A Dislike of Broccoli Might Be Genetic (But the story isn't that simple)
「あの子が野菜を食べないのは、わがままではなく遺伝子のせいかもしれない」(in translation: "Maybe that child isn't refusing vegetables out of stubbornness — maybe it's their genes") — half of that is true. And the other half is where popular belief has run ahead of the science. This note sorts things into three tiers:the tier that genuinely holds up, the tier that has been disproven, and the tier that disappears in meta-analysis.
What this page establishes
That the 5 worldwide haplotypes formed by 3 SNPs in the TAS2R38 gene explain 55–85% of the individual difference in bitterness sensitivity to a substance called PTC has been confirmed across multiple studies of this gene locus.
The 「スーパーテイスター」 (supertaster) theory — that people with more papillae (fungiform papillae) on the tongue perceive bitterness more strongly — found no correlation with PROP bitterness ratings in a large-scale follow-up study, and the original report was not reproduced.
That TAS2R38 genotype has a significant effect on the perceived bitterness of rocket and cooked turnip has been confirmed in multiple studies, but whether that effect carries through to overall food preference itself often fails to reach significance within the same studies.
The association between bitterness sensitivity (whether measured by phenotype or genotype) and preference for, or intake of, vegetables and legumes was small and not significant in a meta-analysis pooling 36 studies.
Tier 1: Genetics really does matter (established)
The haplotypes (PAV-type and AVI-type) formed by variation at just 3 sites in the bitter taste receptor gene TAS2R38 explain55–85% of the individual difference in bitterness sensitivity to a substance called PTC. In cultured cells, the PAV-type receptor responds at micromolar concentrations, while the AVI-type stays silent even at 1000 times that concentration — a working/non-working distinction reproduced at the molecular level, one of the most solid results in taste genetics. And the family of pungent Brassica compounds (goitrin and others) really do strike this receptor.
Scope: Linkage-disequilibrium mapping in unrelated subjects (Kim 2003), genotype-stratified psychophysics (Bufe 2005), threshold phenotyping across 6 haplotypes (Behrens 2013)
A small region on chromosome 7 at 7q shows strong linkage disequilibrium between SNP markers and PTC taste sensitivity in unrelated subjects; this region contains a single TAS2R gene with 3 coding SNPs, giving rise to 5 haplotypes worldwide. These haplotypes fully explain the bimodal distribution of PTC taste sensitivity and account for 55–85% of the variance in PTC sensitivity.
Limitations: Abstract level: the sample size, the population behind the 55–85% range, and the phenotype values for each haplotype are not given in the abstract. The full text in Science is not accessible.
Suprathreshold PTC intensity functions showed a clear separation between every AVI/AVI function and every PAV/PAV function, and all diplotype groups differed significantly from one another at every PTC concentration (p < 0.001).
Limitations: The sample per diplotype group is small. Heterozygotes are intermediate in PTC intensity but resemble PAV/PAV in recognition threshold, so which phenotype is measured matters.
In 56 subjects covering the 6 TAS2R38 haplotypes, the association with detection thresholds for 4 thioamide compounds (PTC, PROP, goitrin, and methimazole) was strongly significant, at P<0.005 and β>1.85 for all 3 SNPs, and P<0.001 for the PAV/AVI haplotype. For sinigrin (only a subpopulation of N=25), all 3 SNPs were also statistically significant, but the P-values, at 0.03–0.04, were higher than for the thioamide compounds, and the authors point to insufficient power from the small sample.
Limitations: Abstract level. Haplotypes other than PAV and AVI showed no association with phenotype, which the authors attribute to their low frequency. It is used here to read the correspondence between genotype and threshold.
Context: Mean thresholds varied more than 40-fold among compounds (from about 50 µM for PTC to about 2000 µM for sinigrin), and within-compound individual variation ranged from 40-fold for salicin, goitrin, and sinigrin to 8000-fold for PTC. Verbatim: "Mean thresholds varied more than 40-fold among compounds, from a minimum of ~50 µM (PTC) to a maximum of ~2000 µM (sinigrin), whereas within-compound threshold ranges varied from 40-fold (salicin, goitrin, and sinigrin) to 8000-fold (PTC) (Figure 2)."
Limitations: The number of subjects differs by compound (sinigrin uses only a subpopulation).
Scope: Heterologous expression of the PAV and AVI variants of TAS2R38 with calcium imaging. PROP tested alongside PTC
Micromolar concentrations of PTC raised cytoplasmic calcium in cells expressing the PAV-type receptor in a concentration-dependent manner, and the receptor responded equally strongly to PROP, whereas hTAS2R38-AVI did not respond even to the high concentration of 1 mM of PTC or PROP (this sentence has been checked verbatim against the source).
Limitations: The EC50 values are in Figure 1C but are not given as numbers in the main text, so potency values are not recorded here.
Context: Places the TAS2R38 result in its contemporaneous context and generalizes the design: the authors state that, until their study, the phenylthiocarbamide receptor hT2R38 had been the only gene directly linked to variation in human bitter taste, and go on to report that polymorphisms in 2 further hT2R genes produce different receptor activity and different taste sensitivity to 3 bitter molecules.
Limitations: Does not retest the PAV and AVI variants of TAS2R38; it supports the allele–receptor-activity design rather than Bufe 2005's specific numbers. An abstract-level extraction.
Scope: Allele-specific in vitro functional assay of the 6 TAS2R38 haplotypes, confirmed by full-gene sequencing in 56 white subjects
Note: Gives a receptor-level link connecting Brassica glucosinolate chemistry to the PROP/PTC sensitivity axis. The abstract has no per-compound EC50 or threshold values, so a potency ranking is not recorded.
Genetic diversity was confirmed by full-gene sequencing in 56 white subjects, and allele-specific responses to 5 TAS2R38 agonists (PTC, PROP, goitrin, methimazole, and sinigrin) were analyzed by in vitro assay. However, the number of subjects for detection-threshold measurement differed by compound: all 56 subjects for salicin, PTC, PROP, goitrin, and methimazole, but only 25 (a subpopulation) for sinigrin. The genotype association with threshold was strongly significant for PTC/PROP/goitrin/methimazole (the thioamide compounds), at P<0.005 and β>1.85 for all 3 SNPs, whereas for sinigrin all 3 SNPs were significant but with higher P-values (0.04, 0.03, 0.03), and the authors themselves state that power is limited by the small sample of N=25. The P<0.001 for the PAV/AVI haplotype is mainly the value for the thioamide compounds.
Limitations: It should be noted explicitly that the genotype–threshold association for sinigrin is statistically weaker than for the thioamide compounds and comes from a small sample (N=25).
Verbatim: 'Functional expression studies demonstrate that five different haplotypes from the hTAS2R38 gene code for operatively distinct receptors.' The PAV receptor responded with an EC50 of 1.1 uM (PTC) and 2.1 uM (PROP), while 'hTAS2R38-AVI did not respond to PTC or PROP concentrations as high as 1 mM.' PVI, AAI, and AAV responded with EC50 values close to PAV. Concentration–intensity functions and recognition thresholds were measured in 32 subjects of known TAS2R38 diplotype, and one PAV subject's lowest discriminable concentration corresponded to the PAV receptor's in vitro EC50.
Limitations: This study did not test goitrin or sinigrin, so it independently supports the allele-dependent TAS2R38 function axis of this claim (via PTC/PROP), but does not itself support goitrin/sinigrin agonism.
Context: The full-text Discussion of Eib et al. (2020) mentions the relationship between sinigrin's bitterness and TAS2R38. Verbatim: "Sinigrin is associated with bitterness; this is a basic taste, which is predominantly perceived via the TASTE 2 receptor (TAS2R) family on the tongue (Bell et al., 2018). In particular, sinigrin binds preferably to TAS2R38 (Meyerhof, Kuhn, Chudoba, et al., 2010)." This is not Eib's own measurement but a secondary mention citing the receptor-preference finding of Meyerhof et al. 2010.
Limitations: The primary data are in the Meyerhof 2010 TAS2R source. That source currently has access_status=restricted and review_status=abstract_checked, and the sources.csv notes do not yet mention sinigrin. This line follows the pattern of 『別ソース全文で見つかった一次文献への二次引用』 (in translation: "a secondary citation, within the full text of a different source, to primary literature found there"); once the full text of Meyerhof 2010 itself becomes readable, this line should be replaced with a primary evidence line, or treated, for now, as a provisional secondary citation that needs to be supplemented.
Context: In vitro, allelic differences in EC50 for each agonist were not statistically significant, whereas the magnitude of the response (ΔF/F) within the same agonist differed greatly between alleles. PAV responded strongly to all agonists, and AVI did not respond above baseline to any of them. Verbatim: "allelic differences in EC50 within agonists were not statistically significant. In contrast, within-agonist comparisons revealed highly divergent functionality among alleles as measured by ΔF/F."
Limitations: The lack of a significant difference in EC50 may partly reflect a power issue. The correspondence with human thresholds is covered in a separate section of the main text.
Tier 2: 「舌のブツブツが多い人は敏感」 (in translation: "people with more bumps on the tongue are more sensitive") — disproven
A classic piece of folk wisdom —「舌の茸状乳頭が多い人ほど味に敏感（スーパーテイスター）」(in translation: "people with more fungiform papillae on the tongue are more sensitive to taste [supertasters]"). This was refuted by a study of n=394, with R²=0.003(essentially no explanatory power). Genotype, in the same data, had R²=0.514. In other words, a single study shows that genetics matters and papilla count does not. The extended reading of the supertaster concept as 「PROPに敏感な人は何でも強く感じる」(in translation: "someone sensitive to PROP feels everything more strongly") has also been refuted by follow-up research, including work by the original authors.
6-n-propylthiouracil — perceived bitterness is associated with: The density of fungiform papillae and taste buds on the front of the tonguecontested
Scope: Human psychophysics with papilla counts on the tongue surface. Originally reported as a significant correlation supporting the supertaster concept
Note: The correlation that had been the anatomical pillar of the supertaster concept was not reproduced in a direct n=394 verification (Garneau 2014) (R-sq = 0.003). Not significant within diplotype groups either.
The density of taste receptors on the front of the tongue (fungiform papillae, taste buds) correlates significantly with the perceived bitterness of PROP, supporting the supertaster concept. Women have more fungiform papillae and taste buds.
Limitations: At the abstract level; the correlation coefficient and n are not given. This is the founding report of the association that Garneau 2014 later failed to reproduce.
Contradicts: "We failed to find any evidence that FP associates with PROP response [R-sq = 0.003, F(1,392) = 1.23, p = 0.267]" (verbatim); FP was not predictive within diplotype groups either (AVI/AVI p = 0.311, PAV/AVI p = 0.418, PAV/PAV p = 0.447), and FP was not significant in the full model either (p = 0.306), whereas diplotype, age, and sex predicted PROP bitterness (model R-sq = 0.514); "these data argue against the use of papillae density in predicting taste sensitivity and caution against imprecise use of the term supertaster" (verbatim).
Limitations: Only PROP was tested, so this null result does not directly address the relationship between papillae and other bitter substances. The papilla-counting method differs from the Bartoshuk-era studies (no flattening of the tongue), and while the authors position this as an improvement, it remains a methodological difference between the two conflicting reports.
supertaster is identified by: A view that treats high perceived bitterness of PROP as a marker of a globally amplified oral sensory world (stronger taste from a variety of bitter and sweet substances, stronger burning sensation from alcohol and capsaicin)contested
Scope: Suprathreshold intensity scaling of PROP in humans. The generalization claim extends from PROP to other oral stimuli
Note: Hayes and Keast (2011) directly critique the generalization supported by the Bartoshuk lineage: heightened responses beyond bitterness are independent of PROP bitterness, and TAS2R38 alleles cannot explain supertasting. They propose detaching the concept from PROP as "hyperguesia".
Scaling of PROP bitterness identified a subset of tasters who rate PROP as strongly bitter (supertasters); supertasters also perceive stronger taste from various bitter and sweet substances and a stronger burning sensation from oral irritants (alcohol and capsaicin); the frequency of supertasters is higher in women.
Limitations: Abstract level: n, the scaling method, and effect sizes are not given. The generalization to other oral stimuli is the part that later became contested. An erratum (Physiol Behav 1995;58:203) exists, but its content has not been obtained.
Contradicts: The heightened responses of TT and iBT (thermal tasters, and those who report a bitter side-taste from irritants) are independent of PROP bitterness, and important insight comes from carefully distinguishing the construct of supertasting -- an amplified, broadly tuned oral sensory response -- from its historical operationalization via PROP (verbatim). The examination of molecular genetics in this context shows that TAS2R38 alleles cannot explain supertasting (verbatim). The authors propose the term hyperguesia for a broadly tuned heightened taste response.
Limitations: A review's argument, not a new dataset. The independence claim rests on cited primary studies (references 8 and 32), which were not individually obtained in this survey.
Show dataconcept-supertaster —is_identified_by→ "A view that treats high perceived bitterness of PROP as a marker of a globally amplified oral sensory world (stronger taste from a variety of bitter and sweet substances, stronger burning sensation from alcohol and capsaicin)"
Tier 3: So Does This Explain Vegetable Dislike? — The Effect Thins Out in Meta-Analysis
The crucial bridge to 「だから野菜が嫌い」 (in translation: "and that's why they dislike vegetables") turns out to be surprisingly thin. That people with higher sensitivity perceive Brassica vegetables as more bitter has been reproduced (see the cards in tiers 2–4). But once it comes to liking or intake, reports keep appearing that it was pungency rather than bitterness that separated acceptance of rocket, and that bitterness ratings differed while preference did not, and a meta-analysis of 72 studies found no significant pooled effect across all food groups in adults.「感じ方は遺伝で違う。でも、何を好きになるかはその先の話」 (in translation: "how strongly you feel it differs by genetics — but what you come to like is a separate question further down the line") — that is where things currently stand.
Scope: UK adult consumers who rated rocket (arugula) lines (genotyped n=69) and cooked turnip (n=74). TAS2R38 PAV/PAV, PAV/AVI, and AVI/AVI diplotypes
Note: The main effect of genotype on bitterness was significant in both studies, but in neither study did pairwise comparisons within individual samples separate the genotypes.
Sum-of-squares ANOVA showed that TAS2R38 genotype had a significant effect on bitterness perception (P < 0.02). While the authors note that this suggests a significant effect on bitterness perception, they also state that the ANOVA showed no significant difference between genotypes within any particular rocket line, that the diplotype effect was less pronounced than originally hypothesized, and that there was a general tendency for non-tasters to score rocket bitterness lower than medium tasters or supertasters.
Limitations: The main effect does not survive as a pairwise difference within lines. Recruitment was around a single university, with 76.7% of participants female, limiting generalizability. Other TAS2R genotypes and fungiform papillae density were not measured, and the authors list these as factors that could contribute.
TAS2R38 genotype had a significant effect on bitterness perception (F(2,256) = 4.14, p = 0.02); across all samples, PAV/PAV consumers tended to rate bitterness intensity higher than PAV/AVI (p = 0.07) and AVI/AVI consumers (p = 0.05) (27.6 versus 16.9 and 16.4). There was no significant difference in bitterness intensity scores between PAV/AVI and AVI/AVI consumers (p = 0.99). It was only for stir-fried turnip that PAV/PAV consumers rated bitterness intensity significantly higher than PAV/AVI (p = 0.02) and AVI/AVI (p = 0.001) consumers.
Limitations: Pairwise contrasts between PAV/PAV and the other diplotypes remained at only p = 0.05 and p = 0.07 across all samples, and significant separation held for only 1 cooking method. 7 consumers with rare genotypes were excluded from the analysis. The authors themselves state that the sample is small and UK-only.
Scope: Consumer liking and acceptance of Brassica vegetables as a function of TAS2R38 diplotype
Note: This review-level framing was directly tested in 2 UK consumer studies and was not supported: the genotype effect on perceived bitterness does not carry through to liking. For rocket, the main basis for acceptance was not bitterness but perceived hotness (pungency).
Contradicts: The authors conclude that the hypothesis that all consumers reject bitter cultivars is not fully supported by the data presented, and the same holds even accounting for human TAS2R38 diplotype; the main basis for acceptance/rejection seems in fact to relate more to the leaves' perceived hotness than to bitterness. The effect of consumer genotype on taste preference was significant in sum-of-squares ANOVA (P < 0.004), but the pairwise comparison scores accounting for interaction with sample were not significant. An effect of TAS2R38 diplotype on liking was observed, but this classification alone was not an accurate predictor of whether an individual would like or dislike Brassica-type vegetables. Diplotype frequency did not differ significantly between the discriminating cluster and the indifferent cluster (C1 P = 0.918; C2 P = 0.564; between clusters P = 0.919). The correlation between taste preference and bitterness perception across the whole cohort was significant but weak (r = -0.227, P < 0.0001).
Limitations: Rocket is a Brassica but not Brassica oleracea, so extension to cabbage-type crops is an inference. The genotype's main effect is significant in the omnibus test and fails only in pairwise comparisons, so this is a failure to localize the effect rather than a demonstrated null. The correlation between bitterness and liking is weak but in the expected negative direction.
Contradicts: Although TAS2R38 genotype had a significant effect on bitterness perception, this bitterness genotype had no significant effect on either taste liking (F(2,70) = 0.99, p = 0.38) or overall liking (F(2,70) = 1.05, p = 0.35) of cooked turnip, and the interaction between cooking method and TAS2R38 was not significant for either taste liking (F(6,210) = 0.77, p = 0.59) or overall liking (F(6,210) = 1.38, p = 0.23). The ratio of tasters to non-tasters was similar across all 3 liking clusters (63/37, 67/33, 67/33 percent), and the authors concluded that the separation of these clusters was not driven by bitterness genotype.
Limitations: This is a null result in a sample of 74 with 3 genotype groups, and the authors themselves explicitly note that it is small and UK-limited. The absence of a significant effect is not evidence of the absence of an effect. Even combining bitterness and sweetness perception explains only 15.7% of the variance in taste liking (adjusted R2 = 0.157), and most of the variance in liking is explained by neither taste attribute.
Context: This review states that the main sensory attributes involved in these vegetables are a characteristic sharp bitterness and a distinctive aroma, that these characteristics often lead consumers to reject this group of vegetables, and that because some people do not perceive the unpleasant sensation, individual variation in bitterness perception and sensitivity may plausibly be involved.
Limitations: Abstract-level and narrative. Included only to record the framing that the primary consumer studies go on to test; it reports no quantitative association between genotype and acceptance, and is not treated as support for this claim.
Contradicts: The pooled association between bitterness sensitivity in adults (as the PROP/PTC phenotype or the TAS2R38 genotype) and vegetable and legume liking and intake was small and non-significant across the 36 studies meta-analyzed, and the authors conclude that bitterness sensitivity explains little of the variation in adult dietary patterns.
Limitations: Because the vegetable and legume category is broader than Brassica, this is not a direct test of Brassica but a population-level counter-evidence to the general claim. Studies of children and adolescents, where genotype effects are often reported to be stronger, are excluded by design.
TAS2R38 is expressed outside the tongue too — in the epithelium of the nasal sinuses — where it responds not to food but to bacterial "conversation" molecules (quorum-sensing compounds), triggering a defense response. A bitter taste receptor turns out not to be「まずいものセンサー」 (in translation: "a bad-taste sensor") but part of a much broader chemical surveillance system — this is one more reason why NIGAMI LAB does not treat bitterness as a single, monolithic thing.
TAS2R38Bacterial acyl-homoserine lactone quorum-sensing molecules (C4HSL 200 uM, C12HSL 100 uM) in the human sinonasal epithelium. Drives calcium-dependent nitric oxide production, an increase in ciliary beat frequency, and direct bacterial killing is activated by supported
Scope: Polarized primary human sinonasal epithelial cultures genotyped for the TAS2R38 PAV/AVI haplotype. In killing of Pseudomonas aeruginosa strain PAO1, surviving bacteria were 14% +/- 1% for PAV/PAV, 80% +/- 5% for PAV/AVI, and 70% +/- 5% for AVI/AVI
Note: The clearest published case in which a candidate natural ligand for an extra-oral TAS2R is not a food-derived bitter compound but a bacterial signaling molecule. The authors point to the possibility that T2R38 has multiple binding sites, and do not claim that AHL is its only ligand.
Verbatim: "Cultures were stimulated with the gram-negative quorum-sensing molecule N-butyryl-L-homoserine lactone (C4HSL) at 200 uM, or N-3-oxo-dodecanoyl-L-homoserine lactone (C12HSL) at 100 uM." Bacterial killing was genotype-dependent. Verbatim: "Percentage of green (viable) PAO1 after exposure to saline (no epithelial cells; negative control; 97% +/- 2%), colistin (no epithelial cells; positive control; 4% +/- 2%), PAV/PAV cultures (14% +/- 1%), PAV/AVI cultures (80% +/- 5%), AVI/AVI cultures (70% +/- 5%)." Killing required nitric oxide, and was blocked by L-NAME pretreatment or by adding cPTIO.
Limitations: An in vitro human primary culture; the authors state that T2R38 may have multiple binding sites, and do not claim that AHL is its only ligand.
Context: Connects Lee 2012's in vitro genotype-dependent defense result to a clinical outcome. Among 123 patients with chronic sinusitis (82 with nasal polyps, 41 without), TAS2R38 genotype correlated with surgical outcome only in the group without polyps: PAV/PAV homozygotes (n = 7) improved by a mean of 38 +/- 21 points on the SNOT-22, versus 12 +/- 22 points for heterozygotes and non-functional homozygotes (n = 34) (p = 0.006). In a multivariate regression incorporating scores at 1 month and 3 months (n = 207), the mean improvement for PAV/PAV patients without polyps was 20 points greater (95% CI 8 to 32; p = 0.001). Among patients with nasal polyps, mean improvement was roughly the same across genotype groups.
Limitations: An observational outcomes study, not a test of the acyl-homoserine-lactone mechanism itself, so it cannot show that AHL sensing is the causal link. The decisive subgroup is small (7 of 41 patients without polyps were PAV/PAV), and the authors list sample size, follow-up duration, unmeasured sinus microbiology, and variation in surgical technique between operators as limitations. No effect was found in the larger group of patients with polyps.
Show datareceptor-tas2r38 —is_activated_by→ "Bacterial acyl-homoserine lactone quorum-sensing molecules (C4HSL 200 uM, C12HSL 100 uM) in the human sinonasal epithelium. Drives calcium-dependent nitric oxide production, an increase in ciliary beat frequency, and direct bacterial killing"
The key statistics in this note (R²=0.003 / 0.514, variance of 55–85%, and others) have been checked against the original text in PMC full text. Some sources are still at the abstract-checked stage (see each card's verification-status display).
The pooled effect in the meta-analysis has high heterogeneity (I²=82%), and the accurate reading is not 「効果がない」 (in translation: "there is no effect") but rather 「集団平均では検出されない」 (in translation: "not detected at the population-average level").
This cannot be used to make a determination about an individual child's eating behavior based on genetics.
Frequently asked
Is a dislike of vegetables genetic?
That the strength of bitterness perception itself is explained to a considerable degree by the TAS2R38 gene type has been confirmed across multiple studies. However,「苦く感じること」 (in translation: "tasting it as bitter") and「嫌って食べないこと」 (in translation: "disliking it and not eating it") are separate questions, and whether genotype actually moves real-world preference for, or intake of, vegetables and legumes was small and not significant in a meta-analysis pooling 36 studies. In individual studies using rocket and cooked turnip, too, genotype had a significant effect on bitterness-intensity ratings, while liking/disliking ratings often showed no significant difference. In short, the genetics of sensitivity and the behavior of disliking and not eating something need to be considered separately.
Is bitterness sensitivity determined by the number of bumps (papillae) on the tongue?
One report showed a correlation in which people with more papillae (fungiform papillae) or taste buds at the front of the tongue rated a substance called PROP as more strongly bitter, and this became the origin of the term 「スーパーテイスター」 (supertaster). But afterward, a large-scale follow-up study that recounted papillae found no such correlation, and papilla count did not predict PROP response even when broken down by genotype. A methodological difference has also been pointed out — the papilla-counting method itself differs between studies (for example, whether the tongue is flattened first) — so it remains possible that the discrepancy between the two reports comes from a difference in counting method. At present, it has not been confirmed that papilla count alone can explain bitterness perception.
What is not known
How far bitterness-sensitivity genotype actually connects to the behavior of 「嫌って食べない」 (in translation: "disliking and not eating") vegetables and legumes cannot be confirmed from this material alone, because the studies that measure sensitivity strength and the studies that measure not-eating behavior are separate.
How bitter taste receptor genes other than TAS2R38, or the density of papillae and taste buds on the tongue, affect bitterness sensitivity has not yet been measured — the authors themselves state explicitly that they 「測定していない」 (in translation: "did not measure") this.
Whether the supertaster concept can be extended to the diverse bitter substances beyond PROP, or to oral sensations in general such as pungency and burning heat, remains at the level of review discussion and has not been confirmed by independent primary data.
It is known that receptor activity strength (EC50) differs by TAS2R38 genotype, but how far that difference explains actual differences in human bitterness intensity has not yet been confirmed, since some studies found no significant difference.