BitterMasS: Predicting Bitterness from Mass Spectra.

Bitter compounds are common in nature and among drugs. Previously, machine learning tools were developed to predict bitterness from the chemical structure. However, known structures are estimated to represent only 5-10% of the metabolome, and the rest remain unassigned or "dark". We present BitterMasS, a Random Forest classifier that was trained on 5414 experimental mass spectra of bitter and nonbitter compounds, achieving precision = 0.83 and recall = 0.90 for an internal test set. Next, the model was tested against spectra newly extracted from the literature 106 bitter and nonbitter compounds and for additional spectra measured for 26 compounds. For these external test cases, BitterMasS exhibited 67% precision and 93% recall for the first and 58% accuracy and 99% recall for the second. The spectrum-bitterness prediction strategy was more effective than the spectrum-structure-bitterness prediction strategy and covered more compounds. These encouraging results suggest that BitterMasS can be used to predict bitter compounds in the metabolome without the need for structural assignment of individual molecules. This may enable identification of bitter compounds from metabolomics analyses, for comparing potential bitterness levels obtained by different treatments of samples and for monitoring bitterness changes overtime.

Identification of Compounds That Impact Consumer Flavor Liking of American-European Hazelnut Hybrids Using Nontargeted LC/MS Analysis.

American-European (Corylus americana × Corylus avellana) hazelnut hybrids are being developed for the Midwest-growing region of the United States. However, an inadequate understanding of the compounds that impact the consumer acceptance of hazelnuts limits breeding programs. Nontargeted liquid chromatography/mass spectrometry (LC/MS) chemical profiles of 12 roasted hybrid hazelnut samples and the corresponding consumer flavor liking scores were modeled by orthogonal partial least squares with good fit and predictive ability (R2Y > 0.9, Q2 > 0.9) to identify compounds that impact nut liking. The five most predictive compounds (1-5) were negatively correlated to flavor liking, selected as putative markers, purified by multidimensional preparative LC/MS, structurally elucidated (nuclear magnetic resonance, MS), quantified, and validated for sensory relevance. Compound 1 was identified as 1″-O-3'-b-glucofuranosyl-1'-O-1-b-glucofuranosyl-(2,6-dihydroxyphenyl)-ethan-4-one. Compounds 2 and 4 were identified as rotamers of 2-(3-hydroxy-2-oxoindolin-3-yl) acetic acid 3-O-6'-galactopyranosyl-2″-(2″oxoindolin-3″yl) acetate, whereas compounds 3 and 5 were identified as rotamers of 1″-O-1'-b-glucofuranosyl-9-O-6'-b-glucopyranosyl-2″-(2″-oxoindolin-3″yl) acetate. Sensory evaluation determined that all compounds were characterized by bitterness and/or astringency. The sensory threshold values of compounds 1-5 were determined to be below the concentrations reported in 91, 83, 41, 25, and 41% of all 12 hybrid hazelnut samples, respectively, indicating they contributed to aversive flavor attributes.

Impact of rebaudioside A degradation compounds on flavor perception.

Rebaudioside A, a sweet-tasting steviol glycoside, is known to degrade in food products during storage and thought to contribute to flavor instability. The impact of rebaudioside A degradation compounds on flavor perception was investigated. Sensory descriptive analysis indicated rebaudioside A degradation compounds, at concentrations below detection thresholds, modified the perception of taste, somatosensorial, and retronasal aroma attributes of a strawberry-flavored model beverage. Gas chromatography/mass spectrometry analysis and orthonasal sensory tetrad tests further indicated the addition of the degradation compounds did not significantly alter the volatile aroma composition or orthonasal perception, respectively. Altogether, subthreshold unimodal and cross-modal integration of multisensory percepts were supported to impact the flavor performance of rebaudioside A.

Characterization of the impact of chlorogenic acids on tactile perception in coffee through an inverse effect on mouthcoating sensation.

Coffee "body" is acknowledged by coffee industry professionals to be an attribute which contributes meaningfully to overall coffee quality and is defined as the collective tactile sensation imparted by the beverage. Currently, there is limited knowledge of the chemical compounds that contribute to tactile attributes in coffee. In the present work, coffee body was determined to be comprised of 4 sub-attributes including mouthcoating, astringency, chalkiness, and thickness and the specific constituents contributing to the tactile sensation of mouthcoating were further pursued using sensory-guided fractionation via preparative-scale liquid chromatography. Signal detection-based sensory methodologies were employed to characterize the sensory effects elicited by selected compounds in water and coffee matrices. Two chlorogenic acids, 3-O-caffeoylquinic acid (3-CQA) and 4-O-caffeoylquinic acid (4-CQA), were observed to impart subtle but significantly perceptible mouthcoating effects in water and/or coffee. Counterintuitively, sensory perception was inversely related to compound concentration. Complex receptor-ligand interactions or salivary lubrication dynamics are discussed as two potential mechanisms to explain this inverse relationship. Taken together, the outcomes of the present study (1) provide new targets for coffee tactile sensation optimization and modulation, (2) identify a novel dimension of sensory impact for two compounds of the chlorogenic acid family, and (3) present a need for deeper investigation into 3-CQA and 4-CQA mechanisms of sensation.

Identification of compounds contributing to umami taste of pea protein isolate.

The umami taste of pea protein ingredients can be desirable or undesirable based on the food application. The compounds contributing to the umami perception of pea protein isolate (PPI) were investigated. Sensory-guided prep-liquid chromatography fractionation of a 10% aqueous PPI solution revealed one well-known compound, monosodium glutamate (MSG), however, it was reported at a subthreshold concentration. Two umami enhancing compounds 5'-adenosine monophosphate (AMP) and 5'-uridine monophosphate (UMP) were subsequently identified after the LC fractions were re-evaluated with MSG. Sensory recombination studies, utilizing the aqueous PPI solution as the base, confirmed AMP and UMP were umami enhancers of MSG and contributed approximately 81% of the perceived umami intensity. However UMP was only reported to enhance umami perception in combination with AMP (not individually) indicating synergistic interactions were observed between the two enhancer compounds. Therefore the presence of all three compounds are important for umami perception and provide an improved basis to tailor the flavor profile in PPI products.

Identification of a Bitter Peptide Contributing to the Off-Flavor Attributes of Pea Protein Isolates.

The aversive bitter taste of pea protein ingredients limits product acceptability. Compounds contributing to the bitter perception of pea protein isolates were investigated. Off-line multi-dimensional sensory-guided preparative liquid chromatography fractionation of a 10% aqueous PPI solution revealed one main bitter compound that was identified by Fourier transform ion cyclotron resonance mass spectrometry and de novo tandem mass spectrometry (MS/MS) sequencing as the 37 amino acid peptide PA1b from pea albumin and further confirmed by synthesis. Quantitative MS/MS analysis reported that the concentration of the bitter peptide was 129.3 mg/L, which was above the determined bitter sensory threshold value of 3.8 mg/L and in agreement with the perceived bitter taste of the sample.

Identification of compounds that enhance bitterness of coffee brew.

The bitterness perception of coffee is a key attribute that impacts consumer acceptance. Nontargeted liquid chromatography/mass spectrometry (LC/MS) flavoromics analysis was applied to identify compounds that enhance the bitter perception of roasted coffee brew. Orthogonal partial least squares (OPLS) analysis was used to model the comprehensive chemical profiles and sensory bitter intensity ratings of fourteen coffee brews with good fit and predictivity. Five compounds that were highly predictive and positively correlated to bitter intensity were selected from the OPLS model, further isolated, and purified using preparative LC fractionation. Sensory recombination testing demonstrated that five compounds significantly enhanced the bitter perception of coffee when presented as a mixture, but not when presented individually. In addition, a set of roasting experiments revealed the five compounds were generated during the coffee roasting process.

Mono-n-butyl Malate-Derived Compounds from Camu-camu (Myrciaria dubia) Malic Acid: The Alkyl-Dependent Antihyperglycemic-Related Activity.

Malic acid derivatives from camu-camu (Myrciaria dubia) fruit exhibited a strong in vitro inhibitory activity toward pancreatic α-amylase and α-glucosidase enzymes. During a bioguided chromatographic fractionation process of the whole fruit (pulp and peelings) polar extract, isomers (S)-4-butoxy-2-hydroxy-4-oxobutanoic acid (1) and (S)-4-butoxy-3-hydroxy-4-oxobutanoic acid (2) (84:16) were isolated and identified as a potent inhibitor of α-amylase (IC50= 11.69 ± 1.75 µg/mL) and α-glucosidase (IC50 = 102.69 ± 4.16 µg/mL). The chemical structures were confirmed by HPLC-ESIMS and 1H and 13C NMR (one- and two-dimensional) analyses. The structure-based virtual screening demonstrated that the aliphatic moiety plays a significant role in the binding mode of the test alkyl malate esters. Compound 1 exhibited the best interaction profile to bind both enzymes, having key structural features to form relevant contacts by involving adequate enzyme-ligand complex stabilization and compactness over time.

Identification of subthreshold chlorogenic acid lactones that contribute to flavor instability of ready-to-drink coffee.

Flavor instability of ready-to-drink (RTD) coffee during storage negatively impacts product quality. Untargeted liquid chromatography/mass spectrometry (LC/MS) analysis was applied to identify chemical compounds that degraded during storage and impacted the flavor attributes of RTD coffee. LC/MS chemical profiles of non-aged and aged coffee samples were modeled against the degree of difference sensory scores by orthogonal partial least squares with good fit (R2Y = 0.966) and predictive ability (Q2 = 0.960). The top five predictive chemical features were subsequently purified by off-line multidimensional Prep-LC, revealing ten coeluting chlorogenic acid lactones (CGLs) compounds that were identified by LC/MS and nuclear magnetic resonance (NMR). The concentrations of eight CGLs significantly decreased in the coffee during the 4-month storage. Sensory recombination testing revealed the degradation of 3-O-caffeoyl-É£-quinide and 4-O-caffeoyl-É£-quinide significantly impacted the flavor stability of RTD coffee at subthreshold concentrations.

Identification of Non-Volatile Compounds Generated during Storage That Impact Flavor Stability of Ready-to-Drink Coffee.

Coffee brew flavor is known to degrade during storage. Untargeted and targeted LC/MS flavoromics analysis was applied to identify chemical compounds generated during storage that impacted the flavor stability of ready-to-drink (RTD) coffee. MS chemical profiles for sixteen RTD coffee samples stored for 0, 1, 2, and 4 months at 30 °C were modeled against the sensory degree of difference (DOD) scores by orthogonal partial least squares (OPLS) with good fit and predictive ability. Five highly predictive untargeted chemical features positively correlated to DOD were subsequently identified as 3-caffeoylquinic acid, 4-caffeoylquinic acid, 5-caffeoylquinic acid, 3-O-feruloylquinic acid, and 5-O-feruloylquinic acid. The increase in the six acidic compounds during storage was confirmed by sensory recombination tests to significantly impact the flavor stability of RTD coffee during storage. A decrease in pH, rather than an increase in total acidity, was supported to impact the coffee flavor profile.

Mechanisms of Rebaudioside A Degradation and Ingredient-Sweetener Interactions in Beverages during Storage.

The instability of rebaudioside A (Reb A) in food product applications during storage challenges their utilization. The pathways of Reb A degradation in aged acidic beverages were investigated. Three Reb A degradation compounds of known sensory importance were monitored, consisting of (1) a rearrangement, (2) a hydration, and (3) an epoxidation/rearrangement product. Using deuterium-labeled water (D2O) experiments, compounds 1-2 were reported to be generated by acid-catalyzed mechanisms involving the formation of a carbocation on carbon position 16, followed by either deprotonation via E1 elimination on C15 to form the more thermodynamically stable trisubstituted alkene (compound 1), or by the Markovnikov addition of water via SN1 substitution to form a tertiary alcohol (compound 2). Compound 3 was generated by epoxidation of the exomethylene at the C16-17 positions, followed by the opening and rearrangement of the ring to form a new alkene bond between C15-C16 and a primary alcohol on C17. Further analysis of the effect of beverage ingredients indicated the addition of caramel color significantly increased (p < 0.0001) the concentrations of compounds 1-2 compared to the aged control by 89 and 83%, respectively, whereas a specific coffee flavor and caramel color were reported to significantly reduce (p < 0.0001) the formation of compound 3 compared to the aged control during storage by 90 and 79%, respectively.

Identification of Non-Volatile Compounds That Impact Flavor Disliking of Whole Wheat Bread Made with Aged Flours.

Whole wheat flour has a shorter shelf life than refined wheat flour due to off-flavor development. An untargeted liquid chromatography/mass spectrometry (LC/MS) flavoromics approach was applied to identify compounds that negatively impact the flavor liking in whole wheat bread made from aged flours. The chemical profiles of thirteen breads made from aged flours were obtained using LC/MS and modeled by orthogonal partial least squares (OPLS) to predict flavor liking. Top predictive chemical features (negatively correlated) were identified as pinellic acid (9S,12S,13S-trihydroxy-10E-octadecenoic acid), 12,13-dihydroxy-9Z-octadecenoic acid, and 1-(9Z,12Z-octadecadienoyl)-sn-glycero-3-phosphocholine. The sensory analysis confirmed the three compounds increased the bitterness intensity of the bread samples. The formation of the trihydroxy fatty acid bitter compound, pinellic acid (9S,12S,13S-trihydroxy-10E-octadecenoic acid), was impacted by the lipoxygenase activity of the flour; however, there was no influence on the formation of 12,13-dihydroxy-9Z-octadecenoic acid or 1-(9Z,12Z-octadecadienoyl)-sn-glycero-3-phosphocholine. Additionally, the concentrations of all bitter compounds were significantly higher in bread made from aged flour versus non-aged flour.

Identification of non-volatile compounds that impact consumer liking of strawberry preserves: Untargeted LC-MS analysis.

Non-volatile compounds that impact the acceptability of strawberry preserves were investigated by untargeted LC-MS flavoromics analysis. Chemical profiles for fifteen strawberry preserves were modeled against consumer liking scores by orthogonal partial least squares (OPLS) with good fit (R2Y = 0.995) and predictive ability (Q2 = 0.918). Four chemical compounds predictive of acceptability were identified, by accurate MS and NMR, as secoisolariciresinol monoglucoside, (+)-isolariciresinol monoglucoside, 1-hexanoyl-phloroglucinol-2-O-ß-d-glucoside, and the novel compound decanoic acid-4-O-ß-d-glucoside. Sensory recombination testing of preserve samples with added levels of the four predictive LC-MS compounds indicated perceivable sensory changes in the flavor profile. Female consumers significantly preferred the recombination preserve with added levels of both predictive GC-MS and LC-MS compounds as compared to the control preserve, demonstrating the applicability of the approach for understanding product liking.

Identification of volatile compounds correlated with consumer acceptability of strawberry preserves: Untargeted GC-MS analysis.

Volatile compounds that impact the acceptability of strawberry preserves were investigated by untargeted GC-MS flavoromics analysis. Chemical profiles for fifteen strawberry preserves were modeled against consumer liking scores by orthogonal partial least squares (OPLS) with good fit (R2Y = 0.998) and predictive ability (Q2 = 0.853). Ten highly predictive compounds were selected, nine of which were identified, and eight of which were quantified using tandem mass spectrometry. Sensory recombination difference testing revealed that the addition of predictive compounds to an average-liked preserve at levels found in the most-liked preserve induced a perceptible difference to frequent consumers of strawberry jams. One of the highly predictive compounds was identified (MS and synthesis) as a novel linalool oxide product 2-methyl-2-vinyl-tetrahydrofuran with a fruity, herbal-minty, piney aroma and a low odor threshold value.

Untargeted LC-MS based identification of Rebaudioside A degradation products impacting flavor perception during storage.

Untargeted LC-MS flavoromics chemical profiling was used to identify compounds predictive of Rebaudioside A (Reb A) flavor instability in an acidified beverage after 6 weeks at 35 °C. High-quality orthogonal partial least squares analysis models were developed from the chemical data and d' values from tetrad sensory panel testing with good predictive ability. The top four highly predictive compounds were selected and identified as Reb A (negatively correlated) and three Reb A degradation compounds (positively correlated), which included a rearrangement, hydration, and an epoxidation/rearrangement of Reb A, termed compounds 1, 2, and 3, respectively. The concentrations of compounds 1-3 in the aged beverages were determined to be below the sensory recognition threshold values. However, sensory recombination testing of compounds 1-3 as a tertiary mixture revealed a perceivably significant flavor change that was aligned with the aged beverage.

Characterization of bitter compounds in native and hybrid American hazelnuts.

The main objective of this study was to identify compounds that contribute to the bitter taste attributes of American hazelnuts (C. americana) and derived hybrids. Sensory-guided Prep-LC fractionation analysis selected four compounds as bitter that were further identified by MS and NMR as carpinontriol B, 2-(3-hydroxy-2-oxoindolin-3-yl)acetic acid 3-O-6'-galactopyranosyl-2″-(2″oxoindolin-3″yl) acetate, giffonin-2-O-[(α-d-glucopyranosyl)-6'-O-(3-hydroxy-3-methylglutaric acid) and Afzelin, termed compounds 1-4, respectively. The concentrations of compounds 1, 2, 3, and 4 were determined across 88 American hybrid hazelnut genotypes and reported to be present at levels above their bitter threshold values in 94, 75, 25, and 0% of the samples, respectively. Moreover, the concentrations of compounds 1 and 3 were significantly higher for nuts dehusked prior to drying during post-harvest handling, which is expected to result in a higher perceived bitter intensity.

Identification of non-volatile compounds that negatively impact whole wheat bread flavor liking.

Untargeted LC-MS flavoromic analysis was utilized to identify chemical compounds that impact consumer liking of whole wheat bread. Chemical fingerprints of thirteen whole wheat breads were modeled against consumer flavor liking scores by orthogonal partial least squares (OPLS) with good fit (R2Y = 0.98) and predictive ability (Q2 = 0.95). The four most predictive features (negatively correlated) were identified as 9S,12S,13S-trihydroxy-octadec-10E-enoic acid (pinellic acid), 9S,12S,13S-trihydroxy-octadeca-10E,15Z-dienoic acid, 8R*,9R*,10S*-trihydroxy-octadec-6Z-enoic acid, and 1-(octadeca-9Z,12Z-dienoyl)-sn-glycero-3-phosphocholine. Sensory validation studies including bitter threshold determination and recombination tests confirmed the contribution of these compounds to the perceived bitterness intensity of the bread samples and the overall negative impact on flavor liking. Lipoxygenase activity of the flour was reported to have a significant impact on the formation of the three bitter compounds (trihydroxy fatty acids) in the bread samples.

Identification of inhibitors of pinellic acid generation in whole wheat bread.

Bitterness is a common aversive flavor attribute of foods associated with low consumer acceptance. Untargeted LC-MS flavoromic profiling was utilized to identify endogenous compounds that influence the generation of the bitter compound 9,12,13-trihydroxy-trans-10-octadecenoic acid (pinellic acid) during bread making. A diverse sample set of wheat germplasm was chemically profiled. The corresponding pinellic acid concentrations after dough formation were modeled by orthogonal partial least squares (OPLS) with good fit (R2Y = 0.8) and predictive ability (Q2 = 0.6). The most predictive feature (negatively correlated), postulated to interfere with the biosynthetic pathway, was identified as schaftoside, an apigenin di-C-glycoside. Recombination experiments involving the addition of schaftoside to flour prior to breadmaking resulted in a 26% decrease in pinellic acid formation and significantly lower perceived bitterness intensity in whole wheat bread. This work provides novel understanding of bitter generation pathways in wheat products and new strategies to improve flavor profiles and consumer acceptability.

Identification of coffee compounds that suppress bitterness of brew.

Untargeted LC-MS flavoromic profiling was utilized to identify compounds that suppress bitterness perception of coffee brew. The chemical profiles of fourteen brew samples and corresponding perceived bitterness intensities determined by descriptive sensory analysis were modeled by orthogonal partial least squares (OPLS) with good fit (R2Y > 0.9) and predictive ability (Q2 > 0.9). Ten chemical markers that were highly predictive and negatively correlated to bitter intensity were subsequently purified by multi-dimensional preparative LC-MS to conduct sensory recombination testing and/or confirm compound identifications by NMR. Three of the ten compounds evaluated, namely 4-caffeoylquinic acid, 5-caffeoylquinic acid, and 2-O-ß-d-glucopyranosyl-atractyligenin were identified as bitter modulators in coffee, and significantly decreased the perceived bitterness intensity of the brew.

Impact of temperature and water activity on the aroma composition and flavor stability of pea (Pisum sativum) protein isolates during storage.

Flavor stability of pea protein isolates (PPIs) during storage was investigated. Two commercial PPIs were stored at three water activities (0.128-0.501) under refrigerated (7 °C) and accelerated (37 °C) temperatures for 12 weeks. Eleven aroma compounds were monitored by gas chromatography-tandem mass spectrometry (GC/MS/MS) and results revealed significant changes in the aroma concentrations among the PPI samples during storage. In agreement with the chemical changes, significant differences in orthonasal aroma profiles were demonstrated using a sensory difference-from-control test. The sample stored under accelerated storage temperature (37 °C) and at the highest water activity showed the greatest degree of aroma change. An aroma recombination sensory study indicated the generation of two specific compounds, 1-octen-3-ol and nonanal, along with the degradation of 2-4-decadienal resulted in sensory changes during storage indicating lipid oxidation was the main mechanism of flavor instability in the PPI samples.