Few laymen know what the word “teratology” means, but for the unborn fetus and its parents, it involves the difference between normality and disability, life and death. It is the science that studies the causes, mechanisms and patterns of abnormal development in animals and humans, as well as plants. Certain congenital developmental disorders (CDDs) may be caused by various mechanical effects, fetal diseases and arrested development of the embryo and fetus. 

The term was borrowed in the mid-19th century from French tératologie, which itself was taken for the Greek words “sign sent by the gods, portent, marvel and monster” and -ologie, used to designate a discourse, treaty, science, theory or study of some topic. 

Teratology as a modern science was born in the 1930s with the publication of a set of experiments in which pregnant pigs were fed a diet deficient in vitamin A. All of these piglets suffered a variety of malformations, especially a lack of eyes.

With the growth of understanding of the origins of birth defects, the field of teratology began to overlap with other fields of science including developmental biology, embryology and genetics. Until the 1940s, teratologists regarded birth defects as primarily hereditary, but them, the first well-documented cases of environmental agents being the cause of severe birth defects were reported. 

Known teratogens in human fetuses include thalidomide, retinol, mercury, alcohol, lead and polychlorinated biphenyls (PCBs), but there are many more and numerous more substances whose detrimental effects  have not been proven. Fetuses whose mothers consumed significant amounts of alcohol during pregnancy, for example, can develop craniofacial malformations; this is the leading cause of birth defects and neurodevelopmental abnormalities in the US. 

Understanding how a teratogen causes its effect is not only important in preventing congenital abnormalities; it also has the potential for developing new therapeutic drugs safe for use with pregnant women.

In humans, vaccination has become readily available and is important to the prevention of some diseases like polio, rubella, smallpox and COVID-19, among others. There has been no association between congenital malformations and vaccination against influenza and COVID-19 vaccines, for example.

Thalidomide, which was approved by the US Food and Drug Administration in the late 1950s as a sedative, was also used to treat nausea and vomiting during pregnancy. Its use during pregnancy caused irreversible damage to the fetus, and nearly 10,000 children were born with severe congenital malformations

Worldwide. The thalidomide crisis was followed by the establishment of stringent guidelines for drugs targeted at pregnant women, and these enabled the relatively safe use of drugs during pregnancy, and medication use during pregnancy is common. However, it also prevented pregnant women from benefiting from drugs they needed or their own health.

Now, collaboration between information systems engineering researchers from Ben-Gurion University (BGU) of the Negev in Beersheba and the clinical pharmacology unit at Shamir Medical Center (Assaf Harofeh) in Tzrifin in the Dan Region has yielded a method for predicting drug safety, using artificial intelligence (AI). Based on this method, an English-language website has been set up that is accessible to the general public and allows for information on about 10,000 drugs. It is located at https://icc.ise.bgu.ac.il/medical_ai/drug_preg/full. 

Drug safety concerns everyone who has ever taken medication, and this is especially relevant to pregnant women who also need to guard the health of the fetus in the womb. Today, only a small proportion of drugs have an unambiguous classification regarding the safety of their use during the months of pregnancy. 

The BGU research group, which included student Guy Shtar, Prof. Bracha Shapira and Prof. Lior Rokach from the university’s department of software and information systems, examined the possibility of offering pharmacological and medical information through AI prediction of drug-drug interactions. 

Based on its research, the research group developed a first-of-its-kind model that allows pregnant women and their doctors to learn about the safety of taking medications during pregnancy.

The information presented on the site was compiled using algorithm-based technology that helps predict the safety of taking the medication while weighing the data collected during the study. To train the proposed model, we created a set of labeled drugs by processing over 100, 000 textual responses collected by a large teratology information service. Structured textual information is incorporated

into the model by applying clustering analysis to textual features.

The model is accessible to the public and provides detailed explanations of its decisions and determinations o help physicians and patients understand the benefits and dangers of taking medication during pregnancy

For a significant number of drugs, the safety of pregnancy is not known with high certainty because one can’t use woman like lab animals, and in most cases, there is partial information collected on animals only. Very few of the drugs have been tested in a controlled manner in pregnant women in the past, and even most of the drugs do not have an unambiguous classification as to the safety of taking them, said Shtar. “The information was collected from about 100,000 inquiries received at the pharmacology unit at Shamir Medical Center over a period of a decade including side effects, medical uses and taking several drugs at the same time. 

The Health Ministry in Jerusalem has a number of pharmacological counseling centers for pregnant and lactating women, and they make the information available to the public for each drug. But in many cases, there are not enough data, the recommendation of professionals for these drugs is sometimes to avoid the drug only because it lacks information. “Through in-depth data analysis, we have developed a model that may help many,” Shtar continued. 

The findings of the study were published under the title “Explainable multimodal machine learning model for classifying pregnancy drug safety” in the prestigious journal Bioinformatics, which is the official journal of the International Society for Computational Biology and Bioinformatics.