An artificial intelligence system to predict the taste of new drugs

A team of experts at University College London are working to make drugs more palatable – and at a speed never seen before

Scientists are using artificial intelligence to make medicines taste better, combining years of data to break through one of the biggest obstacles in paediatric care.

Poor taste is one of the most significant barriers for children taking medicine, not just in daily life but in adherence to drugs that have to be taken on a long-term basis, like HIV antiretrovirals and antibiotics for tuberculosis.

It can often result in treatment failure and complications, as well as fueling antimicrobial resistance (AMR).

But now a team of experts at University College London are working to make drugs more palatable – and at a speed never seen before. 

Data collected by an “electronic tongue” is being used to create an AI model for predicting taste in medicine.

The new model breaks down a drug into a series of chemical descriptors that determine taste and is trained to map these out to predict levels of bitterness. 

“We run a machine learning algorithm to basically try and see what’s the chemical structure, what’s the molecular structure, what are the other chemical physical parameters that make it bitter, and try to see if there’s a relationship,” said Dr Hend Abdelhakim, an assistant professor at the UCL Global Business School for Health.

• ‘Mainly a problem with children’

Other taste qualities that can be detected include salty, sweet, sour, umami and astringent.

Bitterness and astringency are the main focus as they are the tastes that make it less likely that patients will adhere to their medicines.

The AI will speed up the drug development process by skipping the need to run human trials for taste in the early stages of drug development. Eventually it is hoped the AI won’t even need the e-tongue. 

Ordinarily, drugs are assessed in a lab and are given a rating for their taste before being tested in taste trials, but this can be time consuming and expensive. 

The AI model being developed will be an open access tool, meaning that pharmaceutical development around the world can benefit from the data on palatability of commonly used drugs.

For many long-term medications, including those for HIV or diabetes, taste is one of the most important factors in how likely people are to adhere to the regimen they are prescribed. 

It is particularly important for vulnerable groups, including the old and very young. 

“It’s mainly a problem with children, because they have a heightened sense of taste. With chronic medications it does impact compliance. It’s not just the child being fussy,” said Dr Abdelhakim.

A total of 63 per cent of children aged 10 to 18 identified “bad taste of medicines” as a barrier in a study in the European Union. 

“It’s a problem for longer term diseases, so for example, HIV,” explained Dr Abdelhakim.

“Antiretroviral medicines don’t taste very well. So if the patient has to take those pills every day for life, it’s a much bigger problem, especially if they start them very, very young.

“Even if it’s a wonder drug, if the patient doesn’t take it, it won’t work.”

Treatment adherence with antibiotics is particularly important given the risk of AMR if a complete course of treatment is not taken. 

“With antibiotics, if the patient doesn’t take it, yes, they won’t get better. But also you’re going to contribute to antimicrobial resistance,” said Dr Abdelhakim.

“It’s actually a bigger problem for the rest of us. It might be one small cog in the machine, but if we can tackle it, then that’s good.”

Follow Emirates 24|7 on Google News.