Don’t like bitterness? Let the machine “taste” for you first

Produced by: Science Popularization China

Author: Denovo Team

Producer: China Science Expo

When we think of the taste of food, we often think of sour, sweet, bitter, salty, and fresh, among which bitterness is often the least popular. In fact, one of the important factors that cause bitterness in food is a biological molecule called bitter peptides. People have been trying to stimulate or inhibit these bitter peptides to achieve people's pursuit of food flavor.

Why do people taste bitter?

Humans and other mammals have a highly developed sensory system, in which taste plays an important role in many aspects. Among them, the perception of bitter taste is mainly due to a specific type of cell surface receptors present on the tongue, throat and other taste-related tissues in the mouth - TAS2Rs taste receptors .

These receptors are a type of G protein-coupled receptors (GPCRs) that can recognize and bind to specific types of bitter molecules. When this binding occurs, a series of signal transduction pathways within the cell are activated, including the activation of cAMP and calcium ion channels. These signals are ultimately transmitted to the taste center of the brain, and after further information processing, people produce a subjective feeling of bitterness.

TAS2Rs taste receptors in the cavity bind bitter molecules

(Image source: Generated by the author using AI)

So what are the bitter peptides that trigger the perception of bitterness?

This needs to start with peptides.

Peptides are a class of biological macromolecules composed of two or more amino acids connected by peptide bonds (-CO-NH-). This connection occurs between the carboxyl group (-COOH) of one amino acid and the amino group (-NH2) of another amino acid.

Amino acids can be thought of as pearls, and if you string the pearls together with short strings, they become polypeptides. If the length of the peptide is long enough, it will fold and form the protein we are familiar with.

Amino acids are like pearls, peptides are like short pearl bracelets, and proteins are like long, stacked pearl necklaces.

(Image source: Generated by the author using AI)

Peptides have diverse functions in organisms. They can act as signaling molecules and participate in various biological processes such as immune response and hormone regulation.

Bitter peptides are a special type of peptides. They are small molecule peptides that make us feel bitter by stimulating taste buds . The structure and length of these bitter peptides may vary. Usually, proteins in food are hydrolyzed to produce these short peptides during processing, storage or digestion.

Therefore, fermented products such as bean cakes, soy sauce, cheese, caviar and many natural foods contain a large amount of bitter peptides. Some foods will also produce bitter peptides after protease hydrolysis in our mouth.

Biological functions of bitter peptides

Bitter peptides typically have one or more biological functions, the most notable of which may be acting as a natural warning mechanism to help animals avoid ingesting toxic or harmful substances.

In the context of evolutionary biology, bitter taste is often considered a defense mechanism , with many plants producing bitter peptides to deter consumption or reduce the likelihood of being eaten.

In addition, bitter peptides may also be involved in some biological regulatory effects , such as regulating digestion, affecting the absorption of nutrients, or acting as precursors or antagonists of certain biologically active substances (such as hormones or neurotransmitters).

Since bitter peptides produce unpleasant tastes and reduce the taste quality of food, people will instinctively avoid such substances. However, the bitter taste in substances such as beer, coffee, and cheese is a very important sensory criterion that helps to increase people's preference for the food or beverage.

Therefore, bitter peptides have been widely studied in order to improve human taste experience of food.

Bitterness in beer, coffee, cheese, etc. is a very important sensory criterion

(Image source: Generated by the author using AI)

How to remove bitter peptides from food?

Bitter peptides are widely found in various foods, especially in bitter melon, some green leafy vegetables (such as spinach and broccoli), soybeans and their products, mature or fermented cheese, and certain specific organ meats (such as liver and spleen). In addition, some natural plants and herbs, such as Sophora flavescens and Coptis chinensis, also contain high concentrations of bitter peptides. Beverages such as some teas and coffees may also contain a certain amount of bitter peptides.

Momordica charantia

(Photo source: veer photo gallery)

Studies have shown that the bitterness of most bitter peptides is caused by the hydrophobic amino acids in them, and the intensity of their bitterness is mainly determined by factors such as the types and arrangement order of the hydrophobic amino acids . Therefore, artificially discovering such bitter peptides and removing/reducing bitter peptides through separation and purification or adsorption, and enzymatic modification can regulate the bitterness in food.

The processing and cooking methods of food may affect the content and properties of its bitter peptides , mainly because high temperature, enzyme activity, pH changes and other additives can change the structure or stability of peptide molecules.

For example, high-temperature cooking may cause peptide molecules to degrade or change their three-dimensional structure, thereby reducing bitterness, while enzymes introduced during fermentation or processing can decompose or transform bitter peptides. At the same time, changing the pH value of food or adding other seasonings and additives may also affect the solubility of bitter peptides or their interaction with other ingredients, further changing their content and taste in food.

How foods are processed and cooked can affect the amount of bitter peptides

(Photo source: veer photo gallery)

Therefore, even though bitterness is an inevitable consequence of protein hydrolysis, efforts are still made to reduce the bitterness of our foods by masking, removing or preventing it.

Can machine learning also predict bitter peptides?

After ultra-high temperature (UHT) sterilization, milk can be stored for a long time, but some UHT milk often tastes bitter during shelf storage. Studies have shown that this is because casein in milk may be hydrolyzed by some thermostable proteases to produce bitter peptides. Identifying bitter peptides in spoiled UHT milk will help analyze the cause of the bitter taste of UHT milk and improve its quality.

Traditional laboratory methods are reliable methods for identifying bitter peptides. Peptidomics methods based on liquid chromatography tandem mass spectrometry can effectively analyze peptides in spoiled UHT milk, but identifying bitter peptides is a difficult task that is usually time-consuming and expensive.

To solve this problem, a team from the Dalian Institute of Chemical Physics, Chinese Academy of Sciences, used peptidomics technology and machine learning to build a prediction model (CPM-BP) to efficiently identify bitter peptides. Peptidomics refers to a technology used to analyze protein fragments (i.e., peptides), while machine learning uses algorithms to "teach" computers how to make complex predictions.

Machine learning algorithm predicts bitter peptides

(Image source: Reference [7])

The researchers used an algorithm called "Light Gradient Boosting Machine" to successfully build a predictive model that had an accuracy of 90.3% for predicting bitter peptides in an independent test set.

To verify the validity of the model, the researchers also compared spoiled and fresh UHT milk. They found 180 potential bitter peptides in spoiled milk and further verified some of them using HEK293T cell lines, which were modified to express a specific human bitter receptor (hT2R4), and three of the potential bitter peptides had the ability to activate this bitter receptor (hT2R4). Ultimately, the experiment successfully demonstrated the validity of the model.

In short, this study not only provides a faster and more economical method to identify bitter peptides, but also successfully verifies the accuracy and reliability of its prediction model, which has important application value in food science, drug development and nutrition research.

Conclusion

As social and economic levels continue to improve, consumers have higher and higher expectations for food quality, not only pursuing health but also paying more attention to taste. Machine learning has therefore been applied to more scenarios. Since it involves a variety of bitter compounds and bitter receptors, as well as their complex interactions with other flavor components, research in this area still faces considerable challenges, and scientists are also working hard to discover more "secrets" about bitterness.

References

[1] Huang Lei, Zhou Qiyang. Research progress of bitter peptides in soy protein hydrolysate[J]. Modern Food, 2022, 28(21): 22-24.

[2] Lin Tong, Zhao Jichun, Lei Xiaojuan, et al. Research progress on preparation of flavor peptides from edible fungi and their flavoring mechanism[J]. Food and Fermentation Industries, 2022, 48(19): 313-319.

[3] Si Kuolin, Li Zhiguo, Li Lingyu, et al. Research progress on the formation and debittering methods of cheese bitter peptides[J]. Food Industry, 2021, 42(03): 267-271.

[4] Wang Zhifei, Lin Lu, Sun Weifeng, et al. Research progress on bitter peptides and bitter taste receptors[J]. Chinese Condiments, 2016, 41(09): 152-156.

[5] Feng Hongxia, Lu Zhaoxin, You Hua. Research on the formation of bitter peptides and debittering methods [J]. Food Science, 2002(05):151-154.

[6] Höhme, Lisa, Christin Fischer, and Thomas Kleinschmidt. "Characterization of bitter peptides in casein hydrolysates using comprehensive two-dimensional liquid chromatography." Food Chemistry 404 (2023): 134527.

[7] Yu, Yang, et al. "Identification and prediction of milk-derived bitter taste peptides based on peptidomics technology and machine learning method." Food Chemistry 433 (2024): 137288.