Metabolic & Cardiometabolic · Gut & Microbiome
food-derived peptides inhibits DPP-4
In plain terms: Can food proteins block DPP-4, the enzyme that breaks down the blood-sugar hormone GLP-1?
Strong support Metabolic & Cardiometabolic 🐭 Non-human evidence💰 Industry COI noted
RefutedContestedStrong support
consensus score 1.00
Yes, but only shown in the lab and animals so far, and not yet proven in bread or people.
📅 Last reviewed: 2026-07-15 ⓘ
Evidence ladder
How far up the ladder this claim has climbed. A high consensus on a low rung means "consistent so far," not "proven in people."
Top evidence so far: Animal studies (Animal)
MechanismIn-vitroAnimalObservationalRCTMeta-analysis
How the studies fall
26 support 0 contradict 0 tested null 0 mixed · 26 sources, 26 independent groups
What the evidence shows
Peptides from whey, cereal (Coix) and other plant proteins inhibit DPP-4 in vitro and raise active GLP-1 in mice. The bread-relevant ones required targeted enzymatic hydrolysis rather than LAB fermentation, and their release in bread, gut survival and absorption remain unproven. Potency is the thing to hold on to.
The evidence (32)
| Source | Grade | Stance | Quality | Finding |
|---|---|---|---|---|
| Wang, Huang, Marnila, Hiidenhovi, Valimaa, Granato, Makinen 2024 · Food Res Int | animal | supports | moderate | Read from full CC-BY text (Luke/jukuri repository; not in PMC). Real but weak, and mostly predicted. Only 4 of the 166 proposed peptides were ever wet-assayed against purified DPP-4: 'peptides PPVEEP (score 456.6) and GPAGDPA (score 321.8) were selected from BHBH and synthetized, which had in vitro DPP-4 inhibitory activity IC50 value 936 uM and 1049 uM, respectively', against the paper's own benchmark that 'Potent DPP-4 inhibitory peptides have IC50 less than 100 uM'. The other two failed outright: 'peptide HyP-HyP-GRPGF (score 355.3) and GADPEDVIVS (score 267.9) did not have in vitro DPP-4 inhibitory activity or even being a substrate for DPP-4 enzymes' — and one of those was ABOVE the in-silico threshold, so prediction did not track inhibition. The abstract's '86 peptides from BHBH and 80 peptides from BHH' are in-silico filter output, not measurements; the previous extract ('predicted DPP-4 inhibitors') was quoting that. In vivo the exposure was the whole hydrolysate, never an isolated peptide: 'forty male C57BL/6J mice were divided into five experimental groups (n = 8/group)' on high-fat/high-fructose diet for 16 weeks, with sitagliptin as positive control, no randomisation or blinding described. The on-claim result is high-dose-only — 'Although LD-BHH treatment did not affect DPP-4 activity in the plasma of mice fed a high-fat, high-fructose diet, HD-BHH treatment significantly decreased DPP-4 activity' — as is the GLP-1 rise (HD-BHH 17.47 pM vs HFD 14.84 pM, p<0.05), since 'the levels of active GLP-1 in the mice from the HFD and LD-BHH groups exhibited no statistically significant differences (p > 0.05)'. Glucose tolerance did improve at both doses ('both LD-BHH and HD-BHH were able to decrease (p < 0.05) glucose levels compared to the HFD group'), but the abstract's companion phrase 'insulin tolerance improved' is not supported by the results text: ITT 'effects were not statistically significant compared to the HFD group'. Body weight was a reported null: 'BHH was ineffective in preventing body weight gain'. Supports the property claim on measured enzyme inhibition; carries no potency and no isolated-peptide in-vivo evidence. Caution: this paper's Conclusion section contradicts its own Results ('While plasma-like DPP-4 levels increased...') and must not be quoted. |
| Takemori K, Nakamura Y, Sato K, Shiratsuchi E, Kometani T, Masuda S 2026 · Nutrients 2026;18(11):1759 (published version of this preprint) | animal | mixed | moderate | Bonito elastin peptides were assayed for DPP-4 inhibition at the concentrations they actually reach in rat plasma: 'The observed inhibitory activities were as follows: GOG (0.25–0.3 µM), 8.73%; VP (0.02–0.04 µM), 11.1%; PG (2 µM), 10.4–11.4%; cGP (1.5 µM), 10.8%; and cPV (0.1 µM), 0.67%.' IC50 was 163.2 µM (GOG) and 10.68 µM (VP) against sitagliptin's 9.5 × 10−10 µM, and could not be determined at all for PG, cGP and cPV — so the property is real but 300–600× above achieved exposure. In vivo the peptides did raise GLP-1 and early insulin and improve glucose tolerance (n = 6/group), and lowered glomerular and leukocyte Dpp4 mRNA (n = 9 vs 10, 4 weeks) — but that is transcript, not enzyme activity, and the authors state they 'did not quantify active DPP4 protein in renal tissue or peripheral blood' and that 'the direct pathway of EPs → DPP4 inhibition → higher GLP-1 levels remain unvalidated', concluding the effects occur '[r]ather than acting through strong DPP4 inhibition or GLP-1 receptor agonism'. Hence mixed: inhibition exists, but this study argues against it being the operative route. DUPLICATE at re-read #469: this note is the preprints.org version of the SAME study now published as Nutrients 2026;18(11):1759 (PMID 42280402, PMC13258851), which the vault also holds as s42280402 with its own appraisal on this same claim. One study, two votes. Zero-weighted here; the published row is the live one. Peer review changed the title's 'Weak' DPP-IV inhibition to 'Modest' and left the numbers unchanged. |
| Ojha P et al 2025 · Food Chemistry: X | in-vitro | supports | low | NARRATIVE REVIEW, no primary data — zero-weight per CONVENTIONS §5. Full text (PMC12686732) read: JATS article-type 'review-article'; the authors state 'This review focuses on developing innovative strategies to incorporate these peptides into dietary interventions for managing T2DM.' It runs no DPP-4 assay, no docking and no in-silico screen of its own; Table 1 COMPILES other groups' measured IC50s ('Recent studies on the enzyme inhibition properties of IVSD-derived peptides are presented in' Table 1) and Table 2 compiles their docking energies. ISOLATED-PEPTIDE potency it carries: goat-milk GPFPLL 130.68 µM, LPYPY 179.52 µM (body text: 'was 179 and 474 for the peptides LPYPY and RPWR, respectively'), i.e. ~4 orders of magnitude weaker than sitagliptin at ~19 nM. HYDROLYSATE rows are mass units (bovine casein 0.62–2.08 mg/mL, camel casein 0.66–1.75, pea isolate 1.37, rapeseed 1.01, sunflower 0.89 mg/mL) and carry DIRECTION ONLY, not potency. The goat-milk/whey/casein primaries it restates are already appraised on this claim, so weighting it double-counts by construction. Its own translational verdict is cautious: 'However, no clinical studies have specifically evaluated IVSD-derived peptides for the treatment of T2DM' and 'For instance, peptide WR with DPP-IV exhibits poor absorption properties'. Funding: 'This study received no funding'; competing interests declared none, though one co-author is affiliated to Zhangjiakou Yongchangyuan Kernel Food Co., Ltd. |
| Zhang et al. 2023 · J Agric Food Chem | in-vitro | supports | moderate | Purified-enzyme assay: stepwise papain+alcalase hydrolysate of coix prolamins gave '97% inhibition' of DPP-IV vs single enzymolysis; three isolated peptides LPFYPN, TFFPQ, ATFFPQ had 'IC50 = 70.24, 176.87, 268.31 uM'. IN-VITRO only, weak potency vs gliptins; required targeted proteolysis, not fermentation. |
| Patlan-Velázquez LF et al 2026 · Food Science & Nutrition | in-vitro | supports | low | Peptide-rich <10 kDa extract of co-culture-fermented Requeson cheese (a crude mixture, not isolated or synthesised peptides) gave 14.2% DPP-IV inhibition, significantly above the monocultures (p<0.05). The extract was reconstituted at 15 mg/mL and 100 uL entered a 400 uL reaction (~3.75 mg/mL in-assay, arithmetic from the methods), so this is a single-mass-dose percentage, not an IC50 and not a molar potency; it cannot be placed on the claim's potency ladder. The six sequences labelled 'DPP-IV inhibitor' are UniDL4BioPep deep-learning predictions (probability >=0.5), not measured inhibitors, and the authors state chemical synthesis would be needed to validate them. In-vitro only, triplicate, ANOVA/Tukey. |
| Zhao X et al 2026 · Food Chem | in-vitro | supports | low | MEASURED DPP-IV inhibition (ladder rung 1), but of a crude fraction, not an isolated peptide: 'YTPH1 and YTPH2 were isolated using G-15 gel fractionation, showing considerable inhibitory activities against α-amylase (AAM) and dipeptidyl peptidase IV (DPP-IV), respectively' — so the DPP-IV entity is YTPH2, a Sephadex G-15 fraction of an alkaline-protease hydrolysate of yellowfin tuna trimmings (genuinely food-derived; YTPH1 is the α-amylase fraction and does not bear on this claim). The abstract reports the result qualitatively, so this row carries DIRECTION ONLY: no IC50 or % inhibition is available at abstract grade and it cannot be placed against the isolated-peptide span of roughly IC50 10 µM to ~1 mM, let alone sitagliptin at ~19 nM. The three named peptides sit a rung lower on the authors' own wording: 'In silico analysis identified three bioactive peptides: GEVE, LDDVEE and VSGAELE. These peptides potentially regulate blood glucose levels through targeted inhibition of AAM or DPP-IV' — an in-silico prediction with the hedge 'potentially', not a measurement. REGRADED animal → in-vitro because the grade must follow the arm that tests THIS claim: the in-vivo arm measured other endpoints entirely — 'YTPH noticeably reduced random and fasting blood glucose levels, increased serum insulin levels, enhanced antioxidant enzyme activities, and alleviated liver and kidney damage' — and never DPP-4 activity, so this is an in-vitro DPP-4 source. Quality low: crude-fraction entity, qualitative readout, in-silico attribution for the named peptides, abstract-grade text. Not a review — the lab runs its own hydrolysis, fractionation and assays. COI declared none; all-academic affiliations (Chinese Academy of Sciences, South China Sea Institute of Oceanology; Zhongkai University); funding not seen. Food Chem 500:147497, closed access — abstract-grade appraisal. |
| Zheng Q, Kang Q, Zhang H, Tu L, Yu Y, Yang J, Tang Y 2026 · Biosci Biotechnol Biochem | in-vitro | supports | moderate | ABSTRACT ONLY - genuinely closed; nine routes exhausted (local fulltext-cache and paper-drops absent, no PMCID, NCBI elink pubmed->pmc empty, NCBI BioC 'No result can be found', Europe PMC 'Subscription required', Unpaywall oa_status 'closed', OpenAlex any_repository_has_fulltext false, Semantic Scholar isOpenAccess false). LADDER RUNG 1 - MEASURED, not docked: 'The most active fraction (F1-2) had IC50 values of 0.0910 mg/mL (DPP-IV) and 0.0189 mg/mL (ACE).' That IC50 belongs to a chromatographic FRACTION of a Bacillus subtilis-protease hydrolysate of Lentinula edodes (shiitake) stem protein - a mixture - so it has no molar value. The three named peptides were selected with the aid of sequence DATABASES, then chemically synthesized, and are reported only as 'Following the sequencing of fraction F1-2, screening, and solid-phase synthesis, three novel peptides (EW-1, WR-2, and LP-3) exhibiting >60% inhibition of both DPP-IV and ACE at 1 mg/mL' - a percent at a stated but very high MASS concentration, with no peptide IC50, sequence or MW in the abstract, so this row yields no uM figure and cannot be compared with sitagliptin (~19 nM); do NOT enter it on the claim's potency ladder. Tension to resolve against the full text: the 'excavated' peptides are quoted at 1 mg/mL, ~11x the parent fraction's own DPP-IV IC50, so nothing here shows they beat the crude fraction they came from. Positive control and its value, replicate n, enzyme source and substrate are all cannot-tell from the abstract. The cell endpoint measured was Caco-2 cytotoxicity - 'These peptides showed no significant toxicity to Caco-2 cells' - a viability screen in one immortalised line; the previously stored 'Caco-2 safety confirmed' adopted the authors' gloss 'indicating safety' and is corrected. No in-vivo arm is reported in the abstract. Funding: Key R&D Program of Guangdong Province (2023B0202030002, 2021B0707010002) and the MARA Key Laboratory of Functional Food; no COI statement visible in the PubMed record. |
| Fanzaga M, d'Adduzio L, Bollati C, Musco MS, Boschin G, Aiello G, Lammi C 2026 · Biomedicines 2026;14(3):589 | in-vitro | supports | low | WHOLE HYDROLYSATE, NOT ISOLATED PEPTIDES, and MEASURED not docked. Test articles are two commercial bulk collagen hydrolysates — CH ('CollaSel TRIPEPTIDE') and an unnamed benchmark BCH — dosed as mixtures; no isolated or synthesised peptide was assayed and no IC50/Ki is reported, so this row carries DIRECTION only and places nowhere on the claim's per-molecule potency ladder. Purified fluorogenic DPP-IV assay: 'CH peptides inhibited DPP-IV by 29.87 ± 0.45%, 46.61 ± 0.45%, 79.74 ± 0.01%, 85.61 ± 0.98%, and 84.06 ± 1.61%' at 0.5, 1.0, 5.0, 10.0 and 20.0 mg/mL — every percentage is a MILLIGRAM-PER-MILLILITRE mixture concentration; BCH ran 15.14% to 89.82% across the same range and was numerically higher at 20 mg/mL. The cell-based result is an order of magnitude smaller: at the identical 10 mg/mL in Caco-2, 'a reduction of 8.12 ± 4.25%' (SD ~half the mean), the authors noting that 'this activity was reduced when tested in situ on Caco-2 cells'. STC-1 GLP-1 rose 16.9 ± 5.88% at 10 mg/mL (BCH +11 ± 2.8%, ns). n = 'three determinations performed in triplicate'. Tripeptide content is cannot-tell from this paper: the peptidomics retained 'only peptides of ≥4 amino acids ... for protein inference'. QUALITY LOW: 'The research was funded by Sel Sanayi Ürünleri Ticaret ve Pazarlama A.Ş.', the same company that supplied the winning product ('CH was provided by Sel Sanayi...'), while 'The authors declare no conflicts of interest.'; Results 3.2 also states 'CH peptides are statistically more active than BCH peptides, suggesting a higher hypoglycemic effect of BCH', naming the wrong product against its own figure, and Methods lists four concentrations where Results reports five. |
| Amigo-Benavent M, Rivero-Pino F, Villanueva-Lazo A, Montserrat-de la Paz S, FitzGerald RJ, Millan-Linares MC 2026 · Food Funct | in-vitro | supports | moderate | Cell-free DPP-IV assay, Diprotin A 5 uM positive control, n=3 per IC50. HYDROLYSATES: IC50 1.59 mg/mL after Alcalase+Flavourzyme vs 2.29-2.54 mg/mL Alcalase alone; the unhydrolysed hemp protein isolate was inactive (>5 mg/mL). ISOLATED PEPTIDES (chemically synthesized, >95% pure): of the two selected by the in-silico screen, TNGPQLIH gave IC50 1.70 mg/mL (MW 878 Da, i.e. ~1.9 mM - roughly 100,000-fold weaker than sitagliptin at ~19 nM) while GKLDLVKPQ gave IC50 >5 mg/mL (~>5 mM), which the authors call 'a disconnect between in silico prediction and in vitro validation'. Post-INFOGEST digestion the HYDROLYSATE IC50s were 1.53-2.19 mg/mL (conserved or improved) and the isolate itself became active at 1.74 +/- 0.03 mg/mL; the synthetic peptides were tracked through digestion by LC-MS detection and their reported IC50s are pre-digestion. Every measurement is cell-free enzyme assay - no cell, animal or human arm. Public funding (PAIDI P20_00661, ERDF US-1381492, CYTED Pro-Alt); authors declare no conflict of interest. |
| Liu 2026 · J Dairy Sci | in-vitro | supports | moderate | Goat milk-derived peptides MMSF, MPFPK, GPFPIL competitively inhibited DPP-4 (IC50 0.32-2.83 mM) in vitro. |
| Gan J, Chen J, Shen X 2026 · Food Chemistry | in-vitro | supports | moderate | Two synthesized quinoa-derived pentapeptides (WLAFR, LLPFR) competitively inhibited DPP-IV in vitro with IC50 ~78µM and ~130µM respectively; kinetics/spectroscopy plus docking elucidated binding mechanism, no in-vivo testing. |
| Fleury 2022 · Int J Mol Sci | animal | supports | moderate | Rat gavage (n=8/group, randomised allocation) plus INFOGEST digestion and Caco-2 models. In vivo, 4 of the 5 proteins dosed lowered plasma DPP-IV at 30 min: 'Pea proteins were the most efficient, and decreased DPP-IV activity by 20.2% (p<0.0001)', ovalbumin 9.2%, fish gelatin 8.2%, hemoglobin 7.4%; caseins (slow-digesting) did not. In vitro all seven digested proteins inhibited DPP-IV vs the blank-digestion control (IC50 1.5-10 mg/mL). Only five went in vivo — whey and gluten were dropped 'because their solubility was too low for force-feeding', and pea alone was pepsin pre-hydrolysed, which the authors offer as the reason it led. Critically, 'no correlation was found between the in vivo and in vitro results'. Enzyme activity only; no glucose, insulin or GLP-1 measured. |
| Xu et al. 2019 · J Agric Food Chem | in-vitro | supports | moderate | Rapeseed napin hydrolysate IC50 0.68mg/mL; peptides dock DPP-IV. IN-VITRO only |
| Takemori 2026 · Nutrients | in-vitro | supports | moderate | Direct cell-free DPP-4 enzyme assay on the five EP degradation peptides recovered from rat portal plasma: 'The DPP4 inhibitory activity of each peptide was evaluated at concentrations corresponding to their plasma levels.' Percentages are reported with their doses - GOG 8.73% at 0.25-0.3 uM, VP 11.1% at 0.02-0.04 uM, PG 10.4-11.4% at 2 uM, cGP 10.8% at 1.5 uM, cPV 0.67% at 0.1 uM. Molar IC50s, directly comparable with this claim's ladder: sitagliptin 9.5e-10 uM, GOG 163.2 uM, VP 10.68 uM; 'values could not be determined for PG, cGP, and cPV because of their low inhibitory activity.' VP sits near the potent end of the claim's own 6 uM-13 mM span - 'notably, the VP peptide identified herein exhibited stronger inhibitory activity than previously reported sequences' - while remaining ~10^10-fold weaker than sitagliptin on a molar basis. STANCE mixed -> supports at re-read (2026-08-23), reversing re-read #469. Grading the arm that bears on this claim: this is a claim_class:property row about whether food-derived peptides inhibit DPP-4, and the in-vitro assay measures exactly that, affirmatively. The paper's 'Despite Modest DPP4 Inhibition' framing and its admission that 'the present study does not establish a definitive causal link between this modest DPP4 inhibition and the subsequent in vivo metabolic improvements' concern whether inhibition EXPLAINS the animal glucose and renal outcomes - a mechanism-to-outcome question this claim does not ask, and the reason #469 landed on mixed. The claim's own scope note already stipulates the potency gap. Honest counterweight, held in this text rather than in the stance: three of the five peptides were too weakly active for an IC50 to be fitted, and the orally dosed material is a <1000 Da hydrolysate, not an isolated peptide. The animal arms (OGTT, nephrosclerosis, Dpp4 mRNA) are recorded as separate candidates and are NOT counted toward this property row. |
| An, Yang, Zhang, Suo, Song 2025 · J Dairy Sci | in-vitro | supports | moderate | PRIMARY in-vitro bench study, not a review and not an animal study (the MeSH 'Animals' term is the milk's source species). Flavorzyme hydrolysis of buffalo casein raised DPP-4 inhibition with hydrolysis time - 'the DPP-4 inhibitory effect of buffalo casein was found to increase with longer hydrolysis times' - peaking at the 3-h maximum degree of hydrolysis (20.05 +/- 0.14%): 'The hydrolysate obtained after 3 h of hydrolysis demonstrated the highest level of inhibition, with a half-maximal inhibitory concentration (IC50) value of 1.04 mg/mL.' The abstract names ONE sequenced peptide (whether others were assayed cannot be told from an abstract): 'The DPP-4 inhibitory peptide YPFPGPIPN, with an IC50 value of 0.88 mg/mL, was identified in the 1 to 3 kDa fraction of the 3-h hydrolysate', the abstract describing its active-site interactions 'via hydrogen bonds, hydrophobic interactions, salt bridges, and pi-cation interactions'. POTENCY, which the previous extract omitted entirely: the paper reports mass units only; at ~1,001 Da for YPFPGPIPN, 0.88 mg/mL is ~880 uM (conversion mine, not the paper's), i.e. roughly 4.6 x 10^4-fold weaker than sitagliptin at ~19 nM, and at the weak end even of this claim's own peptide range. The 1.04 mg/mL figure is a crude hydrolysate, so its per-molecule potency is worse again. WHAT IT DOES NOT SHOW: the endpoint is bare enzyme activity in a cuvette - no cells, no animals, no humans, no digestion or absorption step, and no glucose, insulin or GLP-1 measured; inhibition is measured rather than merely docked, which is why this outranks the cluster's docking-only rows, A positive control (sitagliptin/diprotin A), replicate counts and dispersion are NOT VISIBLE in the abstract - which is not the same as absent from the study. Assay rigour cannot be audited from the text obtainable here, and this row's provenance is abstract by necessity: full text is unreachable after seven routes. Note the named in-vivo disconfirmer in this same cluster: s35955493 gavaged rats and found 'no correlation was found between the in vivo and in vitro results', with caseins specifically failing to lower plasma DPP-IV. QUALITY moderate, and honestly labelled: read from the PubMed abstract only - this is a gold-OA CC-BY paper with no PMC record whose sole full-text location, the publisher PDF, returns 403 behind Cloudflare - so assay rigour could not be audited. Single Southwest University group (An, Yang, Zhang, Suo, Song); no funding or competing-interests statement reachable; dairy-favourable conclusion in the American Dairy Science Association's journal, noted and not acted on. |
| Laosam P et al 2026 · Foods | in-vitro | supports | moderate | Mixtures only, no isolated peptide: goat-blood hydrolysate HBN-8 gave 81.13% DPP-IV inhibition and its ≤3 kDa fraction 87.8 ± 1.3%, both at a standardized 1 mg Leu eq./mL; best SEC sub-fraction F3 IC50 0.89 ± 0.09 mg Leu eq./mL — mass-based, not molar, so it does not convert onto a molar potency ladder. LC-MS/MS identified 14 peptides but 'chemical synthesis and direct enzyme inhibition assays of the individual peptides were not performed'. Separate endpoints: simulated GI digestion RAISED inhibition 34.08%→58.75% (gastric)→53.87% (intestinal); Caco-2 gave 10.47% transepithelial recovery with DPP-IV IC50 rising 11.29→15.99 mg Leu eq./mL (~58% potency retained, i.e. an order of magnitude weaker than the purified fraction). No in-vivo glycemic endpoint; SHR/T2D-model work named as future. |
| Ramirez-Lozano F et al 2025 · Foods | mechanism | supports | low | OFF-SCOPE on OUTCOME. This paper never measured and never even docked DPP-4. Its assays were α-amylase inhibition, α-glucosidase inhibition and DPPH/ABTS scavenging, and its docking targeted only those same two enzymes: 'Docking calculations were performed with AutoDock Vina (v1.2.3), and the theoretical binding energies (kcal/mol) were calculated.' The ONLY DPP-IV content in the entire paper is a sequence-motif database annotation — 'The biological potential of individual peptides was predicted using the BIOPEP-UWM database' — which surfaces as a text column in Table 1 and as a single sentence each in the abstract and Results: 'Identified bioactive peptides showed potential for inhibiting ACE and DPP-IV, as well as exhibiting antioxidant potential.' and 'Predicted biological activities included: (1) Angiotensin-Converting Enzyme (ACE) inhibition (most prevalent), (2) dipeptidyl peptidase-IV (DPP-IV) inhibition (linked to antidiabetes effects), and (3) antioxidant capacity (free radical neutralization).' A motif annotation is a weaker inference than a docking score, and an Xaa-Pro motif cannot distinguish a DPP-4 INHIBITOR from a DPP-4 SUBSTRATE. The paper reports no DPP-IV IC50, no DPP-IV percent inhibition and no DPP-IV binding energy. ATTRIBUTION of the numbers it does report: −10.0 and −11.8 kcal/mol for VNDNGEPTL are DOCKING SCORES, not inhibition — 'Notably, sequence VNDNGEPTL exhibited binding energies of −10.0 kcal/mol (α-amylase) and −11.8 kcal/mol (α-glucosidase).' — while the measured potencies belong to a CRUDE BULK HYDROLYSATE, not to any isolated peptide, and are given per mg of protein rather than molar: 'Also, the dialyzed hydrolysate demonstrated the highest enzyme inhibition, with IC 50 values of 0.78 mg/mL for α-amylase and 0.60 mg/mL for α-glucosidase.' So this source places NOWHERE on this claim's potency scale (isolated peptides roughly IC50 10 µM to ~1 mM against sitagliptin at ~19 nM): it contributes no DPP-4 datum at all. Prediction-to-assay attrition is total — 'Ninety-seven peptides were designed using MarvinSketch' and ZERO of the 97 were synthesised or assayed against any target. SUBJECT is on-topic: the peptides are genuine food-derived in-silico digests of Phaseolus vulgaris storage proteins (phaseolin, arcelin, lectin, α-amylase inhibitor), not endogenous fragments. OUTCOME is not. Compartment note for any future re-use of this source: the enzymes were 'α-amylase from porcine pancreas, α-glucosidase from Saccharomyces cerevisiae' — yeast, not mammalian intestinal enzyme — in a cell-free microplate, and the authors defer efficacy themselves, calling for 'in vivo evaluations and clinical trials to validate biological efficacy and determine the optimal dose'. Read from full open-access text (PMC12651079, 244,941 bytes in fulltext-cache/xml), not the abstract, despite the source note's 'provenance: abstract'. |
| Han B et al 2026 · Food Res Int | in-vitro | supports | moderate | ABSTRACT ONLY - the article is genuinely closed (no PMCID; NCBI elink/BioC, Europe PMC 'Subscription required', Unpaywall oa_status 'closed', OpenAlex any_repository_has_fulltext false, Semantic Scholar openAccessPdf empty; nothing in fulltext-cache or paper-drops). LADDER RUNG 1 - MEASURED. Goat milk protein was digested by simulated gastrointestinal digestion, then 'Following fractionation by ultrafiltration and preparative RP-HPLC, twelve peptides were identified from the most active fraction (F2) using UPLC-MS/MS.' ISOLATED PEPTIDE, not hydrolysate: 'Among these, IPPK exhibited the most potent DPP-IV inhibitory activity, with an IC50 value of 120.17 +/- 7.18 uM.' IPPK = Ile-Pro-Pro-Lys, average MW 453.58 Da, so 120.17 uM = 0.0545 mg/mL = 54.5 ug/mL, and ~6,325x weaker on a molar basis than sitagliptin at ~19 nM. That still makes it the most potent goat-milk-derived peptide this claim holds (cf. GPFPLL 130.68 uM, LPYPY 179.52 uM, FPL 321.5 uM, FPHFDL 465.6 uM, MMSF/MPFPK/GPFPIL 0.32-2.83 mM), and four orders of magnitude short of a gliptin, both at once. The F2 fraction is called only 'the most active fraction' with no number of its own, so the fraction-vs-purified-peptide potency check cannot be run; the positive control, the replicate n behind the +/- 7.18, and how many of the twelve peptides were actually assayed are all cannot-tell from the abstract. GRADE FOLLOWS THE ARM THAT TESTS THIS CLAIM: DPP-4 inhibition is measured in vitro, so this is graded in-vitro rather than animal. The mouse arm's measured endpoint is BLOOD GLUCOSE, and the paper's own link back to the enzyme is a hedge, not a result - 'in-vivo studies demonstrated that IPPK effectively lowered blood glucose levels in mice, which MAY BE associated with DPP-IV inhibition and subsequent increases in GLP-1 levels'; the verb 'demonstrated' governs blood glucose alone, and whether serum DPP-4 activity or GLP-1 was assayed is cannot-tell from the abstract. The prior extract dropped that hedge and stored the authors' speculation as an observed association. No effect size, dose, mouse n or glycaemic model is given for the in-vivo arm, so it carries direction only and is logged as a candidate rather than read into this claim. DOCKING IS RUNG 2, off-scope on outcome: 'Molecular docking analysis indicated that IPPK interacts with key residues (Arg125, Glu205, Glu206, Tyr547, and His740) within the DPP-IV active site' - genuine active-site residues, but only one peptide was docked, so it neither ranks anything nor validates anything, and the rung-1 assay is what carries this row. Primary experimental paper, not a review, so no double-counting of primaries. Food Research International (Elsevier) 2026; Binsong Han and Cunzheng Zhang, Jiangsu Academy of Agricultural Sciences, with Lina Zhang and Peng Zhou, Jiangnan University; funding not stated in the PubMed record (no GrantList) and OpenAlex grants null; 'The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.' INDEPENDENT of the goat-milk row s41887527 despite the shared surname - that is Weiwei Han at Jilin University, different peptides and a virtual-screening method; do not merge into group_id 'liu-han'. |
| Shen F et al 2026 · Food Funct | in-vitro | supports | moderate | Wheat-peptide-derived LPQ and LPQF (proline at position 2) 'exhibited potent dipeptidyl peptidase-IV (DPP-IV) inhibitory activity, mediated by high-affinity binding involving van der Waals forces, hydrogen bonding, and electrostatic interactions' - an in-vitro characterisation with docking. The in-vivo arms measured GLYCEMIA, not DPP-4 activity: WP 50-1000 mg/kg attenuated HFD-induced glucose intolerance and insulin resistance in mice, and LPQ/LPQF at 0.1-5 ug/mL 'normalized glycemic levels' in hyperglycemic zebrafish. Direction only: the abstract gives no IC50, so potency is not settled here. Abstract-grade (closed-access Food Funct 2026;17(1):288-303); two of ten authors hold food-industry posts (Beingmate; Hangzhou Kangyuan) and the funding statement is behind the paywall. |
| Rai & Priyadarshini 2023 · Eur J Nutr 2023 Sep;62(6):2489-2507 | animal | supports | moderate | Bovine whey protein hydrolysate (bmWPH) — a MIXTURE, 'produced using simulated in situ digestion, purified using RP-HPLC', not an isolated sequenced peptide — showed 'the dose-dependent inhibitory effect of bmWPH on the catalytic activity' of DPP-4 in a cell-free assay; no IC50 or test concentration is recoverable from the available text, so this row carries NO potency comparable with molar isolated-peptide rows. In 3T3-L1 cells it lowered adipogenic 'factors and DPP-4 protein levels, leading to a negative effect on preadipocyte' differentiation, and after 'co-administration of WPH for 20 weeks' HFD mice 'showed a marked reduction in DPP-4 levels in WAT, liver, and serum'. The weight-loss and 'increased serum and brain GLP levels' results are an ANIMAL outcome arm (mouse, n and dose in mg/kg cannot-tell from available text) and bear only weakly on any human outcome. Abstract-grade; paper is not in PMC open access. |
| Zhu Y, Zhou E, Tang Y, Li Q, Wu L 2025 · Foods | in-vitro | supports | low | Six peptides found by peptidomics of a Tenebrio molitor larval hydrolysate were chemically SYNTHESIZED (Sangon Biotech) and assayed as isolated peptides against DPP-IV: 'However, only three peptides showed inhibitory effects against DPP-IV. Among them, GR-7 demonstrated the highest inhibitory potency against DPP-IV, with an IC 50 value of 0.64 mg/mL, followed by DK-8 (IC 50 = 7.93 mg/mL) and WK-6 (IC 50 = 9.20 mg/mL).' Using the paper's own monoisotopic MWs that is GR-7 753 uM (849.4 Da), DK-8 8.80 mM (901.5 Da), WK-6 12.6 mM (728.3 Da) — 4.0e4 to 6.6e5-fold weaker than sitagliptin at ~19 nM; DK-7, FK-8 and SK-6 gave no IC50 ('-' in Table 2) over 1-5 and 1-10 mg/mL. The abstract and Conclusions both misreport this ('Six unique peptides ... demonstrated significant alpha-glucosidase and dipeptidyl-peptidase IV (DPP-IV) inhibitory effects'), so the results table governs. DPP-IV method is one sentence — 'a commercial kit (Abcam, Cambridge, UK) following the protocols provided by the manufacturer' — with no assay positive control reported (the only named control, WLRL, is a docking comparator) and no replicate n in Methods, Statistics or any legend. Docking anti-correlates with the assay in this same paper: DK-7 and FK-8 docked at deltaG <= -8.0 kcal/mol yet did not inhibit, while DK-8 docked weakest (-6.8) and did. All six raised glucose consumption in IR-HepG2 cells, but three of those six inhibit DPP-IV not at all, so that cell effect is not attributable to DPP-4 inhibition. No in-vivo data: 'Future work will focus on in vivo hypoglycemic efficacy'. |
| Yan C et al 2026 · Food Res Int | in-vitro | supports | moderate | Defatted walnut meal peptides FPAG (IC50 105.6µM) and LPSYQPTP (IC50 132.4µM) inhibited DPP-IV in vitro, retained activity after simulated GI digestion, and enhanced active GLP-1 secretion in STC-1 enteroendocrine cells. |
| Suryaningtyas ITI, Jung WK, Lee SJ, Je JY 2025 · Int Immunopharmacol | animal | supports | moderate | Read from the PubMed abstract only - this is a closed-access Elsevier paper with no green-OA deposit; seven routes (local cache, NCBI idconv, elink pubmed->pmc, NCBI BioC, Europe PMC, Unpaywall, OpenAlex/Semantic Scholar) all returned hard negatives, so assay rigour could not be audited. PRIMARY experimental work reporting its own data, not a review: paired Caco-2 assays plus a 4-week in-vivo arm. ENDPOINT CORRECTION at re-read #491: the previous extract read 'inhibited DPP-IV, raised GLP-1, improved glycemia in T2D mice', hanging all three endpoints - DPP-IV included - on the mouse arm. The paper splits them. The only DPP-IV assay the abstract locates is in vitro: 'In vitro, P1 and P2 were assessed for their ability to inhibit α-glucosidase and DPP-IV activity in Caco-2 cells, alongside glucose uptake and transporter protein expression (SGLT-1 and GLUT2).' The in-vivo arm is 'HFD/STZ-induced diabetic mice were administered P1 or P2 (1 mg/kg or 10 mg/kg) or metformin (200 mg/kg) for four weeks', and its reported result is 'Peptide treatment significantly improved glycemic control by inhibiting α-glucosidase and DPP-IV, increasing GLP-1 levels, and reducing intestinal glucose uptake' - the authors' own causal gloss over both limbs. Whether DPP-IV activity was separately assayed in the mice cannot be told from the abstract. Grade left at animal: the in-vivo arm is real and the effect is attributed to DPP-IV inhibition; but the measured enzyme endpoint behind this claim is the cell assay. POTENCY: none. No IC50 and no percent inhibition is reported for either peptide, so this row supports the property without sizing it - which matters, since this claim's note warns that a bare 'inhibits' is exactly how a hydrolysate's '~97%' once got mis-read as isolated-peptide strength. COMPARATOR: metformin, which is not a DPP-IV inhibitor, so the study carries no gliptin benchmark for the DPP-IV limb. INDEPENDENCE: single Pukyong National University lab (Suryaningtyas, Jung, Lee, Je corresponding) using its own house mussel peptides; genuinely independent of the other 31 appraisals on this claim, none of which is PKNU or mussel-derived. Not n-of-1 - group-design rodent work. Also reported and not claimed here: alpha-glucosidase inhibition, reduced intestinal glucose uptake with SGLT-1/GLUT2 modulation, suppressed NF-κB with lower IL-6/TNF-α/IL-1β, raised SOD/GPx/CAT, and preserved beta-cell insulin secretion. |
| Antony 2024 · Int J Mol Sci | mechanism | supports | low | OFF-SCOPE on both subject and outcome. SUBJECT: the previous extract called these 'hemorphin food peptides'; the paper says the opposite — 'Hemorphins are a group of endogenous bioactive peptides produced by the physiological or pathophysiological degradation of all hemoglobin chains except the alpha chain' — and no food, dietary, whey, casein or cereal protein appears anywhere in the body text. OUTCOME: nothing was inhibited. The study is purely in-silico — 'computational approaches such as protein-peptide molecular docking and extensive molecular dynamics (MD) simulations were employed to identify the binding pose and stability of peptides in the active site of DPP IV' — with no enzyme assay, IC50, cell, animal or human data. What it reports is docking score and MM-GBSA energy: 'H7 (YPWTQRF) had a GScore of -11.44 kcal/mol and an MM-GBSA binding energy of -119.64 kcal/mol', against positive control 'Diprotin A produced a docking score of -8.05 kcal/mol and MM-GBSA binding energy of -83 kcal/mol', with contacts held through triplicate 500 ns MD. The authors themselves defer the inhibition question: 'Future research on hemorphins using in vitro and in vivo analysis will shed additional insight into their underlying inhibitory properties.' Two further cautions: the screen was sequence-specific, not class-wide — 'LVV and VV additions at the N terminal lack proline, which could account for their weak DPP IV binding' — and active-site occupancy by an Xaa-Pro peptide cannot separate inhibition from cleavage, as the paper concedes of its own control: 'diprotin A can be hydrolyzed by DPP IV into Ile-Pro and Ile'. Read from full open-access text (PMC10932442), not the abstract. |
| Handayani FW et al 2026 · Plant Foods Hum Nutr | in-vitro | supports | moderate | Bitter leaf (Vernonia amygdalina) thermolysin hydrolysate yielded peptide VAYPQCV (VV7), a competitive DPP-IV inhibitor with IC50 19.8µM; in-vitro assay plus docking, remained intact after simulated GI digestion (not tested in vivo). |
| Chen W et al 2026 · Foods (MDPI) 2026;15(9):1493 | mechanism | supports | low | Entirely computational. 11 oligopeptides were released from the mushroom lectin MEP5 by in silico proteolysis and their DPP-IV inhibition PREDICTED by the StackDPPIV server plus docking into PDB 6B1E (QPPR, DGTY, DPDSH at the S2 pocket). The authors state 'this study is entirely computational' and that 'docking scores and confidence metrics cannot be directly translated into quantitative inhibition constants or in vitro enzymatic activity'; IC50 assays are described as planned. Predicted binding, no measured inhibition. |
| Mitsui 2025 · Foods | in-vitro | supports | moderate | Cell-free assay, recombinant HUMAN DPP-4 (Gly-Pro-MCA): buckwheat flour hydrolysate 60.8+/-1.6% inhibition at 10 mg/mL vs 8.3+/-1.2% for the unhydrolysed control; ODS fraction Fr.2 IC50 1.67+/-0.11 mg/mL (n=3). ISOLATED SYNTHETIC tripeptides purified from it: Ile-Pro-Trp IC50 6.2 uM (2.6 ug/mL, MW 414) and Ile-Pro-Leu 20.3 uM (6.9 ug/mL, MW 341), vs in-house diprotin A (Ile-Pro-Ile) 2.3 uM and sitagliptin ~19 nM elsewhere, i.e. ~330x and ~1070x weaker. Both peptides kept ~100% activity after pepsin/pancreatin. The rat OGTT arm measured blood glucose only, never DPP-4, and was NULL: no difference at 300 mg/kg, p=0.07 at 30 min for 600 mg/kg. |
| Mongkolnkrajang U et al 2026 · Sci Rep | in-vitro | supports | low | MEASURED (rung 1), not docked: synthesized ATHALLA (AA-7; 95.8% HPLC purity) in a cell-free Gly-Pro-p-nitroanilide assay gave 'an IC50 value of 52.63 ± 2.32 µM ... the positive control, diprotin A, which produced the stronger IC50 value of 22.4 ± 1.29 µM' — positive control quantified, 3 biological x 3 technical replicates. At the paper's own [M+H]+ 696.46 Da (MW 695.4) that is ~36.6 ug/mL, ~2.35x weaker than its own control and ~2,770x weaker than sitagliptin (~19 nM); it misses the '<50 uM = strong inhibition' bar the discussion itself imports, though the abstract still says 'strong'. HYDROLYSATE numbers are separate and mass-only: crude protein 81.55, whole BPPH 55.64, <0.65 kDa fraction 1.10 ug/mL. Three internal inconsistencies in this arm: (1) the Dixon 'Ki value of AA-7 (0.099 mM)' = 99 uM EXCEEDS the 52.63 uM IC50, impossible for the competitive inhibition the same paragraph asserts; (2) the purified peptide (~36.6 ug/mL) is ~33x weaker by mass than the <0.65 kDa fraction it was isolated from (1.10 ug/mL), so AA-7 cannot account for that fraction's activity; (3) docking again ranks backwards against the assay — 'greater binding affinity for AA-7 ... diprotin A, for which the binding energy was lower at -6.4 kcal/mol' (AA-7 -7.3) while the measured IC50s put diprotin A 2.35x ahead. Caco-2 uptake and SGLT1/GLUT2 arms are off-scope for this claim, and the paper states no cell-based DPP-IV assay was run. |
| Qin X et al 2026 · Foods (MDPI), 2026;15(9):1489 | in-vitro | supports | moderate | Sea buckthorn leaf protein was trypsin-hydrolysed; the peptide VG-11 (VAYPLDLFEEG) was then identified, solid-phase-synthesised and assayed on its own. In a triplicate cell-free DPP-IV assay against a sitagliptin positive control, VG-11 gave IC50 0.3270 +/- 0.026 mg/mL, 'equivalent to approximately 0.2611 mM' - about four orders of magnitude weaker than sitagliptin's (2.633 +/- 0.586) x 10^-5 mg/mL. The mixture is a separate, mass-based number and is not this peptide's potency: the SLPPs-T hydrolysate read IC50 0.1286 +/- 0.012 mg/mL. The second peptide from the same hydrolysate, PM-8, was inactive - 'an inhibition rate of merely 2.98% at the maximum tested concentration of 5 mg/mL'. Docking was a screening step that predicted binding only, and the paper reports it mispredicting its own alpha-glucosidase result: 'differences in binding energy do not equate to equivalent differences in in vitro activity.' Dual-target paper - the alpha-glucosidase figure (0.3885 +/- 0.015 mM) is kept separate. No cell, animal or human data. |
| Khaliq M et al 2026 · Annals of Medicine | observational | supports | low | Narrative review (Annals of Medicine) of algal peptides, synthesising in-vitro, animal and limited human studies. On DPP-4 it states 'By inhibiting DPP-IV, algal peptides prolong incretin activity, thereby improving insulin release and glucose homeostasis', citing Chlorella sorokiniana peptides (Val-Trp, Trp-Val, Ile-Trp, Leu-Trp) that 'demonstrated DPP-IV inhibition along with GLUT-4 upregulation', Limnospira platensis LRSELAAWSR at IC50 167.3 ug/mL, Palmaria palmata hydrolysates after simulated digestion, and Caulerpa sp. docking predictions. All rows are in-vitro or in-silico and are secondary reports of primary studies; the review's own conclusion is that 'their clinical translation is limited by variability in peptide characterization, uncertain bioavailability, and lack of robust human trials'. Zero-weighted as a review. |
| Takenoyama S et al 2026 · Anim Sci J | in-vitro | supports | moderate | MIXTURE, not isolated peptides: crude Protamex (90 min) hydrolysates of beef, pork and chicken, plus a commercial chicken meat hydrolysate tablet (CMP), assayed against DPP-IV in vitro. 'The IC50 values for DPP-IV inhibition were 3.2, 3.4, and 3.9 mg/mL for beef, pork, and chicken hydrolysates, respectively', and 3.3 mg/mL for CMP — these are MASS-BASED potencies for undefined mixtures and do not convert to molar (uM/mM) values, so they cannot be compared with purified-peptide IC50s. The hydrolysates 'exhibited approximately 70% DPP-IV inhibitory activity' at a concentration the abstract does not state, and a 'fivefold diluted solution' of the tablet still gave >80% inhibition (relative dilution only, absolute dose not given). No individual peptide is reported as identified, synthesised or separately assayed. The <=3000 Da fraction (65% of the tablet by mass) is described as 'enriched in peptides with potent blood glucose-regulating' activity — an authors' characterisation, not a reported inhibition measurement. In-vitro enzyme assay; no glycemic outcome reported. Co-author affiliated to Marunaka Foods Inc., maker of the tested tablet. |
| Tan J et al 2026 · Foods | in-vitro | supports | moderate | FULL TEXT (PMC12840617) read; LADDER RUNG 1 - MEASURED. Two ISOLATED synthetic peptides from bromelain-hydrolysed goat blood protein were assayed against recombinant human DPP-IV with Gly-Pro-pNA: FPL IC50 321.5 +/- 11.3 microM (MW 375.25 Da = 0.121 mg/mL), competitive (Km 0.2684 -> 0.4196 microM, Vmax unchanged, Ki 709 microM); FPHFDL IC50 465.6 +/- 11.6 microM (MW 744.70 Da = 0.347 mg/mL), non-competitive (Vmax 1.4428 -> 0.4580 dA/h at 400 microM, Ki 186 microM). Purities 99.35% and 98.66%, MW confirmed by LC-MS/MS. POSITIVE CONTROL present: 'Alogliptin at 295 nM was used as the positive control', giving 97.35% inhibition - so the best peptide needs ~1,090x more compound than the control's working concentration, and sits ~16,900x above the ~19 nM sitagliptin benchmark; no alogliptin dose-response was run, so a strict IC50 ratio is not computable. REPLICATES: 'All experiments were repeated three times independently'. DOCKING RANKED THEM BACKWARDS - FPHFDL docked better (-9.2 kcal/mol blind, -33.43 MM-GBSA) than FPL (-7.3, -16.51) yet was the WEAKER inhibitor; the authors write 'this is contrary to the observed IC50 values'. Treat the docking rows as off-scope on outcome. HYDROLYSATE, NOT PEPTIDE: the whole hydrolysate gave 66.23% (5 h) / 68.00% (7 h) inhibition but its assay concentration is nowhere reported, so those percentages are direction-only and must not be read as potency. Only 2 of the 9 peptides the Conclusions call 'potent' were ever synthesised or assayed; the other seven are PeptideRanker/LightGBM scores alone (rung 3). After INFOGEST digestion FPL weakened to 425.5 microM (enzyme-free control 426.0 -> acid/aggregation, not proteolysis) while FPHFDL strengthened to 247.9 microM (enzyme-free control 477.2 -> genuine proteolytic gain). No cellular, animal or human arm: 'in vivo hypoglycemic effects require further experimental validation'. Foods (MDPI), Southwest University Chongqing; funded by Chongqing provincial goat-industry programmes (CQMAITS202513, cqnyncw-kqlhtxm, CYS25174); 'The authors declare they have no competing interests.' |
Disagree, or know a study we missed?
We grade by evidence, not opinions. The way to weigh in is to point us to a study we haven't cited (check the evidence table above first), or to flag a problem with one we have. Every submission is reviewed; if it holds up, the grade updates and shows in Science Changes Its Mind.
Opens a short form. You'll sign in with Google so submissions are tied to a real account — we don't display your identity, and we only accept a link we can verify (PubMed, DOI, ClinicalTrials.gov).
Educational only, not medical advice. Grades and scores reflect published evidence weighted by study design and quality; see the methodology.