Anyone who's tried a sugar-free drink or snack knows that sweetness doesn't always taste quite like sugar. Some substitutes leave behind a bitter or metallic aftertaste, while others have a sweetness that lingers a little too long. What if scientists could find alternatives that taste more like the real thing?
Researchers at the Hebrew University of Jerusalem are using artificial intelligence to explore that possibility. In a new study published in Food Chemistry, they screened more than 50 million chemical compounds and identified 92 promising candidates for future sugar substitutes. But there's a catch: The findings are based entirely on computer predictions, so researchers don't yet know whether the compounds actually taste sweet or are safe to consume.
Finding an alternative that delivers the sweetness of sugar without its calories has long been a challenge for food scientists. Popular substitutes such as aspartame, saccharin and sucralose can help reduce added sugar in foods and beverages, but differences in taste can make them less appealing to some consumers.
To search for better options, researchers Shayma El-Atawneh and Amiram Goldblum developed computer models that learned to recognize characteristics of substances known to taste sweet or bitter.
They then used those models to search through approximately 50.7 million compounds, looking for ones that might deliver sweetness without an unpleasant aftertaste.
The initial search identified hundreds of promising possibilities. After additional screening, the researchers narrowed the list to 92 candidates they considered worth investigating further.
Some of the compounds have chemical structures unlike those of familiar sugar substitutes, suggesting that scientists may have new possibilities to explore rather than simply modifying ingredients already in use.
Still, finding potential sweeteners on a computer is very different from developing ingredients that can be added to foods.
Researchers will first need to determine whether the compounds actually produce a sweet taste, how pleasant that sweetness is and whether they have unwanted aftertastes. They'll also need to establish whether the substances are safe to consume and can work effectively in foods and beverages.
The computer models aren't foolproof, either. In fact, the researchers found that their predictions failed to correctly identify sucralose, an existing sugar substitute, as sweet.
There's also no evidence yet that any of the new candidates would be healthier than today's alternatives. The study didn't examine how the compounds might affect blood sugar, weight or other aspects of human health.
For consumers, that means there's no new sugar substitute on the horizon just yet. But the research suggests AI could make the search more efficient, helping scientists identify promising ingredients without having to test millions of possibilities individually.
Whether any of those possibilities eventually finds its way into a favorite sugar-free drink or snack remains to be seen.
The study was conducted by researchers affiliated with the Hebrew University of Jerusalem. No specific funding source was identified. Goldblum disclosed a pending patent as a potential competing interest, although further details were not provided.