AI model predicts risk of birth defects from new drugs

Researchers in the US have developed an artificial intelligence (AI) algorithm that can be used to predict whether a drug has the potential to harm developing foetuses in the womb.

The team from the Icahn School of Medicine at Mount Sinai in New York say the “knowledge graph” tool can also identify existing medicines that are not currently classified as harmful, but may in fact cause congenital disabilities.

Birth defects are seen in around one in every 33 births in the US and, for most, there is no known cause, according to the scientists, who have published their work in the Nature journal Communications Medicine.

Possible causes for birth defects include genetic mutations, as well as environmental factors such as drugs, cosmetics, food, and pollutant substances that women may be exposed to while pregnant.

“We wanted to improve our understanding of reproductive health and foetal development and, importantly, warn about the potential of new drugs to cause birth defects before these drugs are widely marketed and distributed,” said Avi Ma’ayan, director of the Mount Sinai Center for Bioinformatics and lead author of the paper.

The researchers trained their ReproTox-KG AI using data from the literature on genetic associations, drug-induced gene expression changes in cell lines, known drug targets, genetic burden scores for human genes, and the ability of small molecule drugs to cross the placenta.

It uses semi-supervised learning (SSL), a branch of machine learning that uses a small amount of labelled data to guide predictions for much larger unlabelled data.

The AI identified more than 30,000 preclinical small molecules that may cause birth defects, as well as over 500 different molecular mechanisms or ‘cliques’ that connect birth defects, genes, and drugs.

“Although identifying the underlying causes is a complicated task, we offer hope that through complex data analysis like this […] we will be able, in some cases, to better predict, regulate, and protect against the significant harm that congenital disabilities could cause,” said Ma’ayan.

There have been a number of instances in which a drug’s tendency to cause congenital abnormalities isn’t recognised during the clinical testing phase and only emerges after it has been placed on the market and used by thousands of patients.

The outbreak of limb malformations caused by thalidomide in the late 1950s and early 1960s is the most prominent example, but there have been others in recent decades. More recently, for example, Sanofi’s epilepsy therapy Depakine (sodium valproate) was linked to neurodevelopment delay (NDD) issues, such as spina bifida, in children who were exposed to the drug in the womb.

Photo by freestocks on Unsplash.