Researchers decode the chemistry behind a deadly genetic disease

Northeastern University researchers used a novel machine learning tool to predict how genetic mutations cause a rare metabolic disease known as OTC deficiency, uncovering some of the underlying biochemical mechanisms at play and laying the groundwork for future treatments.

Ornithine transcarbamylase (OTC) deficiency is a genetic disorder that impairs the body’s ability to safely eliminate ammonia, a byproduct of normal protein recycling that occurs in cells. Ammonia buildup can be toxic and cause serious harm, including brain or liver damage, or even death.

A research team led by Northeast chemistry and chemical biology professors Mary Jo Ondrechen and Penny Beuning combined its original machine learning tool, called Partial Order Optimum Likelihood, or POOL, with laboratory biochemical experiments to study dozens of OTC gene mutations. This gene produces the OTC enzyme, a protein that speeds up chemical reactions, which is part of the cycle that converts nitrogen into urea so the body can excrete it in urine.

“Professor Ondrechen’s machine learning method is extremely effective in predicting the effects of mutations on the function of a protein,” explains Beuning. “This is the second time we have used this method to analyze hundreds of mutations in a disease-associated enzyme, and the experimental analysis in both cases showed that the predictions were accurate.”

The study, published in ACS Chemistry Biologysheds light on how certain mutations disrupt the normal activity of the enzyme. This deeper understanding, at the molecular level, of disease mechanisms is an important step towards the development of personalized treatments in the future.

Understanding the OTC deficit and its impact

“Right now, we’re basically asking ourselves, ‘Why is this mutation bad?’ Why does it cause disease?’” Ondrechen says. “And then you could try to think of a drug, a small molecule that would bind to the protein and counteract the effect of the mutation.”

Each year, between 14,000 and 77,000 people are diagnosed with OTC deficiency. A severe form of the disease affects some newborns, usually boys, shortly after birth. A milder form of the disorder may appear later in childhood or adulthood.

“Here in Massachusetts, every baby is tested for various hereditary mutations, including over-the-counter deficiency,” Ondrechen says.

Symptoms vary in severity but can include vomiting, fatigue, seizures, developmental delays and psychiatric problems. Current treatments focus on managing ammonia levels through a low-protein diet, medications that remove excess nitrogen, and, in severe cases, liver transplants.

Exploring genetic mutations and their effects

The Human Gene Mutation Database reports 486 known mutations in the OTC gene. Of these, 332 involve a change in a single building block of DNA and can weaken or completely disable the enzyme.

“It is also possible that only certain mutations actually reduce the activity of the enzyme,” says Beuning. “Some of them will appear randomly and not be associated with disease, even if they are in a person’s cells.”

In their experiments, the researchers discovered something unexpected: Some disease-related mutations behaved normally in test tube experiments, but were impaired when tested in living cells.

The researchers focused on specific amino acids in the enzyme that can turn their electrical charge on or off, a property that allows the protein to catalyze chemical reactions. They calculated a measurement called μ4, which describes how strongly these charged amino acids interact with their environment and helps predict which genetic mutations might interfere with the enzyme’s function.

“One of the advantages of the machine learning method is to reduce the set of mutations to identify those most likely to change OTC activity,” says Beuning.

POOL learns patterns in biological data (for example, which mutations are most damaging), even when scientists don’t have complete information about each case. It predicts which variants are likely to cause stronger or weaker effects based on the available evidence.

Machine learning reveals new insights

The scientists selected 17 mutations associated with the disease and one additional mutation to study in detail. Half of them were predicted to cause disease by directly altering the enzyme, and the other half by some other mechanism.

POOL combined with μ4 analysis correctly predicted which 17 out of 18 mutations hindered the enzyme’s ability to do its job. Most mutations that did not hinder the enzyme in the glass hindered the enzyme in the cells.

The team also proposed possible explanations for how the disease develops after analyzing the affected enzymes in cell cultures.

“The scale on which we carried out this study would not have been possible without the innovations of our doctoral students,” explains Beuning. “Their very dedicated efforts made it possible to obtain the enzyme and study its mutations in the laboratory to determine the effects of the mutations at the molecular level.”

The results show that μ4 measurement can complement existing bioinformatics tools to predict how mutations affect enzyme activity.

Next steps and ongoing research questions

Ondrechen says research has already revealed some reasons why certain mutations directly impair the enzyme’s ability to speed up chemical reactions. The next challenge is to understand why other mutations, which do not directly affect catalysis, nevertheless lead to diseases.

“That’s the hardest question, and that’s what we’re studying now,” she says.

There are many possible explanations, Beuning says, from how much protein the cell produces to its interactions with other proteins in the pathway.

“We are currently working to understand these other factors contributing to enzyme activity to determine how these different mutations affect activity, which, of course, could be different factors for the different mutations,” she says.

This story is republished courtesy of Northeastern Global News news.northeastern.edu.