The challenge of bitter peptides is particularly acute in the growing market for plant-based protein foods. As manufacturers turn to peas, soy, and other plant proteins to replace animal products, the enzymatic breakdown of those proteins during processing generates peptide mixtures whose sensory profiles are difficult to anticipate. Undesirable bitter notes are a frequent reason that plant-based alternatives fail to win over consumers. At the same time, bitter peptides are not simply a defect to be eliminated; many of them carry physiological properties and may, for example, play a role in regulating hunger and satiety. Understanding which sequences taste bitter, and why, therefore has value both for improving flavor and for designing foods with functional benefits.

Antonella Di Pizio, principal investigator of the new study and head of the Molecular Modeling research group at Leibniz-LSB@TUM, underscores the broader stakes of the work. To make plant-based protein sources more attractive for food production and to use them more sustainably, she explains, researchers need to understand which peptides taste bitter and what structural features characterize them. AI-based methods, she argues, can make an important contribution to exactly that understanding. Her team, which also included researchers from the Technical University of Munich and Pompeu Fabra University in Barcelona, set out to build a computational pipeline that could move the field beyond slow, trial-and-error sensory testing.

The technical core of the new approach lies in the combination of two complementary machine learning components. The first is a protein language model, a type of neural network trained to capture the statistical patterns of amino acid sequences in much the same way that large language models learn the structure of human text. The team trained this model using approximately 500 known bitter-tasting peptides, allowing it to internalize the sequence features associated with bitterness. The second component is BitterPep-GCN, a prediction model the group had developed in earlier work and described in 2024 in the Journal of Cheminformatics. BitterPep-GCN is a Graph Convolutional Network, a specialized form of artificial neural network designed to analyze structured data. In this case, the structured data are peptide molecules themselves, represented as graphs in which amino acid residues function as nodes and the chemical relationships between them as edges.

Graph-based representations give the model an advantage that purely sequence-based methods can lack. Because the network processes the peptide as a structured chemical object rather than a simple string of characters, it can learn how the arrangement, identity, and interactions of residues influence binding to bitter taste receptors. By fusing the knowledge encoded in the protein language model with the structural sensitivity of the graph convolutional network, the researchers created a system capable of both classifying existing peptides and generating novel candidate sequences. This dual capability, known in the field as de novo design, is what distinguishes the new method from earlier bitterness predictors that could only score peptides already in hand.

Putting the system to the test, the researchers first used it to generate 161 new peptide sequences that had never been experimentally characterized. The pipeline then filtered this set, identifying the candidates that, according to the model predictions, were highly likely to taste either strongly bitter or clearly non-bitter. Selecting the most promising of these designed molecules, the team had them chemically synthesized and submitted to a trained sensory panel for evaluation. Human tasters, rather than receptor assays alone, provided the ground truth, which is a demanding standard for any computational model of flavor.

The results were striking. Of the 31 designed peptides ultimately tasted, the trained panel confirmed the AI predictions in 25 cases, correctly classifying them as bitter or non-bitter. In the course of the experiments, the researchers also identified numerous previously unknown bitter-tasting and non-bitter-tasting peptides, expanding the experimental dataset available to the field. For a property as subtle and receptor-specific as bitterness, a prediction accuracy of roughly 80 percent in a blind de novo design setting represents a substantial step forward, and it demonstrates that generative models can produce chemically meaningful candidates rather than merely ranking known compounds.

Alexandra Steuer, first author of the study and a doctoral student in Di Pizio’s group, emphasizes what the results mean for the discipline. The findings show, she notes, that not only can the bitterness of peptides be predicted, but that the new AI-based method can also be used to specifically design new bitter-tasting peptides. Di Pizio adds that this brings researchers significantly closer to the goal of proactively controlling taste characteristics, rather than reacting to off-flavors after they appear in a finished product. The distinction between reactive quality control and proactive molecular design captures the practical promise of the approach.

Di Pizio also stresses that the research is ready to be implemented in application frameworks. In the long term, the new findings could help to specifically control the formation of bitter-tasting peptides during food production, a capability she describes as particularly relevant for plant-based, protein-rich foods, whose acceptance often suffers because of undesirable flavor notes. If manufacturers can predict, early in product development, which peptides will emerge from a given protein source and processing regime, they could adjust fermentation starters, enzyme choices, or formulation strategies to steer the flavor outcome. Conversely, the ability to design bitter peptides on demand could support research into appetite regulation and satiety, where bitterness may play a functional physiological role.

The study, titled De novo design and experimental characterization of bitter peptides, was published in npj Science of Food on June 25, 2026, with a author team including Steuer, Ferri, Eckrich, Heidenkampf, Mittermeier-Kleßinger, Schaefer, Behrens, Ferruz, Dawid, and Di Pizio. Training data for the language model came from the Bitter Peptide Space (BPS)-1000 database maintained by the Leibniz Institute, a curated resource that underpins much of the group’s computational work. The research was performed computationally at its core, with experimental validation through synthesis and human sensory testing, and the authors declare no competing interests. As machine learning continues to move from analyzing existing molecules to creating new ones, the Munich-led study offers a concrete demonstration that the taste of tomorrow’s foods, down to the individual peptide, can increasingly be designed rather than discovered.

Subject of Research: AI as a tool in flavor research

Article Title: AI as a tool in flavor research

Article References: AI as a tool in flavor research. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: tool, flavor, research, scientific research, peer-reviewed research, research findings

The team tackled a deceptively simple question: when two tomato species look and taste different, how much of that difference comes from the genes themselves versus the molecular switches that control them? To find out, they exploited a clever property of F1 hybrids. When a cultivated tomato is crossed with a wild relative, every cell contains one chromosome set from each parent. Because the two parental copies of each gene sit side by side in the same cellular environment, exposed to identical transcription factors and signaling molecules, any difference in how strongly the two copies are expressed can be attributed to sequences on or near the gene itself. This phenomenon, known as allele-specific expression, allows scientists to separate cis-regulatory effects, which act locally on DNA elements such as promoters and enhancers, from trans-regulatory effects, which arise from diffusible factors encoded elsewhere in the genome.

The scale of the analysis sets the study apart. Rather than examining a single cross at a single moment, the researchers generated hybrids between cultivated tomato and three wild relatives spanning a gradient of evolutionary distances: Solanum pimpinellifolium, the closest wild ancestor; Solanum neorickii at an intermediate distance; and the more distantly related Solanum pennellii. They then profiled gene expression across three distinct fruit tissues and multiple developmental stages, from early cell division through ripening. To support the work, the team assembled new genome sequences and annotations for the parental lines, ensuring that sequencing reads from each hybrid could be assigned accurately to the maternal or paternal copy of every gene.

The resulting dataset delivers an unambiguous headline: cis-regulatory divergence is the dominant force shaping expression differences between tomato species. Across tissues, developmental stages, and levels of relatedness, local regulatory changes consistently explained a larger share of parental expression differences than trans effects. This makes evolutionary sense. Cis-regulatory mutations tend to affect a single gene or small set of genes, allowing fine-tuned changes without catastrophic collateral damage, whereas mutations in trans-acting factors can perturb hundreds of downstream targets simultaneously and are more likely to be deleterious. Natural selection, the authors argue, has repeatedly favored regulatory variants that adjust individual genes without breaking the broader network.

Yet the study also shows how much this simple dichotomy oversimplifies reality. The majority of cis-regulatory effects proved to be tissue-specific, stage-specific, or both. A gene whose wild allele dominates expression in the fruit pericarp may show no such bias in the placenta, or at a different point in ripening. The researchers catalogued genes classified as cis-only, cis-plus-trans, and cis-by-trans, where a cis difference is itself modulated by trans-regulation depending on context. This last category is particularly intriguing, because it implies that local regulatory variants do not act in isolation; their effects can be amplified, suppressed, or reversed depending on the trans environment supplied by the other parent and by the developmental program of the tissue.

Evolutionary distance emerged as a second organizing principle. In hybrids with S. pimpinellifolium, the closest relative, regulatory divergence between the two alleles was comparatively modest and trans effects retained a noticeable share of the action. As the wild parent became more distantly related, cis-regulatory contributions grew progressively larger. The inheritance patterns of expression levels shifted in parallel, indicating that the architecture of regulatory variation is not static across the tomato clade but accumulates and reorganizes over evolutionary time. For crop scientists, this is an encouraging message: the more exotic the germplasm, the greater the reservoir of independent regulatory variants available for breeding.

The practical payoff lies in the pathways the team traced. They found extensive cis-regulatory divergence in genes governing carotenoid biosynthesis, the source of lycopene and beta-carotene that give ripe tomatoes their color and contribute provitamin A; in the phenylpropanoid and flavonoid pathways, which produce antioxidants linked to human health; and in steroidal glycoalkaloid metabolism, which influences bitterness and toxicity in wild fruit. Genes controlling sugar accumulation, a decisive factor in perceived sweetness, and volatile organic compounds, which shape aroma, likewise showed strong allele-specific patterns. In several cases, the wild allele carried regulatory variants that boosted expression of biosynthetic genes in specific tissues or stages, suggesting concrete targets for reintroducing flavor and nutrition into elite cultivars without disrupting yield-related traits.

Among the individual genes highlighted are SlKLUH, a cytochrome P450 involved in fruit growth whose cis-regulation proved strikingly tissue-specific, and ZDS, a key enzyme in the carotenoid pathway. The promoter analysis added a mechanistic layer: a substantial number of differently regulated genes carried structural variations of 30 base pairs or more in the two-kilobase region upstream of the gene, pointing to insertions, deletions, and rearrangements as frequent sources of cis-regulatory novelty. Such structural variants are often invisible to standard SNP-based studies, which may explain why genome-wide association work has undersold the regulatory component of fruit quality variation.

For breeders, the message is that wild tomato relatives are not merely a source of disease-resistance genes but a deep library of regulatory alleles that can tune when, where, and how strongly the fruit’s metabolic genes operate. Because cis-regulatory variants tend to be narrowly scoped, introgressing a wild promoter behind a domesticated gene could, in principle, enhance a specific flavor compound without dragging along the yield penalties that have historically made wild germplasm unattractive. The study’s high-resolution map of which genes are cis-regulated, in which tissue, and at which stage, offers a direct roadmap for such precision breeding, whether through marker-assisted selection or genome editing of regulatory regions.

The work also carries a broader evolutionary lesson about the domestication bottleneck. Decades of selection for size, uniformity, and shelf life narrowed the genetic diversity of cultivated tomato, and with it the diversity of regulatory variants shaping flavor chemistry. By quantifying how much regulatory divergence separates the crop from its wild cousins, and by showing that this divergence concentrates in exactly the pathways that define fruit quality, the study reframes wild relatives as the key to restoring what domestication left behind. As sequencing costs fall and allele-specific analysis becomes routine in crop genomics, the tomato map is likely to become a template for dissecting regulatory variation in other fruits, from pepper to melon, where the difference between a memorable harvest and a forgettable one often comes down to the switches, not the genes.

Subject of Research: Allele-specific gene expression and regulatory divergence between cultivated and wild tomato species

Article Title: Allele-specific expression reveals complex regulatory divergence underlying fruit phenotypic differences between cultivated and wild tomato species

Article References: Zhao, J., Nicolas, P., Xu, Y., Vrebalov, J., Giovannoni, J., Fei, Z., & Catala, C. (2026). Allele-specific expression reveals complex regulatory divergence underlying fruit phenotypic differences between cultivated and wild tomato species. Genome Biology. https://doi.org/10.1186/s13059-026-04279-5

Image Credits: AI Generated

DOI: 10.1186/s13059-026-04279-5

Keywords: tomato, allele-specific expression, cis-regulatory variation, wild relatives, fruit development, flavor, carotenoids, flavonoids, F1 hybrids, gene regulation, Solanum pennellii, crop breeding

The research team, led by Si-Jia Lin and corresponding author Xu-Hui Huang of Dalian Polytechnic University in China, set out to identify the volatile compounds responsible for both the beloved and the objectionable notes in wakame, and then to track how those compounds respond to two commercial levers: progressive pickling and packaging material. China produced 206,100 tons of wakame in 2022, and global output has climbed steadily since 2012, so even small improvements in flavor retention carry substantial commercial weight. The team’s central question was whether the off-odors that plague the industry could be tamed through processing rather than through additives or breeding.

To do this, the researchers combined classical sensory evaluation with an arsenal of instrumental techniques. Gas chromatography-mass spectrometry revealed 55 volatile compounds in the samples, including 19 aldehydes, 10 alcohols, 6 ketones, 5 acids, and 4 esters. Aldehydes, with their characteristically low odor thresholds, emerged as the dominant contributors to wakame’s smell, accounting for 35.2 percent of the volatile profile, followed by alcohols at 16.6 percent and ketones at 13.1 percent. Using odor activity values, a metric that weighs a compound’s concentration against the concentration at which humans can detect it, the team narrowed the field to 18 decisive odorants. Among these were hexanal, (E)-2-nonenal, (E)-2-decenal, beta-cyclocitral, caryophyllene, and 1-octen-3-one, each imparting everything from grassy and oily notes to floral, fruity, earthy, and mushroom-like character.

The experimental design was rigorous. Ready-to-eat wakame supplied by a commercial producer in Liaoning was washed, desalted, blanched at 90 degrees Celsius for two minutes to protect color, and then subjected to two sequential seasoning steps: a first pickling at 10 degrees Celsius for two hours in a 4 percent salinity solution, and a second pickling for half an hour at 0.5 percent salinity. Samples were then sealed either in aluminum foil bags or in polyethylene terephthalate, or PET, casings, sterilized, and stored at 4 degrees Celsius. Six sample categories captured the full matrix of two packaging types across three processing stages. Thirty trained panelists scored smell, taste, color, and texture, while an electronic nose fitted with ten metal oxide semiconductor sensors and an electronic tongue capable of distinguishing 25 taste scales provided objective corroboration.

The sensory and colorimetric data told a nuanced story about salt. Proper pickling turned wakame greener, a effect the researchers attribute to chlorophyll forming more stable complexes with metal ions such as sodium from sodium chloride. Elevated salt also suppresses enzymatic activity and microbial growth by lowering water activity, slowing the oxidation reactions that generate stale aromas. But there is a tipping point. The second pickling round damaged pigment, darkened color, and softened texture, with hardness dropping significantly after the first seasoning and gumminess falling significantly after the second. The lesson, the authors suggest, is that moderate salinity preserves quality while excessive salinity destroys it, a balance producers must strike carefully.

On taste, the electronic tongue revealed that progressive pickling reduced bitterness, a trait traced to mannitol, algin, bitter amino acids, and bitter peptides in the seaweed. As salt levels rose, bitter amino acid content fell, and panelists’ umami, sweet, and salty scores climbed. The electronic nose, meanwhile, recorded rising responses from sensors sensitive to nitrogen oxides and sulfur compounds, reflecting the way salt disrupts cell structures and releases nitrogenous material. Panelists noted declining fishy, rancid, and overripe aromas as pickling progressed, and principal component analysis separated the samples almost perfectly, with cumulative contribution rates of 99.72 percent for odor and 99.66 percent for taste.

The quantitative volatile data were even more striking. After progressive pickling, the content of beta-cyclocitral, a signature algae-derived odorant, surged from 2.52 to 485.42 nanograms per gram, while caryophyllene, a pleasantly floral and fruity sesquiterpene, rose from 1.35 to 578.05 nanograms per gram. These increases enriched wakame’s characteristic marine aroma and helped mask off-notes. Simultaneously, the levels of undesirable compounds declined: 1-octen-3-one, an earthy, metallic ketone formed by oxidation of unsaturated fatty acids, dropped from 641.4 to 371.5 nanograms per gram in aluminum foil packages and from 725.66 to 459.08 nanograms per gram in PET, while fishy (E)-2-nonenal in foil-packaged wakame fell from 11 to 3.21 nanograms per gram.

Packaging proved to be the quiet hero of the study. Aluminum foil, with its low permeability to gases and water vapor and its complete opacity to light, preserved more of the volatile compounds that make wakame appealing while blocking the migration of oxidation-promoting substances into the food. Foil-packaged wakame retained higher hardness, stickiness, and resilience, showed greener color, and contained compounds such as (E,E)-2,4-heptadienal, 1-heptanol, and (E)-3-hexen-1-ol that were undetectable in the PET samples. Beta-cyclocitral levels in foil-packaged wakame reached 485.42 nanograms per gram compared with just 128.67 nanograms per gram in PET. In total, foil-packaged samples after double seasoning contained 16 key odorants against 14 in their PET counterparts, and panelists judged the foil samples more aromatic and less fishy.

To verify that these 18 compounds truly drive wakame’s aroma, the team built six aroma recombination models by adding the key odorants at their measured concentrations to an odorless wakame matrix, and then performed omission tests in which single compounds were removed and 15 panelists ran triangle tests to detect the difference. In freshly processed samples, removing hexanal, (E)-2-nonenal, (E)-2-decenal, or 1-octen-3-one produced significant to highly significant perception changes, confirming their importance. After pickling, beta-cyclocitral became the standout, while compound interactions grew so complex that individual omissions became harder to detect, a phenomenon the authors attribute to synergies among the expanding roster of odorants.

The practical implications are immediate for a global seaweed industry seeking to convert health-conscious consumers into repeat buyers. Choose packaging that blocks oxygen, moisture, and light; calibrate salt levels to stabilize pigment and suppress microbes without wrecking texture; and accept that progressive seasoning, done judiciously, can flip wakame’s chemistry from off-odor generator to flavor enhancer. As seaweed moves from niche health food to mainstream sustainable protein source, studies like this one show that flavor is not an accident of nature but a controllable outcome of engineering, one aluminum foil bag at a time.

Beyond the headline findings, the study’s methodology offers a window into how modern flavor science increasingly operates. By pairing trained human panels with electronic nose and electronic tongue instruments, the researchers followed a growing trend in food analysis sometimes called sensory omics, in which machine-based readings are calibrated against human perception to produce reproducible, quantifiable flavor fingerprints. This dual approach helps offset the subjectivity and fatigue inherent in panel work, since each taster in the study fasted for three hours beforehand and evaluated samples in isolated compartments under controlled lighting and temperature.

The findings also sit within a broader body of research on seaweed preservation. Earlier work on kelp showed that higher salt concentrations and lower temperatures reduce alginate lyase activity, an enzyme that degrades cell wall polysaccharides and accelerates quality loss. Comparable strategies, including the combination of natural antioxidants with fermentation to remove fishiness from fresh kelp, have been reported by other Chinese research groups, suggesting that salt-mediated enzyme inhibition is a recurring theme across brown algae processing.

Priorities for future work follow naturally from these results. Because plastic packaging can permit oxygen and moisture migration that drives oxidation, comparative studies of barrier materials, including glass, multilayer films, and foil laminates, may refine packaging recommendations further. Extending shelf-life trials beyond laboratory storage to real distribution conditions, and testing whether the same pickling thresholds hold for other commercially farmed seaweeds, would help translate this molecular map into industry-wide standards for one of the world’s fastest-growing aquaculture sectors.

Subject of Research: How pickling methods and packaging materials affect the flavor quality of ready-to-eat wakame

Article Title: Effect of pickling and packaging difference on characteristic flavor quality of ready-to-eat wakame

Article References: Lin, S.-J., Zhang, T.-T., Guo, Y., Zhang, K., Qin, L., & Huang, X.-H. (2026). Effect of pickling and packaging difference on characteristic flavor quality of ready-to-eat wakame. Blue Biotechnology, 3(1), Article 8. https://doi.org/10.1186/s44315-026-00059-9

Image Credits: AI Generated

DOI: 10.1186/s44315-026-00059-9

Keywords: wakame, pickling, flavor, volatile compounds, aluminum foil, PET packaging, seaweed, beta-cyclocitral, caryophyllene, off-odor, electronic nose, food quality