Sechenov University and School 21, the Sbera School of Digital Technologies. After the main track, he mastered the field of "Artificial Intelligence in medicine" and applied the knowledge gained in creating a system of toxicological screening.

Prospects for early detection of reproductive toxicity

The development is still a research prototype and does not replace laboratory tests. Its task is to narrow down the range of substances that need to be tested first.

— The next stage is to conduct an extended validation of the model on independent samples of substances and integrate it with existing computational tools for a comprehensive assessment of the effects on the system of steroid sex hormones. This will allow us to assess the prospects for the implementation of the developed approach in the toxicological and hygienic assessment of chemical safety," said Ekaterina Guseva.

Photo: IZVESTIA/Sergey Lantyukhov

Some chemical compounds can affect the progesterone receptor: mimic the action of the hormone or block it. As a result, ovulation, embryo implantation and pregnancy preservation are disrupted, said molecular biologist Arina Kholkina. Known teratogens include, for example, retinoids and thalidomide, substances that can cause fetal malformations.

— The problem is that there is no single regulated framework where you can quickly check any substance for such a risk. There are thousands of compounds that a person comes into contact with at home and at work, and it is impossible to test each of them experimentally. The Sechenov University machine learning system solves this problem: it is trained on thousands of compounds with known results and predicts whether a new substance is capable of binding to, activating or blocking the progesterone receptor, as well as whether the molecule has a general toxicity to cells, she explained.

Enhance reproduction: virtual diagnostics will find the cause of infertility without surgery

3D reconstruction allows you to diagnose dangerous diseases without risky interventions and anesthesia.

In fact, it is a filter that selects from thousands of compounds those that should first be tested in the laboratory. This approach will help protect pregnant women and those planning to have children from the potentially dangerous effects of chemicals in a timely manner, the expert concluded.

Photo: IZVESTIA/Polina Violet

Similar systems can be useful in the treatment of infertility. If an analysis of the microRNA profile in the blood makes it possible to determine which receptors are suppressed at the epigenetic level, the doctor will be able to quickly understand why standard therapy does not work. Olga Yaroshenko, an obstetrician-gynecologist and consultant at the Rational Center for Genetic Research MyGenetics, told Izvestia about this. According to her, this approach can change the tactics of infertility treatment: it will allow to identify progesterone resistance and take it into account when choosing therapy.

"If the system detects a high marker of toxic damage, the couple is sent to a pre—detox stage of preparation before the start of expensive ART (assisted reproductive technology) cycles," the doctor said.